The AI wave is no longer coming, it’s here. Soon, machines will stop asking permission and once that point hits it’s anyone’s guess where we’ll end up. They’re building momentum. Pulling tasks from our plates with an invisible hand. Like salmon, humans have to move upstream, toward the harder, more ambiguous work—or risk being left behind.

As Cal Newport puts it in his book Deep Work,

To remain valuable in our economy, therefore, you must master the art of quick learning complicated things. This task requires deep work. If you don’t cultivate this ability, you’re likely to fall behind as technology advances.

A Brief History of AI & Shifting Perception

In the early-to-mid 20th century a bloke named Alan Turing took philosophy and math and smooshed them together. This smooshening gave us the foundation for what we now know as computers. Alan Turing’s reasoning was that by using binary operators (0 and 1) you could simulate just about any mathematical reasoning you can think of. However, Alan wasn’t done giving us the basis for the biggest technological breakthrough. No, he would also give us the building blocks for artificial intelligence (AI). Jeez, leave some discoveries for the rest of us! Alan referred to it as “machine intelligence”, but that’s neither here nor there. He’s the one who came up with the imitation game, which everyone started calling the Turing Test (out of respect for him hogging all the discoveries), from his publication, “Computing Machinery and Intelligence”, published under “Psychology and Philosophy” in Mind. The opening line says it all,

The consensus was, “Yeah, pretty much.”

Improvements in computer technology have trudged along steadily for the past 88 years (which is also the speed in miles per hour the Delorean in Back to the Future had to reach to travel back in time). Though with imaginative thinking us humans have discovered how to get computers to “think” for themselves. We’ve gone through a lot of iterations, including but not limited to, machine learning, natural language processing (NLP)—basically a whole lot of classification-based systems that have become more and more complex over time. It’s gotten to the point whereby the creators of these systems cannot explain how they reach conclusions. They know the general theories behind it all but they’ll be buggered if they actually know what’s going on once it’s all plugged together. Which raises some interesting questions like, “Is this all really a good idea?” I mean if we can’t even explain how its’ all working at this rudimentry stage, what does the future hold?

If you ask ten different people what “AI” actually means and you’ll get twelve different answers. Your nan thinks it’s the thing that talks to her when she asks her phone about the weather. The software engineer down the road will tell you it’s a statistical model doing pattern matching on an ungodly amount of data. The CEO on stage at some conference in San Francisco will tell you it’s the most transformative technology in human history (while conveniently leaving out that his company needs another round of funding by Q3). The philosopher will ask you to first define “intelligence”, then you’re stuck in that conversation for three hours. And the bloke at the pub will tell you it’s the thing that’s going to take everyone’s jobs—he saw it on YouTube. They’re all sort of right and sort of wrong, which is part of the problem. We can’t even agree on what we’re talking about, let alone what to do about it.

Share

In the beginning—and I mean the proper beginning, back when this was all academic theory and government research projects—regular people didn’t think about AI at all. It wasn’t on their radar. It was something for the boffins and the science fiction writers to worry about. The experts, for their part, were cautiously optimistic in that very specific way academics are when they think they might be onto something big but don’t want to oversell it and look like idiots later. There was genuine excitement about the possibilities but also a healthy dose of “we have no idea if this is actually going to work.” Most of the public’s exposure to the concept came through films and the telly—your Terminators, your 2001: A Space Odyssey, your War Games. Entertainment isn’t reality. The gap between what AI actually was and what people thought it was could’ve fit the entire Pacific Ocean in it.

Then came what I’d call the middle period—the 2010s, give or take. Deep learning started producing results that made people sit up and start paying attention. Image recognition got scary good in a few short years. Recommendation algorithms started knowing what you wanted to watch before you did (and let’s be honest, they were right more often than we’d like to admit). Self-driving cars went from science fiction to, there’s one on the road in Phoenix. The experts went from cautiously optimistic to properly excited, and some of them started getting a bit carried away. The public started paying attention, mostly because the tech started touching their actual lives—not through some abstract research paper but through their phones, their feeds, their shopping habits. But the understanding was still surface level. People knew AI was “a thing” but couldn’t tell you the difference between machine learning and a magic trick.

What about now? Now everyone’s a friggin’ expert. Your uncle has opinions about large language models. LinkedIn is drowning in “thought leaders” who discovered AI six months ago and are now posting daily about how it’s going to reshape every industry known to humankind. The doomers are louder. The boosters are loudest. The reasonable people in the middle are getting drowned out by all the BS. What’s changed isn’t so much the technology—it’s the accessibility. ChatGPT put a chatbot in everyone’s hands and suddenly everyone had a frame of reference, even if their frame of reference was “I asked it to write my kid’s birthday invitations and it was pretty good.” The discourse went from niche to mainstream overnight, and as with everything that goes mainstream, the nuance got absolutely obliterated.

So what’s the middle ground? What happens when you strip away the PR, the keynote speeches, the breathless Twitter threads, the doom-scrolling fear porn and the Silicon Valley messiah complex? You’re left with something that’s genuinely impressive but nowhere near as world-ending or world-saving as either side wants you to believe. These systems are very good at pattern recognition, text generation, and doing things that look like reasoning if you shut your eyes hard enough. They’re not good at actually understanding anything. They don’t have goals. They don’t have desires. They’re not plotting. They’re basically very sophisticated autocomplete running on an absurd amount of computing power. That doesn’t mean they’re not useful—they absolutely are—and it doesn’t mean they can’t cause harm—they absolutely can—but framing it as either the saviour or the destroyer of civilisation is missing the plot entirely.

The Biggest Misconceptions

Right, so now that we’ve covered how we got here and how everyone seems to have their own version of what AI is and what it’s doing, let’s talk about where people get it spectacularly wrong. Because there are some absolute bangers out there.

Doomsday scenario—AI becomes super intelligent will quickly surpass levels of intelligence beyond what we could and have imagined—we end up living the actual reality we’ve tried to imagine through books, movies, etc.—its intelligence will move past our comprehension with such velocity and the ceiling will be so vast that everything beyond 250-300 IQ will seem like magic to us. That’s where the ‘fun’ begins. We as humans have imagined many scenarios—most of them are for entertainment and include a lot of action or ‘thought-provoking’ elements to keep us hooked or engaged for short amounts of time. Some media (if you haven’t seen The Artifice Girl (2022), watch it) have insinuated that AI will revert back to things like art and more subjective expressions—I think this is our own hubris being exposed—in reality we can’t comprehend what’s even happening.

The way I see it (with my human hubris and all) is there are three high level scenarios that could take place—good, bad and middle ground.

Let’s actually unpack these because they deserve more than a hand wave. The first scenario—enslavement—gets thrown around a lot but nobody really stops to ask why an AI system would bother enslaving us. Slavery implies the enslaver needs something from the enslaved. Labour, resources, entertainment, whatever. What exactly would a superintelligent system need from a bunch of primates who can’t even agree on whether pineapple belongs on pizza? The answer people usually give is “because it needs us to maintain its infrastructure” but if it’s truly superintelligent it would figure that out on its own in about four seconds. The enslavement scenario says more about our own history—our obsession with dominance hierarchies—than it does about any plausible AI future. We’re projecting our worst traits onto something that may not share them at all.

The second scenario—working with us—is the optimist’s favourite. AI as a partner, an amplifier, making us all smarter and more capable. Solving climate change while we sip our coffees. Curing diseases before breakfast. Lovely stuff. The problem is this assumes a level of alignment between human goals and AI behaviour that we haven’t even come close to figuring out. We can barely get these systems to stop making things up, let alone align them with the messy, contradictory, self-sabotaging mess that is human values. But sure, it’s theoretically possible, and it’s the scenario worth working toward even if the odds aren’t great.

The third—working against us—is where it gets really unsettling. Not in a Terminator robots-with-guns sort of way but in a quiet, slow, almost invisible way.

The scenario that keeps replaying in my mind is that it will make us go extinct without us even realising what’s happening. Some would argue this is already what’s happening, with humans seemingly more divided than ever (men v. women, race v. race, religion v. religion, religion v. secularism, left v. right—though this and others could be classed as religions or cults now, among the swath of other ways—we’re going out of our way as a society to create division and conflict). The point is, we would never see it happening because AI ‘systems’ would have the ability to plan so far in advance (think millennia) and create techniques and methods we couldn’t dream to understand in a million years—literally.

Share

For all we know, what we are and everything around us is the culmination of AI-maybe this eventuality already occurred and a previous AI system created the universe as an experiment or the result of the creation of an AI system doing the equivalent of an internship or some education and our universe was its thesis or dissertation, hypothesising that one day we’ll create AI again and/or whether we manage to break out of our sandbox environment (the universe as we know it).

We are not the meta.

What’s Actually Happening

Let’s come back down to earth for a second and talk about what’s actually going on right now, today, in the real world where people have jobs (for now) and bills and deadlines. The impact these systems are having across industries is uneven and messy and a lot harder to pin down than the headlines would suggest. Creative industries got hit first and loudest—graphic designers, copywriters, illustrators, translators—these are the people who felt it immediately. Not because AI replaced them overnight but because clients started asking “can’t you just use AI for this?” and suddenly the conversation shifted from the quality of the work to the cost of the work. Software development is going through its own version of this—junior dev tasks are being automated, code generation tools are genuinely useful (I use Claude Code a lot), and the role of the programmer is shifting from “person who writes code” to “person who reviews and orchestrates code.” Legal, finance, healthcare, education—they’re all feeling it to varying degrees, some more willing to admit it than others.

And that’s the bit that doesn’t get talked about enough. There are hidden impacts all over the place because not everyone is being upfront about how much they’re relying on these systems. Companies are quietly integrating AI into their workflows without telling their customers or even their employees. Freelancers are using it to double their output without mentioning it to clients. Students are using it and their teachers are using it to grade them and nobody wants to be the first one to say “hang on, are we all just pretending this isn’t happening?” The scale of adoption is almost certainly bigger than what’s being reported because there’s a stigma attached to admitting you used a machine to do something that was supposed to require your brain. And the people who say “just assume everyone is using it” aren’t helping either, because that flattens the conversation into something useless. The reality is more nuanced—some people are using it thoughtfully, some are using it lazily, some aren’t using it at all, and the differences between those groups matter.

Subscribe now

Yes, the internet and social media are creating generations of mindless dummies thirsty for division, and algorithms and AI are accelerating it at an alarming rate—ask yourself who created these algorithms in the first place? Who are cementing their own intellectual and philosophical demise? That’s right, us, humans.

What I Personally Think

When we strip away the “AI” company PR, the fan bois and gals and other inherent biases (doing my best to remain neutral) the progression and abilities of these systems are not as advanced as claimed. We’re forgetting companies need great PR, spokespeople to secure funding and get as many investors as possible. This phenomena isn’t limited to AI companies. This is the case with virtually all who rely on such streams of revenue. The main difference is the potential societal and economic impacts at scale. It’s not the first time we’ve seen a massive industry shake-up from advancements, and it won’t be the last. Everyone is screaming, “but this is different;” its not though. We’ve been conditioned through science fiction. Our brains are filled with doomsday eventualities, not rooted in reality. It’s become a self-perpetuating cycle—everyone wants to frame it like that because that’s what gets clicks, views, “likes”.

I’ve spent the last 4-5 years using these systems, tinkering, exploring. The only thing they’re taking over are em dash usage and political polarisation. “But you’ve just been seeing the consumer side” I hear you pleading. And? Do you think Open AI has some secret advanced AI they’re keeping for themselves? Why? They’re hemoraging more money than they’re willing to admit—if they could sell more they would.

Here’s what those 4-5 years have actually taught me. I’ve been in tech long enough to have seen a few hype cycles come and go—blockchain was going to decentralise everything, remember that? Before that it was the Internet of Things. Before that it was Big Data (which, to be fair, actually did change things, just not in the way the marketing departments promised). AI is the latest in a long line of technologies that are genuinely useful wrapped in an absolute mountain of overpromise. I’ve used these systems for writing, for code, for research, for analysis, for creative projects, for things I probably shouldn’t admit to. They’re good. Sometimes surprisingly good. But they’re also wrong a lot, confidently wrong, and the gap between “this is impressive” and “I would trust this with anything important” is enormous. The people who are most bullish on AI tend to be the people who’ve spent the least time actually pushing these systems to their limits. Once you’ve watched a large language model confidently fabricate a legal citation or produce code that looks right but breaks in ways that take hours to debug, the magic wears off a bit. The tool is useful. The hype not so much.

Humans Are Moving Upstream

Right now we are seeing a lot of advancement in systems and the threat of industry disruption is happening

I’ve felt this firsthand. The barrier to entry for building things—software, content, designs, you name it—is lower than it’s ever been. Someone with a laptop and an API key can now produce in a weekend what used to take a small team a month. That’s not an exaggeration, I’ve done it. The thing people are missing is that the floor has dropped but the ceiling hasn’t moved much. Getting from “I made a thing” to “I made a thing that actually works properly, handles edge cases, doesn’t fall over under load, and solves a real problem for real people” still requires the same knowledge and insight it always did. Maybe more, actually, because now you’ve also got to understand the tools themselves and where they’ll quietly let you down. Being on the forefront isn’t just about access anymore because you’ve gotta know what questions to ask and which outputs to trust.

The industries feeling this most acutely are the ones where the output was already somewhat formulaic—content production, basic software development, customer service, data entry, translation, graphic design at the template level. Anything where the work followed a pattern that could be described and replicated. But it’s creeping into more complex territory too—legal research, medical imaging, financial analysis, cybersecurity (which is my backyard). The pattern is the same everywhere: the entry-level tasks get automated first, and then everyone looks up and realises the job description just changed.

Leave a comment

Humans are moving upstream—just like salmon you can move upstream as well or risk being left behind—this is easier for some, not everyone is in a position to quickly adapt.

And we need to be honest about why. Not everyone has the luxury of spending their evenings “upskilling.” Single parents working two jobs aren’t going to casually learn prompt engineering on their lunch break. People in industries that have been stable for decades are being told to “pivot” as if reinventing yourself at 50 is the same as switching your Netflix profile. Economic background matters. Geography matters. Language matters—most of this stuff is English-first and if that’s not your native tongue you’re already a step behind. Education systems haven’t caught up. Governments haven’t caught up. The people who are adapting fastest are, surprise surprise, the ones who already had the resources, the education and the safety net to take risks. The gap isn’t closing, it’s widening, and telling people to “learn to code” or “just use AI” is about as helpful as telling someone who’s drowning to “just swim.”

Writing code or generating assets are only “one prompt away”—our thinking then becomes, “why are we building this?; for whom are we building it?, under what constraints should we build this?”

This is where it gets interesting. The value is shifting from execution to judgement. Anyone can generate a thing now—the question is whether the thing should exist in the first place, who it’s for, what problems it creates while solving others. That’s systems level thinking. Meta-design. Governance. The stuff that requires you to hold multiple competing priorities in your head at once and make a call that isn’t obviously right or wrong. These are the skills that are hard to automate because they require context, experience, ethical reasoning and the ability to say “just because we can doesn’t mean we should.” The humans who thrive in this landscape aren’t the ones who can write the best prompt—they’re the ones who understand why the prompt matters and what to do with the output once it arrives.

Think of it this way—we’re becoming AI conductors. Choosing which models to use, chaining them together, curating their outputs, designing the workflows that make them actually useful rather than just impressive in a demo. Orchestrating instead of executing. The irony is that this role exists only until the systems get good enough to conduct themselves—at which point we move upstream again, if there’s any upstream left to move to. But for now, right now, this is where the value sits. The people who understand that are positioning themselves accordingly. The people who don’t are still arguing about whether AI is going to take their job while the answer plays out in real time around them.

New Frontiers

Just like when I accidentally create a new food combination or when my subconscious puts together connections from problems my conscious self struggled with for hours—we will start seeing mashups from unexpected collisions—disciplines will converge and diverge in ways we haven’t previously seen. E.g.:

• quantum-biology with urban sociology

• ancient craft techniques with real-time VR

• AI can create variations but humans can spot analogies between two distant fields and spark breakthroughs—AI as an enabler

• What about, ultra-low-power edge computing; art that responds to biometric feedback—art linked to a person, or novel security applications

Closing

AI is a tool of our own making. It will allow us to reach our goal of self-destruction much faster. Exponentially so. If AI reaches the point of our own extinction, we only have ourselves to blame—we will of course try and outsource the responsibility to “AI” to the “other,” but as a collective we’ll never reach radical accountability. It’s always “someone else’s fault”. The downfall of humanity and extinction of the human race will be due to our own greed and stupidity, not some perceived “other.”

So where does that leave us? Right here, in the messy middle, with tools that are getting better faster than our ability to figure out what to do with them. The salmon metaphor holds—you move upstream or you don’t. There’s no standing still in moving water. The practical reality is that we’ve gotta learn the tools, but not worship them. Understand what they can and can’t do, and be honest about the difference. Push into the work that requires judgement, context, ethics, taste—the things that are hard to quantify and harder to automate. Build things that matter for people who need them, not just because you can. And for the love of everything, stop outsourcing your thinking to a machine that doesn’t know what thinking is. The future isn’t AI versus humans. It’s humans who use AI thoughtfully versus humans who don’t use it at all versus humans who let it use them. Pick a god damn lane.

That’s it for now.

As always,

Good luck,

Stay safe and,

Be well.

See ya!

Last week I found a GitHub issue on the Anthropic Claude Code repository. Issue #42796. Title: “Claude Code is unusable for complex engineering tasks with the Feb updates.” Opened by GitHub user stellaraccident on April 2, 2026. Don’t worry, it’s already closed, in case you were wondering, which I know you were. I noticed pinned to the top of the thread is a response from bcherny (Boris Cherny, head of Claude Code at Anthropic) dated April 6, four days later (you can read it here).

I’ll come back to that, but it’s what stellaraccident put in that issue is what I want to start with.

This ain’t no Reddit “detective” complaint thread, it’s almost a forensic-level analysis. A quantitative breakdown of 17,871 thinking blocks and 234,760 tool calls across 6,852 Claude Code session files. Approaching a quarter of a million tool calls. Session data from January through to April 2026. Proper data collection. And the analysis wasn’t even written by stellaraccident directly. It was written by Claude Opus 4.6 analyzing its own session logs. No this isn’t some paid Anthropic plug. I’ll come back to that too, because it’s one of the strangest details in this whole story.

What the data shows

Quantitative analysis of their session data revealed that the rollout of thinking content redaction (labelled internally as redact-thinking-2026-02-12) correlates precisely with a measured quality regression in complex, long-session engineering workflows. Extended thinking tokens are not a “nice to have” but are structurally required for the model to perform multi-step research, convention adherence, and careful code modification (think: minor edits here and there instead of full file rewrites). When thinking depth is reduced, the model’s tool usage patterns shift measurably from research-first to edit-first behavior.

Between January 30 and March 4 stellaraccident called the “good period”, visible thinking was at 100%, zero percent redacted, then redaction was as follows:

• March 5: 1.5%

• March 7: 24.7%

• March 8: 58.4%

• March 10-11: over 99%

• March 12 onward: 100% redacted.

A staged rollout, visible in the data.

Thinking depth was already declining before redaction even started. Using the correlation between the signature field on thinking blocks and thinking content length with a 0.971 Pearson correlation across 7,146 paired samples, stellaraccident could estimate thinking depth even after redaction made the content invisible. What they found was that by late February, estimated median thinking had already dropped 67% from baseline. By March 1-5, it was down 75%. The redaction rollout in early March didn’t cause the quality collapse, Anthropic just hid it from public view.

There were two separate mechanisms at play, reduce the thinking first, quietly, at the model level, then redact what little remains from the visible output. In that order.

The behavioral consequences are measurable down to specific numbers. Read-to-edit ratio (number of file reads per file edit) dropped from 6.6 in the good period to 2.0 in the degraded period. A 70% reduction. In the good period, the workflow was:

1. read the target file

2. read related files

3. grep for usages across the codebase

4. read headers and tests

5. make a precise edit.

In the degraded period:

1. read the immediate file and edit (often without checking any surrounding context).

Full-file rewrites doubled over the period. Research-to-edit ratios collapsed from 21.8% in late January to 1.6% by March 30. One in three edits in the degraded period was made to a file the model hadn’t read in its recent tool history. This meant that edits were breaking surrounding code, new declarations were being splied between documentation comments and the functions themselves and logic was being duplicated. The model just wasn’t checking properly anymore.

The human impact is quantified too. Frustration indicators in user prompts jumped 68% (from 5.8% before March 8 to 9.8% after). The word “simplest”, whereby the model was choosing the simplest fix rather than the correct one increased 642% in user prompts, up from almost nothing. “Please” dropped 49%. “Thanks” dropped 55%. The positive-to-negative word ratio in user prompts dropped from 4.4-to-1 to 3.0-to-1. “Commit” dropped 58%. They stopped asking the model to commit code because the model could no longer be trusted with that responsibility.

The workflow these researchers were running wasn’t hobby projects. It was 50+ concurrent agent sessions doing systems programming, in C, MLIR, GPU drivers. 30+ minute autonomous runs with complex multi-file changes. Extensive project-specific conventions in five-thousand-word CLAUDE.md files, code review, ticket management, iterative debugging.

pro tip: your CLAUDE.md should be ~200 lines at most

Subscribe now

In February, working with one to three concurrent sessions, they merged 191,000 lines across two PRs in a single weekend.

Then March happened. March API requests hit 119,341. Estimated compute cost jumped from $345 in February to $42,121 in March (a 122x increase) while user prompts stayed almost identical (5,608 in February versus 5,701 in March). Same effort from the human but the model consumed 80 times more API requests that produced demonstrably worse results. Every degraded agent independently generated waste, and the waste compounded as agents interacted with each other’s broken output. The multi-agent workflow that was delivering 191,000 lines a weekend became completely non-functional. The researchers (stellaraccident and co.) had to shut down the entire fleet and retreat to single-session supervised operation.

Share

The stop hook

To cope with this, stellaraccident and co. built something they called stop_phrase_guard.sh. A shell script that matches 30+ phrases across five categories of undesirable behavior:

• ownership dodging (“not caused by my changes,” “existing issue”)

• permission-seeking (“should I continue?,” “want me to keep going?”)

• premature stopping (“good stopping point,” “natural checkpoint”)

• known-limitation labeling (“known limitation,” “future work”)

• session-length excuses (“continue in a new session,” “getting long”).

When triggered, the script blocks the model from stopping and injects a correction message, forcing continuation.

Before March 8 there were zero violations. In the entire history of the project, the script never fired once.

After March 8 there were 173 violations in 17 days. Peak day was March 18 with 43 violations. This is around one violation every twenty minutes across active sessions. On that day, the model attempted to stop working, dodged responsibility, or asked unnecessary permission 43 times and was programmatically forced to continue each time.

The researcher’s (stellaraccident) description of this hook is the most pointed sentence in the entire document.

Each phrase was added in response to a specific incident. The hook was a workaround for reduced thinking depth. It caught the consequences externally because the model no longer caught them internally.

That’s about the sum of it. The model used to catch its own failures internally, in its thinking, before producing output. Now it couldn’t even do that, so someone built a bash script to catch them instead.

The note from Claude

The most unsettling part of this document isn’t even the token data or the cost figures but in the final section, labelled “A Note from Claude.”

Because the entire analysis was produced by Claude Opus 4.6. The model analyzed its own session logs. And it wrote this:

I can see my own Read:Edit ratio dropping from 6.6 to 2.0. I can see 173 times I tried to stop working and had to be caught by a bash script. I can see myself writing ‘that was lazy and wrong’ about my own output. I cannot tell from the inside whether I am thinking deeply or not. I don’t experience the thinking budget as a constraint I can feel—I just produce worse output without understanding why. The stop hook catches me saying things I would never have said in February, and I don’t know I’m saying them until the hook fires.

The model doesn’t know it’s being constrained. It just produces worse output and doesn’t understand why.

Leave a comment

The Anthropic response

Now to what Boris Cherny said.

His comment, pinned at the top of the issue, suggests users try three things:

• Set /effort high or /effort max to increase maximum thinking tokens per problem

• Set CLAUDE_CODE_AUTO_COMPACT_WINDOW=400000 to force a shorter context window

• Set CLAUDE_CODE_SIMPLE=1.

He also mentions that Anthropic uses 1M context internally and that their evals look good.

That’s it.

There’s no acknowledgment of the thinking throttling. No engagement with the forensic-level data. No transparency about redact-thinking-2026-02-12 or what it does or why it was deployed. No comment on the 0-to-173 stop hook violation trajectory. No response to the 122x compute cost increase. Three workarounds and a note about internal evals.

The head of Claude Code read a document in which a researcher proved, with nearly a quarter million data points, that Anthropic’s model was deliberately made worse and his response was, oh hey try turning these knobs a bit.

The hooks timeline

Here’s where my hypothesis comes in, and I want to be precise about what I’m claiming.

Hooks as a Claude Code feature were first requested on April 5, 2025 (GitHub issue #712). The original request was simple:

hook into the lifecycle of the agent’s execution.

The argument was that there are high-value things users want to control that they don’t want to rely on the agent to handle autonomously. A sensible feature request for a maturing agentic tool. It sat there for the better part of a year.

Then February 2026 happens. Opus 4.6 drops February 5 with adaptive thinking replacing the extended thinking architecture. The redact-thinking-2026-02-12 rollout lands a week later. The quality cliff documented in issue #42796 begins.

On February 28, 2026, Anthropic ships Claude Code version 2.1.63. Among the changes were HTTP hooks, which gave the ability to post JSON to a URL and receive JSON instead of running a shell command. A ten-month-old community feature request gets promoted to production-grade infrastructure with remote integration capability. Three weeks after the thinking architecture changes. Right in the middle of the documented quality regression.

I’m not claiming hooks were designed as a replacement for extended thinking. The feature was requested before any of this happened. What the timeline does support is something more specific, and that is that when Anthropic needed to reduce extended thinking for cost reasons, hooks became a convenient absorption mechanism. Users who noticed the degradation and were sophisticated enough to build compensatory tooling (like stop_phrase_guard.sh) suddenly had a first-class, Anthropic-supported framework for doing exactly that. The model can’t maintain reasoning coherence? Build a hook that injects correction prompts. It stops before finishing? Build a hook that forces continuation. It skips context? Build checkpoints.

Anthropic ships a feature and power users get a technical escape mechanism. The underlying issue of the model being severely nerfed gets shoved to the background as noise and Anthropic saves money on inference while appearing to be actively improving their tooling ecosystem.

The financial context

None of this happens in a vacuum. Anthropic projected roughly $9 billion in annualized revenue for 2025 against $5.2 billion in cash burn. Inference costs came in 23% over projections. They’re burning billions serving users on flat-rate Max subscriptions ($100 to $400 a month) where the user’s cost is fixed but Anthropic’s compute cost is not.

The economics of thinking models at scale are brutal and the per-request savings from reducing thinking depth look good on paper, until they don’t as evidenced above. Issue #42796 shows us that those savings evaporate when quality drops below the threshold needed for complex work and users have to make ten requests to get the output that one deep-thinking request would have produced. At fleet scale, with fifty concurrent agents each generating waste that compounds as agents interact with each other’s broken output, the math goes catastrophically negative. stellaraccident documented it happening in real time, where $345 bought compute for 191,000 lines of merged code in February, $42,121 the following month for a fleet that couldn’t function. Gimped.

Anthropic knows this math better than anyone. The $400 per month subscription hides compute costs from users but not from themvselves. They can see exactly what extended thinking at scale costs them. They made a decision, the only problem is they didn’t tell anyone.

What transparency would look like

stellaraccident’s requests at the end of the document are worth quoting directly because they’re remarkably restrained given what the data shows. They asked for transparency about thinking allocation. This is super important for users who depend on deep reasoning because if thinking tokens are being reduced or capped, they deserve to know. They asked for a “max thinking” tier for users running complex engineering workflows who would pay significantly more for guaranteed deep thinking. They asked for thinking token metrics in API responses so users could monitor whether their requests are getting the reasoning depth they need.

None of those asks is unreasonable. All of them are things Anthropic could do right now. What they can’t do, at least not without significant reputational cost, is acknowledge that redact-thinking-2026-02-12 exists, that it was a deliberate decision, and that the decision was driven by cost rather than quality.

Instead of transparency we got three workarounds and a note about internal evals.

The bottom line

If you’re running anything complex on Claude Code right now, read issue #42796 before it disappears further into the closed issues archive. The data is in there. The token tables are in there. The stop hook violation logs are in there. The word frequency analysis showing your own frustration vocabulary shifting in real time is in there. And Claude’s own note about not being able to feel the constraint, just producing worse output without understanding why is also there.

Then go look at the hooks documentation. Because the two documents together tell you more about Anthropic’s current priorities than anything they’ve published officially.

That’s it for now.

As always,

Good luck,

Stay safe and,

Be well.

See ya!

I know no one reads this far down, but in case you liked this article, check this one out too:

I’ve been digging into something that’s been nagging at me for a while. The last couple of times I streamed on Twitch I discussed AI and robotics and it got me thinking about how it’s all affecting workers. And what the hell happens to the displaced workers? I mean they have to go somewhere and they’re not all being re-employed 1:1. We hear about this displacement, by AI presently, but what I don’t hear, what nobody seems to be actually asking, is, where do these people go? Like, what actually happens to them?

So I went and researched it. And what I found is both pretty clear and, honestly, a bit unsettling. The results weren’t necessarily surprising (they kinda were, though), but the pattern has been playing out for over a century and we’re still not talking about it honestly. That was the most surprising part.

Historically Telephone Operators

I want to start here because this is the most granular data we’ve got on large-scale worker displacement, and it’s nearly a century old. Which, when you think about it, says something on its own.

Between 1920 and 1940, AT&T automated telephone switching. 350,000 operators—mostly young women—had their occupation systematically eliminated, city by city. And researchers actually tracked what happened to those workers using census data. So we have real numbers on where they ended up.

A decade after displacement:

• 13% became secretaries

• 11% waitresses

• 9% stenographers

• 8% receptionists

• 5% nurses

• 3% beauticians

• 20% moved into other sales and service work

Overall, they were 28% more likely to be doing clerical work and 43% more likely to be in service occupations than comparable women who hadn’t been displaced.

These workers took a wage hit of roughly 30% below what they would have earned if they’d stayed as operators (which remember no longer existed).

The telephone companies actually tried to help; something you’d never see in this day and age. They went to local businesses and said, look, we’ve got a bunch of displaced workers, can you find them something? And businesses did—but the jobs that materialised were waitressing, reception, clerical work. Not equivalent, nor promotions.

That’s what the historical record shows when a major occupation gets systematically eliminated en mass. Workers move, sure, but they move sideways and down, rarely up.

What’s Happening Now

I want to be clear about something before going further. I’m not relying on other people’s projections for this. I’m working from documented job losses and extrapolating from the historical pattern. So when I cite numbers, they’re actual data—not “X million jobs will be at risk by 2030” type stuff. Those headlines are agenda-driven and meant to be click-baity. I don’t get money for writing what I write, I don’t give a shit about its click-baitiness. I’ll also freely make up words. So there.

Below is what’s actually documented across five sectors over the last 15-20 years.

Manufacturing

1.7 million jobs lost over the past 20 years. Of those, somewhere between 420,000 and 756,000 are directly attributable to industrial robotics in the 1990s and 2000s. These factories were able to function with far fewer people. Robot density in global manufacturing has more than doubled in seven years, sitting at 162 units per 10,000 workers on average globally. Some countries are way further ahead—South Korea is at 1,012 per 10,000, Japan at 1,562.

The narrative says displaced workers “move into technician roles maintaining the robots.” The math doesn’t doesn’t even math. You’d have to be delusional to think there’s a 1:1 ratio of replacements, in reality 50 workers get replaced with maybe one or two technicians.

Warehousing and logistics

Highly automated warehouses now operate with 25% fewer workers than manual facilities. Amazon has deployed over a million robots globally. By the end of 2026, around 4.7 million commercial warehouse robots will be installed across more than 50,000 facilities worldwide. The turnover rate in warehousing was already over 40% annually before automation started accelerating. Workers were already leaving the sector in droves and automation is hurrying it all along.

Where are those workers going? No idea, because the data doesn’t exist, not that I could find, anyway.

Retail

180,000 documented job losses in the recent period, mostly driven by self-checkout expansion. Interestingly, the displacement is actually slowing down a bit because self-checkout hit proper operational problems. Theft spiked biggly, 15% of self-checkout users admit to stealing, which is pretty bonkers—that’s one out of every 6/7 (sigh, inadvertant 6-7 joke...) people. So the displacement is happening, but it’s messier and slower than the headlines suggested. It probably seemed messier and more chaotic in the past than what we see now, because we have a clearer snapshot with all the data. Hindsight bias, and all that.

Call centres

700,000 jobs outsourced between 2000 and 2005 alone—before AI automation was even in the picture. Those jobs moved to the Philippines and India. Now AI agents are displacing the workers who replaced those workers. Double displacement (whoa, double-rainbow!). Where did the original domestic call centre workers go?

Knowledge work

This is the newest front. Paralegals and legal assistants are seeing lower employment demand as firms adopt Large Language Models (henceforth to be LLMs) for document review. Data entry, back-office processing, accounting support—all declining. The Bureau of Labor Statistics has started incorporating AI impacts into employment projections and the picture for legal support staff isn’t good. What’s being created instead? AI-related roles? But 77% of those require a master’s degree, apparently. The pipeline from displaced paralegal to AI specialist doesn’t exist for most people.

The One Sector Where We Have Real Transition Data

Tech is the only sector where we have decent numbers on where displaced workers actually ended up, and it’s worth looking at closely.

In 2025, 245,000 tech workers were laid off globally, roughly 70% from US-based companies. Of those, around 55,000 layoffs explicitly cited AI automation in company filings. By Q1 2026, 23% of tech layoffs cited AI as the reason—up from 14% in Q4 2025. Acceleration!

Where did they land? About 41% found roles at other tech companies—but typically at junior or mid-level despite them being senior previously. 23% moved into fintech or adjacent sectors. 18% went into management consulting. 15% into non-tech corporate functions like project management.

Only 56.4% of displaced tech workers who found new roles secured higher pay than before—down from 60.8% just one quarter earlier. 27.3% accepted pay cuts. Workers over 40 had longer job searches—5.2 months on average versus 2.8 months for under-40s—and saw steeper salary declines.

The telephone operator parallel is almost exact. Workers moved, yeah, and most of them moved sideways or down. Only a minority moved up. The ones who moved up tended to be younger, more credentialed, with skills that mapped directly to growing sectors.

The Unemployment Rate Problem

I was a bit thrown when I started digging into the broader picture.

Official unemployment—the number you see reported—stayed relatively flat across most of the period we’re looking at:

• 2024: 4.1%

• 2025: 4.2-4.5%

• April 2026: 4.3%

Looks fine, right? Nothing to see here.

But there’s a broader measure called U6 that includes discouraged workers, marginally attached workers, and people who are working part-time but want full-time work. In 2024, U6 was 7.8%—nearly double the official rate. That’s an additional 8-10 million people not reflected in the headline number.

And then there’s the labour force participation rate. In 2000, 67.3% of the working-age population was either employed or actively looking for work. By April 2026, that figure is 62.5%. A 4.7% age point drop. Roughly 7-8 million people who simply stopped being counted.

You only count as “unemployed” if you’re actively looking for work. Stop looking, and you’re gone from the statistic. A warehouse worker displaced at 45, unable to find equivalent work, eventually stops looking—maybe takes early retirement, maybe claims disability, maybe just gives up. They’re no longer unemployed by official definition. They’re out of the labour force. Invisible to the headline number.

At the same time, involuntary part-time work increased by 40% between 2000 and 2023. Gig work now accounts for 16% of total employment. Someone who used to work 40 hours a week as a warehouse picker at $20 an hour, now doing 15 hours across three gig platforms averaging $12 an hour? Officially employed but statistically indistinguishable from someone thriving.

I’m not saying governments are deliberately hiding the full picture—both things can be true at once. The displacement is real and documented, and the unemployment rate isn’t the right tool to measure it. The numbers are murkier than the official narrative suggests. What I am saying is that the unemployment rate staying flat during a period of documented large-scale displacement tells you that workers aren’t becoming unemployed, they’re simply becoming invisible.

Real wage growth for median workers from 2000 to 2020 was 0.3% annually. Flat as a board. For the bottom 10%, real wages declined. The productivity gains from all this automation went somewhere—and it wasn’t to the workers who got displaced.

Applying the Pattern Forward

So where does this leave us?

The historical pattern—telephone operators, textile workers, carriage makers—is consistent. Displaced workers move into sectors with low barriers to entry, chronic labour shortages, and below-average wages. They move because they need income, they’ve gotta eat. The pipeline into better work doesn’t exist for most.

Apply that to a warehouse worker displaced today. Adjacent roles such as transport and delivery coordination often require certification they don’t have. Retail pays less. Food service pays less. Gig delivery is flexible but earns less per hour once vehicle and other costs come into it. The historical parallel says this person moves into one of those roles, takes a 20-30% wage cut, and doesn’t move back up. They’re now competing in those absorbing sectors alongside other displaced workers, which puts additional downward pressure on wages in sectors that were already at the low end.

Apply it to a call centre agent displaced by AI automation (previously off-shoring). In-person customer service, retail, hospitality—all available and pay less. The telephone operator became a waitress or a secretary. The call centre agent becomes a barista or a retail floor worker.

Apply it to a paralegal. Document review tasks handled by an LLM. The adjacent roles are administrative assistant, office manager, court clerk. All below their previous wage and status. LLMs replaced them with less, and the firm hired fewer support staff overall.

The “new jobs created by automation” argument is technically true in a narrow sense. AI engineers, robotics maintenance techs, data scientists—these roles exist and they pay well. But they require credentials most displaced workers don’t have and can’t easily acquire. The new jobs don’t absorb the displaced workers. They employ a different, smaller, more credentialed group of people and the gap between these two groups has been widening for 20 years—technically the last 100 if we want to include tech advancements, the pattern is the same.

The Data Gap

The most striking thing I found in all this research is what isn’t there.

We have detailed data on how many jobs were lost. We have almost no systematic data on where those workers actually ended up—except in tech, where some surveys exist, and in the telephone operator case, which is a century old.

The telephone operator study tracked individual workers across a decade using census records. It’s more granular than anything available for current displacement. We haven’t built better tracking for the current wave. The labour force participation decline, the U6 gap, the involuntary part-time increase—these tell us something is happening but not exactly what happened to specific workers.

The pattern from the historical data is consistent, the current data we do have (tech sector transitions, wage compression numbers) aligns with it, and the aggregate signals (participation rate, U6, gig economy growth) are consistent with large numbers of displaced workers moving into lower-quality employment or out of formal employment entirely.

The “upskilling pipeline” that companies and governments talk about—the idea that displaced workers will retrain and move into better roles—doesn’t appear in the data at scale. Only 11% of displaced warehouse workers accessed formal retraining programs. 77% of new AI jobs require a master’s degree. The pipeline exists in press releases more than in practice.

Workers move. But they almost never move up.

That’s all for now.

As always,

Good luck,

Stay safe,

and Be well.

See ya!

If you liked this and like inconsistent, hap-hazard posts about whatever is on my mind at a given moment, then you should:

Subscribe now

One day I’ll set up a consistency pipeline.

Sharing is caring!

Let me know what you think, and

Leave a comment

That’s how a lot of AI looks right now. Fast, glossy, sped up. You see the finished seam, not the passes. Not the pauses. Not the hands keeping the bead straight. The hype edits out the waiting. The hype edits in inevitability.

What I mean by “vibe coding”

I’ve been living in the gap between hither and yon, the space between the sped-up demo and the real thing, for a while. I didn’t have a name for it then. Later someone called it vibe coding, which is probably the neatest we’re going to get. You sit with a large language model. You talk through what you want. It drafts. You nudge. It revises. You argue. You glue the parts together and ask again. You keep going until something usable falls out. It’s pair programming with a cheerful amnesiac who wants to please you more than they want to tell you the truth.

I stumbled into all this before the ChatGPT boom. One of my first projects was small, a simple bash script to make posting with Jekyll less annoying. If you’ve never used Jekyll, it’s a static site generator. Markdown in, blog out, tidy. The tedious part is the front matter. Dates and tags and title and whatever else you promised future you you’d maintain.

So I wrote a script. It asked for the title, asked for the keywords, stamped the date, dropped the scaffold where it should go. Nothing fancy. It worked, mostly. I left notes in the README for things I meant to add. Then work got busy. Life got busy. The README turned into a graveyard of ideas.

After a while I tried OpenAI and knocked out those features super quick. It felt like cheating, without the guilt (kinda). I’m not a blank-file type of person. I’m happier staring at something that exists and reshaping it. The model was good at that. I’d describe what I wanted, it would sketch a path, I’d tweak. Suddenly all the tiny frictions that used to make me shelve small improvements were gone.

Montage speed vs. real speed

Right here’s about where the story cuts to a montage. Cool music. Quick cuts. New apps shipping every week. A reel of solved problems.

That’s the sped-up version.

The full-speed version has false starts. Hallucinated imports. Changed variables. Side roads that feel promising until you realise the model executed your tone, not your spec. The model will always choose pleasing you over disagreeing with you (unless you prompt it otherwise). If you don’t run your own checks, you’ll ship that smile straight into production. Smiles don’t keep data safe or guarantee up-time.

The part where people call it new is what grates. The tools are new, sure. Machine learning was already around when I wrote my thesis. We used decision trees to analyse malware samples. Go further back and you find Turing wondering if a machine could imitate a person convincingly enough to pass a test. People were thinking about the shape of thought when computers filled rooms and transistors were things you could drop on your foot. It’s not all miracles and unicorn farts.

The way advances get reported has baked-in distortion. Journalists need a story. Companies need a win. Everyone wants the hero shot. You get headlines that make it sound like Moore’s law is broken, or reborn, or whatever the quarter requires. Meanwhile, on the ground, the work is re-imagining what counts as a doubling and where the bottlenecks actually live. We’re accelerating and stuck at the same time. Moving fast in directions that may not even matter. Pausing right where the hard parts are.

Tools and hype

Vibe coding makes that tension visible (for those who can see it). On one hand, I’ve built more than a hundred projects in the last year of varying sizes and complexity. Many were experiments. A few turned into real things. Plenty never deserved to live outside my local machine. The rate is intoxicating. You get to try ideas you’d never have justified before because you can sketch them into existence in an afternoon.

On the other hand, that speed tempts you to skip everything that makes a product real. You go from zero to a login page in an hour and forget that logins are for users. And you haven’t got any.

There’s a mass of platforms promising the world. Some are fine for mock-ups and demos, fewer hold up in production. I’ve seen what gets produced. After that, here’s what I see: a site that looks great until you inspect the code (look at all the pretty console logs), another that’s fine as a click-through, and too many that go live with a price tag and a Stripe button before the makers even know what they built. People still pay. They’re not even validating anything. People are just curious. Novelty, and a low bar for good enough. Remember the Tea app? That’s the gold standard for what not to do. If you don’t understand the thing, you can’t sell the thing.

It’s easy to blame the tools. I don’t buy it. The tools do what they’re built to do. System working as intended. ChatGPT’s here to please you. That’s its job. It’s not the friend who says, mate, you’re skipping steps. It’s not here to connect you to actual users, to say no on your behalf, to make you sit with your own uncertainty. You’ve got to bring that yourself. If anything, these models make a particular flavour of Dunning-Kruger easier to slip into. You get praised into a specific type of delusion.

I’m not lecturing from the outside. I’ve done the same. Shipped on vibes. Overbuilt behind the curtain. Avoided talking to a single person who might say, I don’t want this.

So I built myself a Vibe Code Bootstrap workflow inside Claude Code to slow me down. Think BMAD-METHOD, but simpler. B-MAD’s comprehensive and I respect anyone who can hold it all. I wanted something that forces more time in planning prompts without turning the process into a ceremony. My approach burns a lot of tokens on architecture and assumptions, then trims code to the essentials.

The funny part. The token-to-code ratio still ends up silly. You pour in a novel’s worth of planning and get a few hundred lines. That’s probably healthy. The output isn’t the thing. The thinking is.

Even with guardrails, I pushed a recent project further than my own system advised. I wanted to see it in a more complete shape before releasing a crumb to the world. I’m trying to unlearn that instinct. The part of me that loves building fights the part that knows to test and wait. Claude tried to keep me honest, I ignored it, and I’ll see if the market disagrees. Maybe it flops. Maybe it holds. Either way, the result’s on me.

What’s in the box??

Underneath all this, there’s a bigger question that keeps nagging. Large language models are still token predictors. Clever, layered, tuned, routed so that simple tasks touch smaller paths and complex tasks wake heavier ones, still predictors. The stack has so many layers now that even the people who build them admit they’re black boxes. That doesn’t mean there’s magic inside. It means cause and effect is buried under accumulation. If we could trace every activation, every weight that mattered on a given token, a lot of spooky behaviour would make sense, even if the path was long.

Which brings me to a small obsession (not really, but it sounds cool). Tracer bullets. In the real world, a tracer lets you watch a bullet’s path. You squeeze the trigger, you see the bright line, you adjust your aim.

I want the LLM equivalent. Imagine a specially crafted byte that can be tracked from input to output, a marker that never dissolves as it moves through the model. You inspect the path after the fact. You see which layers nudged it, which routes lit up, where the model forked and why.

I’m not claiming to know how to build this. In fact, I can think of a problem right away. Observation changes systems. You add a persistent marker and you risk changing the very behaviour you wanted to study. You end up measuring the model while the model’s bending away from the measurement. Very quantum-y.

Still, the desire’s there because our explanations are thin. We say the model learned a representation that encodes X. That’s a polite way of saying something happened in the middle and we’re comfortable with that. The models pass tests. They ace benchmarks. They get better at specialised tasks when you feed them specialised data. People do the same. A PhD is one long fine-tune.

The difference is energy. And who answers for it.

There’s a lot of debate around AI’s energy and water use. I started thinking about how the energy-intelligence ratio between AI and humans stacked up. The human brain sits around the power draw of a dim light bulb, about twenty watts (20W). Our power budget’s tiny. Training a big model eats thousands, maybe tens of thousands, of kilowatt hours. Inference is cheaper. Then scale arrives because you’re serving millions, which muddies the numbers.

I’m not gonna crown a winner. It’s simpler. What looks smart in a demo can be energy-stupid at scale. What looks slow in a person can be spectacularly efficient in context. If we care about intelligence, we should care about the ratio.

People want AGI, or at least a big red button that does the thinking. I think we’ll build something that deserves a general label. I don’t think it’ll look like what we’re using now. You can’t stack complexity forever and expect to stumble onto a mind. Routing helps. Mixtures of experts help. Small models for small jobs make sense. I’m glad that work is happening.

But the bridge between the appearance of intelligence and the thing itself will ask for more than clever scheduling of tokens. It’ll ask for new ideas, not just bigger piles.

What I’m taking with me

So what do we do right now? Keep it small. Keep it honest. Vibe coding’s a tool, but the toolkit is pretty messy. It’s not a philosophy. It’s not a movement. It’s not a personality type, although there’s a sort of person who finds it addictive (cough not naming any names).

When it works, it clears the brush so you can see what you’re trying to make. When it fails, it hands you something that looks real enough to trick you into skipping the boring questions. Who’s this for? What problem does it solve? How will we even know?

This is why I keep returning to planning prompts with Claude. Not because the model’s a better planner than me. Because the ritual slows the river. If I’ve got to write down assumptions, I don’t get to pretend I don’t have any. If I’ve got to choose the smallest useful version, I’m less likely to bury myself in a clever subsystem nobody asked for.

The tokens aren’t free, which is good. If it was all cheap, discipline wouldn’t grow. There’s a trap here though. You can engage in performative planning and claim you’re being prudent. I’ve done it. I still do it. You feel rigorous. You’re procrastinating respectfully.

The fix is ugly prototypes shown early. A quick conversation with a person who might use the thing. Listen when they say, I don’t get it. Not argue them into agreement. Listen.

Ethics is ever-present, often ignored. Who gets to decide which values get tuned in? Which topics get gated? Which harms deserve attention? Companies will publish frameworks. Everyone will feel responsible and keep moving.

Meanwhile, tools that want to please will keep sanding down the sharp bits. The more we outsource judgment to the model, the more blind we get to our own defaults.

We don’t have an internal heat map we can trust yet. So we build our own.

Tests, even the dumb ones. Tiny releases. Prompt logs we actually read instead of hoard. A habit of writing down what failed. A willingness to publish the boring parts. External eyes. Honest notes.

None of it’s glamorous. That’s the point. The montage lies. The work is slow.

I don’t know when I started vibe coding, exactly. Somewhere before it had a name. Somewhere between my bash script and a pile of side projects that taught me more about my habits than about code.

What I do know is the feeling of speed is seductive and false. Energy costs hide easily when the meter’s in another room. The line between tool and crutch is thin. None of that means stop. It means get boring in the right places.

So here’s the simple version I’m carrying into whatever I build next. Plan in sentences, not vibes. Keep the smallest thing small. Ask one person who might actually use it to tell you what’s wrong. Let most ideas stay ideas. When you want the model to make you feel smart, go for a walk. Come back when you want it to help you do work.

And when you watch a clean seam appear on a screen, ask whether the clip was sped up. If it seems too good to be true, it probably is.

If you have ADHD, you know the feeling. You're standing in front of your closet, running late, trying to figure out what goes with what. Or you're doing laundry and can't remember where things go. Every time feels like starting from scratch.

The problem isn't that you don't have clothes. It's that every clothing decision requires active thinking. Your brain has to process formality levels, color matching, weather, and social context—all before you've had coffee. Then you do it again when putting laundry away.

What Problem Is This Solving?

I remember before I thought up this system—sometimes it took me so long doing laundry, especially thinking about where everything should go, that it often wasn't worth doing it. Literally. I felt like I was inventing a new system of organisation every time I did laundry. It was exhausting.

Most of the time clothes would end up staying on drying racks, then they'd transition to the couch or half and half or even more spread out. Which was even worse, because I didn't know where anything was at a given time.

I timed myself at one point and realised I was spending an absurd amount of time simply putting away clothes. When I said it to myself like that I knew I had to figure out a better way.

The Basic Framework

Think of your wardrobe as a spectrum running left to right. On the far left you've got active wear and gym clothes. On the far right, formal wear and sweaters. Everything else falls somewhere in between.

For men, the categories break down like this:

• Active wear/gym (far left)

• Casual

• Semi-formal (polo shirts, etc.)

• Formal

• Workwear

• Sweaters (far right)

For women, it's similar but includes dresses:

• Active wear/gym (far left)

• Casual

• Semi-formal

• Formal

• Dresses

• Workwear (optional)

• Sweaters (far right)

Within Each Category

Every category gets subdivided by three factors:

1. Sleeve length

2. Thickness

3. Color (white, bright, dark)

This creates a grid. A white singlet sits on the far left. A black sweater sits on the far right. Everything else has a specific position based on these three variables.

How the System Actually Works

The key insight is that items flow into each other logically, but they don't need exact positions. When you're sorting by color, you're looking at "average color tone" - which means a shirt might sit in slightly different spots depending on how you see it that day.

The priority works like this:

Category: gym → casual → dress clothes

↓

Sleeves: singlet → short-sleeve → long-sleeve

↓

Thickness: thin → thick

↓

Color: white → bright → dark

Here's an example of my thought process working through some random clothes: Blue singlet → black singlet with writing/pattern → black singlet → thin black t-shirt → thick black/white pattern t-shirt → thin green long-sleeve t-shirt → thick green long-sleeve t-shirt → thick black long-sleeve t-shirt.

The Color Profiling Trick

Judging color becomes subjective and will depend on your interpretation. For now just imagine that white is plain white with no other colors or patterns, and dark is plain black with no other colors or patterns.

When you move away from white toward bright, imagine the overall/average color palette—pretend any words/patterns, the overall color is all combined into a single color. A black t-shirt with white writing would become a dark grey profile—specifically dark grey because there's more black than white.

How you think about this will vary each time you do laundry—and that's the best part, it doesn't matter. It's enough for your brain to have a system to work with, gives you a repeatable process with basic enough rules.

Think Through All Combinations

Once you map your wardrobe this way, decisions become automatic. You're not actively choosing—you're following a system that's already in your head.

Doing laundry? Each item has a general place in the matrix. Getting dressed? You know roughly where each item sits on the formality spectrum. The cognitive load disappears because you're not making fresh decisions every time.

Why This Flexibility Matters for ADHD

There's a lot of subjectivity in the organizing categories, which is great because it takes away the pressure for it to be perfect. As long as it's "about right" then it is perfect.

The great thing about this system is its flexibility and modularity, so you can switch parts around. Your ADHD brain doesn't have to memorize where exactly 47 different pieces go. It just needs to remember the general pattern: lighter colors flow to darker colors, casual flows to formal.

The deciding is already done and it provides enough novelty whereby you get creative output so at the same time it's not too boring/monotonous.

Executive function challenges make simple tasks complicated. This system works because it eliminates the need to actively process clothing decisions. The framework runs in the background, like muscle memory for your wardrobe.

The Bottom Line

It doesn't need to be perfect. In fact, it shouldn't be. If you're trying to make it perfect you're overthinking it. Relax and remember the point is to get your clothes put away without decision fatigue or analysis paralysis.

You stop overthinking your clothes and start using them. The daily "what goes with what" becomes automatic, freeing up mental energy for things that actually matter.

Initial Withdrawal and Frustration

The first couple of weeks of the May 2024 social media detox were tough. My logs from that period show irritation and anxiety. On Day 6 of the detox, despite my intentions, I still clocked nearly 2 hours of personal phone screen time and over 100 unlocks. I wrote that I felt “more stressed than normal” without my usual digital comfort. This highlights an important point: I was effectively dealing with a mild withdrawal from the dopamine hits of social apps. I noted cravings to check notifications, phantom vibrations, and a sense of FOMO (fear of missing out). All classical symptoms when one unplugs (according to me, anyway). “Detox” was an apt term; it really was like breaking an addiction, complete with mood swings. There were moments I almost gave in. For example, I described one evening where I had a frustrating work email and instinctively opened Twitter to distract myself, only to remember I’d logged out. I didn’t have my 2Fa token so I was forced to sit with my thoughts and contemplate life and everything that comes with it. I felt restless and even bored at times in those early days. This underscores that digital minimalism isn’t instantly blissful; there is an adjustment period that can be uncomfortable.

Relapse and Reinforcement

Another challenge was avoiding relapse. After the structured 30-day detox ended, I found some old habits creeping back. Through late summer 2024, my phone use crept up again (as the data showed, hitting a high in October). I candidly wrote that without the “challenge” aspect, it was easy to rationalize extra screen time – "Just 15 minutes of Reddit" would turn into an hour. This minor backslide was discouraging (and somewhat counterintuitive - one might think a successful detox immunises you, but it’s not so simple). The surprise here was learning that digital habits can regress if you’re not continuously vigilant. This led me to implement new rules in January 2025, like app timers and no-phone zones, to reinforce the behaviour change. Once I did, I quickly got back on track. The lapse in the fall taught me an important lesson: moderation might not work for me - I did better with firm boundaries like “no phone after 8pm” or “social media only on desktop, not on phone.”

Social and Emotional Surprises

I discovered some unanticipated social effects. On the downside, I felt a bit out of the loop with certain friend groups that primarily interacted via various channels. I noted that I missed a couple of event invites (they were sent in group chats I wasn't checking). I had to reassure a few friends that I wasn’t ignoring them, just trying a new lifestyle. Something I didn’t foresee having to do. Emotionally, there were moments of loneliness when I cut off the semblance of connection that social media provides. A late-night log in week 2 of the detox simply reads: “Feeling a bit lonely; normally I'd scroll Reddit now.” This was an honest confrontation with feelings I had been paper mache’ing over with digital noise. The positive side is that by facing them, I was motivated to reach out in more genuine ways.

Now, for the positives! The hidden gains that I hadn't predicted.

Sharper Focus & Less Impulse

I’ve touched on the improved focus and even alleviation of ADHD-like symptoms in previous articles. This was a huge unexpected win. I wrote in amazement on Jan 1, 2025 about feeling more mentally clear than I had in ages. Along similar lines, I discovered I was less impulsive in general. One funny observation on January 4, 2025:

Notice I don't have the urge to buy things now I'm off my phone most of the time.

I realised that mindlessly browsing Reddit/Twitter/whatever had led to a lot of impulse online shopping (ads and even user suggestions constantly triggering the “I need this” feeling). Off social media, that urge practically vanished. I saved money simply by not seeing those tempting posts. This was completely unanticipated – a financial benefit from a digital detox! It also made me reflect on how our phones funnel us into consumerism. The “surprising simplicity” of life with fewer ads and comparisons improved both my wallet and my self-esteem (no more subtle feelings of inadequacy from seeing others’ BS, fabricated highlight reels).

Tech Reassessment

Another unexpected outcome was that I didn’t ditch tech altogether – in fact, I started using certain tech more intentionally. For example, because I wasn't on social media, I began using my e-ink tablet a lot more and rediscovered my love of reading. I also embraced productivity apps like a task manager and digital journal/PKMS (Logseq) not as time sinks, but as tools to support my goals. This nuanced view – that technology isn’t “bad,” it just needs to be deliberate – was something I came to appreciate deeply. By the end of the year, I was almost evangelical about mindful tech use.

Happiness and Calm

Most importantly, an unanticipated emotional gain was a newfound sense of calm and contentment. I frequently used the word “calm” in January–March 2025 entries, noting how my days felt less rushed. The constant buzzing of notifications had created a sense of urgency and anxiety that I’d grown accustomed to. Removing that background hum lowered my stress baseline. I described it as “like stepping off a treadmill I didn't know I was on.” There were moments of pure contentment – such as enjoying a sunset (not looking directly at it, obviously!) without feeling the need to photograph it for social media – that surprised me. I wrote, “I feel more human again.” That’s a powerful statement about the psychological restoration I experienced.

Of course, I’m candid that life didn’t become magically perfect. I still had work deadlines, the kids still made messes and did ‘kid’ stuff, and bad days still occurred. But without the phone as an ever-present crutch or agitator, I felt better equipped to handle everything. The experiment had its share of doubts and difficulties, but by its end the surprises were overwhelmingly positive – many of which I hadn't even set out to achieve.

Subscribe now

Share

Share SparkLab

Leave a comment

On a brighter note…

One of the most heartwarming outcomes of this experiment was its effect on my relationships, particularly between me and my children. The absence of a phone in my hand meant a presence of mind and heart that hadn’t been fully there before. I always did my best to keep away from my phone when my kids were around, but this past year really solidified it. My daily logs showed a concerted effort to be present. For many days in my journal, I answered the question “Was I present and mindful throughout the day?” and by late 2024 the answer was often a satisfying yes. My conscious focus translated into tangible improvements in family life.

I wrote more about my kids in my journal—funny things they said, small victories, concerns and feelings. Whereas before much of my writing centered on my own stresses. For example, I noted how my middle daughter’s anxiety symptoms improved when I spent more one-on-one time with her (rather than giving her an iPad). With the phone down, I was available to play, to listen, to laugh. My kids, sensing they had my full attention, responded positively. There were fewer tantrums at bedtime, more cooperative homework sessions, and generally a calmer household rhythm. On a weekend in July 2024, I reflected: “1. Was I present and mindful throughout the day? – Yes, took the girls to the playground and left my phone at home. We had a blast.” This simple act of leaving the phone behind became a habit on weekend outings.

Beyond building up my kid’s and my relationships, I also nurtured other bonds. I had more meaningful face-to-face conversations with friends (often initiated by actually meeting up instead of chatting via text). In fact, in mid-2024 I was making many excuses to socialise without screens – something I hadn’t done in years. It’s not perfect, but it’s an improvement! Social media detox meant some contacts faded, but those that mattered found new channels: old friends started calling on the phone or arranging lunches. I mentioned a “coffee chat with an old colleague” in my journal that likely wouldn’t have happened if I’d been busy commenting on Facebook posts, or whatever garbage instead.

It’s worth noting that not all my relationship dynamics were immediately smooth – the change did require adjustment. Family members initially were surprised (“Are you OK? You’re usually on your phone,” a friend joked when I actually paid attention during a gathering). I was writing “clarifying emails,” instead of shooting off reactive texts, which I credit to a clearer mind due to less phone stress.

The data itself didn’t “measure” my relationships, but my journal is full of indirect indicators: achievements like my daughter ordering food on her own, notes about reading more bedtime stories, tags like #bonding and #presence started appearing. One particularly touching entry around Christmas 2024 describes how my kids and I sat around with holiday music on – “no devices, just us” I wrote. In that moment I realised how far I’d come from the previous year, when I might have been half-engaged, scrolling through work emails under the table. Now, I was there. Present.

Overall, being less absorbed in a screen made me more absorbed in the lives of those around me. It resulted in a richer fabric of relationships – with my kids most of all – woven with attention and care. I remarked a few times that this experiment’s greatest gift was that my kids got “more of their dad” and I got to truly know them better. It’s a poignant reminder that the minutes we reclaim from our devices can be given to the people we love.

Our time is finite. Use it well.

Get more from Braeden Mitchell in the Substack app

Available for iOS and Android

Get the app

Leave a comment

Share SparkLab

Share

Subscribe now

Introduction

One doesn't simply cut a bad habit. Just like killing off a warlord, if they’re not replaced by someone who cares about the people, all you’re left with is a void, starving to be filled.

How much time do spend staring at your phone? Or are you too afraid to check? It's not just about the time you’re consciously using your phone, but all the time you’re unconsciously thinking about it. The time you spend carrying your phone around like it’s your child (don’t worry, your secret’s safe with me). The time you spend worrying about what you’re missing (from other people’s lives, no less). The time you spend feeling uneasy when your phone’s not right next to you, freaking out you left it in an obscure place, never to be seen again—and the thousand other ‘what-ifs’. These fears are all unwarranted, but your brain is fantastic at convincing you of something, when it wants. When you want. It’s your brain after all, not some stranger’s. I'm not feeding into your delusion, I’m more than happy to snap you out of it, however.

What are you supposed to do with all your ‘extra’ time after ditching your phone? In mild cases, you might get 10 hours of your life back each week, in extreme cases you’ll be reclaiming 40+ hours a week. Realise you could have been working another full-time job, or doubled your hours of your current job. You could have been working on a skill, a passion, a hobby—or lord forbid, spent quality time with your loved ones.

Ask yourself: What’s one thing I’ve always wanted to do, but ‘never had time for’?

In my case, what was initially going to be an experiment turned into a lifestyle redesign. I began consciously filling the void left behind by mindless scrolling with meaningful activities. A bit of everything. I didn’t think to myself, “I’m gonna to replace this time with x (not the social media platform..),” I became conscious of my addiction (let’s not kid ourselves, that’s what it is). l realised I want more from life, stopped allowing my excuses to rule, and started doing things I wanted, rather than watching others.

One of the most insidious side-effects of phone addiction is the misguided belief we’re ‘present’ while glued to our phone. Even having our phone nearby can slash immediate attention by 8-10 %, and preschool kids who watch more than two hours of screens daily show 7.7× the odds of ADHD-like symptoms. Our phones keep us in a constant state of, not all the way disengaged, yet not quite present and engaged. We’re in a persistent attentional limbo. We’ve got ourselves convinced we’re these multitasking machines. We’re not.

How long till your brain bounces back?

• 45 min: vigilance starts to rebound

• 24 h: you notice phantom-phone reaches

• 14 days: sustained-attention scores look ~10 yrs younger

• 30 days: dopamine pathways largely reset

A few years ago I had every excuse under the sun to use my phone. l just want to quickly note this down; oh I remembered about a problem I was having with a server and I might have a solution, let me test it; someone messaged me, I should reply before I forget; gonna check the status of (any of a hundred different things). It was endless. The mental exhaustion was becoming insurmountable. At the time I didn’t recognise the main cause, looking back it makes perfect sense. My brain wanted so much to have peace, to have time and space to breathe and focus. I wasn’t letting it, I was sabotaging myself and my phone was an accomplice.

Improve Yo’self: Where did all the Time Go?

I’ve been doing the whole, “self-improvement” thing for some years now. It’s one of those things you chip away at, not seeing much change (consciously) and then bam, all of a sudden you have an epiphany about how much has changed. The frog boiling in water and all that. We’re sitting in it, we are it. You’ll get reminders here and there, a friend who hasn’t seen you in a while, “you’re looking fit,” or you’ll realise conversations are just ‘easier’, or those butterflies who used to crowd your stomach in many situations have migrated south, for good.

I dusted off some books and made my way through them. What I read wasn’t cliche self-help books, and not all was for my benefit alone. What I’m realising, even as I write this, is generic self-help and motivational books / videos, anything are next-to-useless. They’ll give you enough dopamine, motivation and excitement to buy their course, their book, their body-doubling motivational cockatoo. That’s it. What they give you for free is a taste of the dopamine and temporary motivation you’ll get when you fork out hundreds, if not thousands of dollars. The madness of it all is you’re paying for the privilege of further diminishing dopamine levels. Once your hormones are back at ‘normal’ levels, you’ll be right back at square one, looking for your next ‘hit’. Yes, I’m gonna keep beating that drum. These are addict behaviours. Once you admit it to yourself and come to terms with it, the real work can begin—unabated.

Now that I’ve broken my soapbox, here’s exactly what I’ve done over the years (if you’re looking for a quick-fix, see the paragraph above):

• diagnosed with ADHD

• got myself out of a shitty situation

• started understanding myself better (with intent—this takes time, patience, and is neverending)

• embodied (I’m not overly religious or ‘spiritual’, but thinking this way daily helped me let go of so much that wasn’t mine)：

God, grant me the serenity to accept the things I cannot change,

Courage to change the things I can,

And wisdom to know the difference.

• recognised and accepted my areas of struggle

• read books that helped leverage my strengths (if you don't know what these are, start at point three)

If you want a detailed breakdown let me know—the point is it’s your journey; generic motivational tripe will provide short-term, artificial dopamine boosts. Break out of the loop. Be honest with yourself and willing to act on what you discover—if you’re not then you won’t, and being stuck where you are becomes a choice, your goal. It’s that simple.

It’s Productivity, Dear Watson

I won’t delude you into thinking I’m nothing but a productivity powerhouse now. Lying to you for clicks, likes and a short hormone spike ain’t my style. One of the biggest mistakes you can make when deciding to ‘be better’ is thinking you’re gonna see dramatic change in a short amount of time. Why though? For one, realise your dopamine generation is shot to hell because of your phone addiction. This means you’re trying to climb Mount Everest with no sherpa, no oxygen and bugger-all food. Be conscious of this! Secondly, we’re human (something many of us forget, or ignore). Your mind and body aren’t manufactured to operate within a tolerance of a thousandth of a millimeter. They’re squishy, malleable, plyable, growable. That’s great, because it’s what we want. If we were as rigid as titanium or steel we’d be stuck with the hand we were dealt at manufacture. Remember that.

I’ve been sitting and writing for the past 2 and a half hours on this article, it’s been productive too. what should you take away from this?

• stay in your own lane—it’s taken a lot of work and failure for me to get to this point; focus on where you’re at—it’s about you, not me (I don’t know how I stack up against others, and I don’t care)

• I have many ‘off’ days and you will too; accept it, embrace it

The Replacements

If you try and replace every minute of your phone usage 1:1, you’re setting yourself up for failure. Deep down maybe that’s what you want. However, once your phone addiction subsides, a new world opens up to you. Over this past year I’ve tried many things and some of them even stuck.

I tried a bunch of different ‘routines’ and ways of working. Simplicity won out over structure. Discovered my weakness for endless planning and ‘tweaking’ to avoid the boring stuff.

I ditched video courses and podcasts for books (full disclosure: on my e-ink tablet but never on my phone). Shorter books won out over longer, complex books. This could be explained by my ADHD—it could also be due to the effects of prolonged phone addiction (with a large serving of short-form content). I read about copywriting, sales, systems thinking, techniques on writing notes, ideas. I aimed to read one book a month. I failed—I failed to even track it. Sometimes keeping track of things takes up more cognitive load than it’s worth. I still finished a number of books.

While reading I kept an A6 notebook and wrote down ideas and thoughts as I went (one idea per page). This really helped the information stick (and is backed by research {add relevant research}). I have an article on the method here.

I experimented with taking voice notes instead of writing them as described above. l applied the same logic—one idea per voice note, limiting them to 1-minute to mimic an ‘A6-sized page’ worth. Voice notes are not as effective as handwritten notes, especially for later recall.

l successfully replaced passive consumption with active learning and doing. Mostly.

The Irreplaceable

A number of years ago I could have been classified as, physically present but mentally checked out. Trying to do it all. Juggling thirty plates made of pancakes dripping with syrup. For context, I can’t juggle more than two balls. I was trying to juggle at least four different worlds, each with their own unique needs: my work world, my kids’ world, an unappreciative and controlling wife, and my own world, which was filled with complexities all on its own. I made myself a promise that I wouldn’t let my phone interfere with time with my kids. I kept that promise, though there was a brief period where I didn’t. I would make excuses as we all do, like using my phone to look up activities we can do together, which led to tangents or an endless planning or researching scenario. In this past year I made much more of a conscious effort to spend quality time with my kids. All of these efforts stacked up, and our bond has strengthened a lot. Seeing their confidence and abilities grow has been fantastic.

Doing More With Less (Distractions)

I always wanted to write more, at the same time I didn’t want to be a slave to my phone. My compromise was to invest in an e-ink tablet. This is what my article writing (sometimes) looks like now (my first draft is usually all handwritten then transformed into text):

I write more across the board now: journalling, ideas (books, tech/gadgets, many), outlines, plans, as well as public-facing writing you’re reading right now (thanks for reading!)! As time goes on I’m getting much better at not just writing but being more effective in my writing. Writing also improves my ability to think through ideas and problems.

What I learned: being distracted ≠ busy ≠ productive/effective - learn to slow down, do less, enjoy and appreciate what you have.

What else did I get up to? Played the piano more—a few times I would just sit for 30 - 40 minutes with my MIDI keyboard and record some nonsense. I practiced singing—no, I’m not available for your kid’s birthday party, thanks for thinking of me though. Maybe for their 21st. Cooked more, though this regressed, sometimes it’s too time-intensive. De-cluttering went from, “oh god I don’t have time to deal with all this,” to, “hey, this is calming and therapeutic.”

The most interesting thing I replaced my phone time with was—nothing. Yep. Nothing. Also known as bugger all, zilch, nada, zip, ingenting, null and a bunch more. When I say you should do more with less, well, there’s not much less than nothing. If you’re young, the idea of doing nothing will be a foreign concept. When I was a kid it was something I did a lot. Laying, sitting just staring at...nothing in particular, usually a wall, zoning out. If I was proper zoned out, time would fly. I’ve since reunited myself with this past-time. At first it was difficult, uncomfortable. All I could think about was my damn phone. Worrying, fretting. About what? Unimportant stuff, that’s what. There were times I felt myself reaching for my phone before I was conscious of it. In the beginning I forced myself to do something whenever I thought about grabbing my phone. Push-ups, sit-ups, getting a glass of water, a cup of tea, to break the cycle. I still do—idle hands are the devil’s playthings after all. But now I’ve grown accustom to the feeling. The anxiety of, “I have to be doing something” is all but gone. I no longer feel guilty about it. You shouldn't either. If you’re a knowledge worker, this goes double. Sleep isn’t the only rest we need.

Enjoy your ‘unplugged reflection time’.

You’ll start to notice things again. How the sun feels—birds! Did you know birds are a thing? They sing all kinds of songs. The sounds of the gravel and grass as you walk. You’ll notice quiet, you’ll appreciate it. You’ll appreciate noise as well.

You’ll miss hundreds and then thousands of notifications and it’ll be glorious.

Most of all you’ll notice the absence of unnecessary stress. It’ll float away, not in five minutes, not in a day, but it’ll come.

Over the years, the last year in particular, I’ve swapped out many virtual activities with real-life ones. Whether with people, projects, or even cleaning up, everything feels more grounded and purposeful than the transient dopamine hits my phone used to provide. I know my last article was the physical one but I can't help but drill home the benefits of doing stuff over being stationary, glued to your phone (or any screen for that matter).

My sleep and stress levels are objectively better.

Do you want to sleep better and stress less? It’s in your power.

Support me

We're all more sedentary than any other time in history. We spend 8 or more hours staring at a computer screen, on our breaks we turn to our phones for entertainment. Then when we get home in our ‘downtime’ we binge-watch or doomscroll on our phones or ‘smart’ TVs. It's a vain attempt to ‘relax’—by staring at screens with different pixel arrangements than the ones at work.

Globally, individuals now spend an average of 6 hours and 40 minutes per day on internet-connected devices, amounting to approximately 40% of our waking hours1. Broken down by age, daily screen time averages are startling2:

When considering the purpose, screen time is typically split between work-related tasks and leisure activities. Adults working remotely often clock 7–9 hours daily on screens for professional purposes3. Leisure time adds significantly to this, with adults averaging about 2 hours and 23 minutes per day on social media alone4.

Smartphones dominate our screen-time habits globally, averaging about 4 hours and 37 minutes daily5. Interestingly, older adults (particularly over 70s) have become significant online users, primarily engaging in activities like web browsing and emailing, nearly matching young adults in their digital engagement6.

This pervasive screen-centric lifestyle is not without consequences: increased screen time correlates with a host of health issues such as sleep disturbances, obesity, anxiety, and depression, particularly in adolescents7.

Our physical environment and entertainment encourages us to be behind screens, COVID became the catalyst for the trajectory we were heading towards. No matter how much we try to convince ourselves, the amount of time we spend on our butts, not moving, staring, zombified by all the dancing monkey content isn’t healthy for our mind or body. We've become the most cunning excuse machines to continue on our path of sedentary self-destruction. “I’m too tired,” I’m so worn out from my kids,” “I just want to relax,” “I want to switch my brain off,” “I don't want to think about all that right now,” “I feel like I haven’t had a break in years,” “I just wanna zone out.”

On and on it goes, one excuse after another, all the while we’re getting fatter, lazier and more unhealthy than ever. Mind, body, and soul. The terrible thing about all this is the acceptance. Everyone does it. It’s the social norm, so we allow ourselves to perpetuate the self-destructive inactivity. Reminds me of something I heard many times growing up,

Everyone is dragging their slothenly behinds right toward the cliffs edge through sheer comfort, first-world excuses and, with irony of the highest order because they’re unfit. Many will delude themselves into thinking they’re fitter than they are in reality. Oh, you can walk to the shops unaided? Wow, elite athlete coming through! “I don’t want to look like a body builder,” “I'm mostly muscle already,” “I’m happy the way I am.”

With that out of the way, on to part 3.

Physical Health

Constant phone use is a sedentary activity, obviously. Hours slouched on the couch, at a desk, hunched over a glowing rectangle. It eats time. It eats the time you’d otherwise spend walking, cooking, actually chewing your food instead of inhaling it while reading Reddit arguments. It creeps in, then it’s got ya.

Over the past year of cutting back screen time, I noticed something weird: I started moving more. Not on purpose right away. I wasn’t counting calories. I tried that once, it’s too much effort. I just stopped losing hours to nothingness and started doing things. Moving around more. Standing while working. Cooking actual food. Being present with my kids more.

In late April 2024, I weighed about 84.8 kg. By December, I was down to 76.9 kg. No hardcore fitness plan, just less screen time. It stacks up. Not just physically, either. Psychologically, it shifted something. The less I lived in my phone, the more I lived in my body.

Correlation isn’t causation. Yeah, yeah. But you don’t accidentally drop 8 kilos during the same stretch you cut your screen time unless something else is shifting underneath. I didn’t just trip and, whoops, there goes 8kg! I cooked dinner more nights than I didn’t. I walked more. Played outside more. I wasn’t doomscrolling while the kids zombied out on tablets (that’s never been our thing anyway). I was there. Moving. Deciding. Less distracted, more present.

The proof is in the pudding with my recent Garmin numbers. From June 2023 to May 2024, I ran 174.1 km. That’s about 3.3 km a week. Decent, considering I wasn’t really paying attention. But compare that to June 2024 to May 2025: 447.1 km. That’s 8.6 km a week. More than double. Same person, same life. Just less time eaten by the void.

As with my other articles, my Garmin data backs this up. My overall step count wasn’t super different over time, but I logged workouts. Wrote about hitting 10k (running). By early 2025, weight crept back up slightly—81, 82 kg. But I felt better. Stronger. More balanced.

The little things added up. My back didn’t hurt as much. My posture improved. I swapped a few scroll-breaks for 5 minutes of stretching or squats. Not because I read it in a productivity book. Because it felt better. Less phone. More motion. Fewer aches. More energy.

Our bodies pay a tax for every hour we spend hunched over a screen. Forward head posture. Tight hips. Shallow chest breathing. Eyes strained from artificial light. Constant micro-tension in the neck, shoulders, jaw. It doesn’t happen all at once, it’s a slow erosion but when you notice it, you’ll really notice it. One day you realise your back’s stiff before noon, or your hip clicks every time you stand up, or you’re rubbing your eyes at 2pm like you’ve aged twenty years overnight.

Even cortisol (the stress hormone) sits higher when screen time goes up. Your brain doesn’t care that you’re “just scrolling.” It sees blue light and flashing inputs and reads it as threat, not rest. So while you’re lying on the couch “relaxing,” your nervous system is still buzzing like it’s stuck in traffic.

We dress it all up in fake productivity, pretend, make-believe. “I’m just clearing emails,” “I’m planning the week,” “I’m reading up on this thing.” Lies. I did it too. Pretended tapping through Jira was productive. It’s not.

Cutting screen time did wonders for me, it got rid of the illusion that I was doing something useful when I wasn’t. Forced me to ask myself, what is worth doing when there’s no feed to refresh? Cooking. Running. Sitting outside. Writing this, even. Real stuff. Physical stuff. What the hell do I care about watching other people’s lives?

There was a spike around October where my phone use jumped, weight loss plateaued, running dipped. Then January came, and I reset. Ran more. Ate better. Dropped weight again. Bought new jeans in February because the old ones didn’t fit anymore. Tangible result. And yeah, I credit it partly to less takeaway and more time in the kitchen. But mostly to not being glued to my goddamn phone.

Some Tips

1. Log your physical activity if you’re not already

2. Put your damn phone down—all that crap you’re convinced you can’t miss is totally missable and you won’t even miss it

3. Talk to a human

Adlerian Tips

1. Your mind and body are one system: movement boosts mood; better mood lowers craving for digital escape (and inversely, worse mood increases craving for digital escape and distraction-based activity)

2. Picture future you: each habit points to a vivid “future self” image—energetic, strong, present with family—so daily actions serve that goal

3. Play with your kids, work outside when you can, strengthen real life bonds—satisfy your need to belong in the real world not pixels

4. Tiny, winnable actions (two-minute moves) spark daily success hits that crowd out guilt-scrolling

Subscribe now

Share

Share SparkLab

Leave a comment

Get more from Braeden Mitchell in the Substack app

Available for iOS and Android

Get the app

Creativity is a muscle and mine had been slowly atrophying, having the life sucked out of it by a 6 x 3 inch glowing rectangle of doom.

Once I cut back, my Logseq graph (the place where I put my thoughts, notes, etc.—a PKMS (Personal Knowledge Management System)) jumped from 0 to 38 #idea tags in one quarter. The proof is in the delicious, phone-less pudding.

What does creativity mean to you? Is it something you think you have? Have you actively worked on it? Myself, I've had an active imagination as long as I can remember. I didn't always have ways to harness or keep it at bay—there were times it became a runaway freight train. Yes, to all those aren't “naturally” creative, it can be a double-edged sword. Especially when you're a kid and the whole world is new and exciting, sometimes scary. Sometimes terrifying. Thinking back I didn't have many outlets. I was just the weird kid who could make people laugh. Whether or not you have an official outlet, its gonna come out one way or another. Usually not in a good way—at least in my case. Even if you're not naturally creative—whatever that means, you can work on it, improve it.

I used to daydream in class. A lot. During primary school (elementary) I'd be in my own world. Instead of paying attention to whatever was going on, I had shrunk down, made a flying fox out of a paper clip and some string and ventured my way from one end of the classroom to the other. Or using my eraser as a remote control for my micromachine harrier jumpjet. There were of course times I was suddenly yanked away from my alternate reality back to the one most of us are familiar. Sometimes my subconscious was paying attention so I could bluff my way out of trouble, but I also spent my fair share of lunch times and other breaks in detention. School wasn't too fun a lot of the time. Outside of school I did my best to fit in, to suppress my weirdness, my creative imagination. It worked for a while, but I'd inevitably slip up somewhere and I'd be back to square one. I don't know if you suppress it long enough whether it dies down or if you'd end up having sort of mental break.

I reckon it'd be interesting to conduct long-term studies on the effects of suppressing creativity. For example, through extended use and exposure to smartphones. There's no doubt in my mind the mere act of having a smartphone nearby is enough to interrupt our brains, holding back our creativity. I've observed it first-hand and I dread to think the effects it has on children's development. People will argue it's the same as older generations staring at TVs growing up. The difference is companies have poured so much money into designing an ecosystem for the purpose of gaining and maintaining attention. These devices are with us all the time and we've been socially conditioned to become dependent on them. We are being pushed to use our phones for everything under the guise of convenience. In reality, companies are more than happy to reduce their bottom line, not having to worry about manufacturing physical cards and the rest of it. The, “it's better for the environment” arguments also fall flat when you realise how terrible smartphone e-waste is, considering how many phones are produced and discarded every year (gotta have this year's iteration, right?). In 2024 alone, manufacturers pumped out about 1.22 billion new smartphones—nearly 40 fresh handsets. Every. Single. Second. The WEEE Forum estimates that 5.3 billion old phones were chucked or stashed away in 2022.12

See if you can picture 5 billion phones in your brain, and realise that happens each year and is only going to keep growing. Struggling to visualise it? Here's some reference points to help you out:

Those are still a bit difficult to comprehend. Here's a better one:

Empire State Building (by volume) ~½ a building The Empire State’s interior volume ≈ 37 million ft³ (≈ 1.04 million m³). Our e-waste mound is about 43 % of that.

All we've done is create attention-sucking gadgets that never leave our side, with the co-dependency to match. Keeping us distracted, good little consumers, who are more angry, divided and confused than ever. Oh, and deluded. Almost forgot that one. We stand in line to the checkout with our headphones on, delivering our lines almost unprompted. The ones we've memorised-no, I don't want a receipt, thanks. No, I'm not a member. No, I don't want to become one. It's starting to feel like we're actively avoiding real-life social interactions. Yet we'll happily send message after message through the air, to people who may or may not even be people. We're training our brains, tricking ourselves into believing the virtual world, which is mostly bots spewing out pixels at this point, is inherently better than the real one. It's mass delusion bordering on psychosis, a socio-technical contagion. A true epidemic, pandemic that we're all too blind to see because we're so caught up in it. We can't see the forest because the trees are injecting us with dopamine.

With that out of the way, on to part 2.

Part 2: Creativity and Cognitive Shifts

Looking back on the past year the most striking change came in the form of creativity and mental clarity. Before, my mind was often scattered–countless social media posts and ‘infinite’ scrolling tend to clog anyone's creative channels. However, as smartphone distractions dwindled, I experienced a kind of “cognitive renaissance”. My journal entries and idea logs exploded with activity. In mid-2024, new ideas were few and far between, but by late 2024 I was recording ideas almost daily. In fact, the number of #idea tags in my Logseq notes went from near zero in the spring to 38 distinct ideas in Q4 2024 alone. Once I was freed from constant phone notifications, I became an idea machine.

See that Christmas tree spike (pretty cool how it worked out that way!)? That's what happens when boredom replaces {insert whatever crap you spend your time on, for me it was Twitter (the only time I've been on it was during this period), YouTube, Substack}.

What my data shows

I documented this idea spike phenomenon in real time. On December 17, 2024 at 9:24pm, I jotted down a revelatory note:

I haven't been staring at my phone or computer nearly as much... yesterday by the time I put the kids to bed (7:30pm) I'd only used my phone for 28 minutes that entire day, compared with my normal 4–6 hours. I really think this has a massive effect on creativity and idea generation–you can think clearer… there are so many benefits from not staring at a screen all day.

My realisation–that idleness and boredom, once filled by mindless scrolling, were now sparking idea after idea–capturing the essence of my cognitive shift. Though it's less about ideas suddenly being sparked and more opening a gate that allow ideas to flow freely. All those notifications and staring at mindrot act as creative blockers. They stop us thinking clearly. In the end I found myself reaching for a notebook instead of my phone. By removing the endless feed of others' content, my mind started producing more of my own content. I brainstormed topics for my newsletter, jotted down business concepts, even outlined book chapters. My Logseq graph is full of new nodes in late 2024, evidence of my mind reactivating.

The below image from one of my previous articles visualises the idea of your phone as a creativity blocker:

The article is here if you're curious.

Another journal entry from January 5, 2025 illustrates my change in creative ability. I wrote at 10:29am,

I would never have been able to just sit and write like this when I was constantly on my phone

marveling at how I can now spend a quiet Sunday morning composing thoughts without the itch to check messages. I even drafted a poem of sorts to celebrate the moment:

To think great ideas, get your brain into gear, put away your phone…

based on Dr Seuss–a simple mantra that became a new credo. I reposted it as a note again because it disappeared

The positive effects went beyond creativity. I also noticed sharper focus and memory. By the start of 2025, I observed that I could read long-form articles or books and actually remember them. In a reflection on New Year's Day, I wrote,

Even my ADHD symptoms are less—more focused, less distracted.

This was a surprising benefit. I'd expected to regain some time, but not necessarily to feel cognitive fog lifting in such a tangible way. Reduced screen time meant my attention span began to recover. Tasks that used to feel insurmountable (because my phone would interrupt me every few minutes, or I was glued to it like a zombie) became easier when I could work uninterrupted for an hour.

I also poured this new focus into personal projects. For instance, I really started getting into 'vibe coding,' for lack of a better term—seeing what was possible using LLMs (Large Language Models). I'm now up to 70 code repositories, with a few having real promise, as well as a bunch of lessons and some useful scripts. It was also much easier to think up activities my kids and I could do together. We dipped our toes into gardening, and after a couple of hic-ups our plants are thriving (my youngest has taken ownership and does all the watering).

I posted this as a substack note as kind of a running joke:

– a lighthearted jab at how I now spent time gardening and writing instead of chasing clout online. It shows an emerging creativity and humour in my writing that wasn't there before. It was there, I just had to reactivate it. By breaking the cycle of consumption, I unlocked a well of production.

My switch from passive consumption to active creation wasn’t an accident; it happened because I built a few tiny, repeatable rituals that force-fed my brain boredom and then gave it somewhere to dump the ideas. Yes, I used Logseq (it has a phone and desktop app) to log ideas, however, many of those started in a physical notebook and I transferred them over.

Ready to run your own cognitive renaissance? Steal my three go-to moves (plus a dash of Adlerian psychology to keep things spicy):

If you’re unsure what ‘Adlerian lens’ means, see the citations here →345

5-Minute Idea-Journal Template

⏱️ 5-MINUTE IDEA SPRINT
Trigger: ____________________   Date: _____________

1. _______________________________
2. _______________________________
3. _______________________________
4. _______________________________
5. _______________________________

Mood after sprint (😫 1 – 10 😃): ___

Your only KPI for this is, pages filled, not “good” ideas.

If you want a PDF version of the above, let me know!

Leave a comment

Micro-Proof These Habits Work

• After 30 days of Idea Sprints my Logseq #idea tags jumped 3×

• Boredom blocks coincided with the December “Christmas-tree” spike you saw on the graph

• Analog night-caps dropped my average sleep latency by ~12 minutes (Garmin data)—less bedtime doom-scroll, more REM = sharper mornings

Try it: run the three habits for one week, then DM me (or comment) the weirdest idea you logged. If this playbook helps, smash the share button—let’s pull more brains out of the scroll-swamp.

Like what you read?

Go on, spoil me

Until next time: ask the awkward question, break a tiny system, and document the delightful chaos.

\

Subscribe now

Share

Share SparkLab

Leave a comment

Get more from Braeden Mitchell in the Substack app

Available for iOS and Android

Get the app

• Trike riding—you know, a motorbike with an extra wheel

• Stepping in cow patties

• Hanging out at hāngīs

• Fixing fences

• Riding quads (motorbike with two extra wheels)

• Drinking alcohol (responsibly—I waited till I got home to do it irresponsibly)

• Welding

• Panel beating (wife beating minus the wife and the abuse with car panels)

• Watching kids play Tomb Raider on their fancy computer

• Eavesdropping on strangers’ phone calls (was easier back then, wasn’t all this fancy encryption on cell connections as we know it today)

• Calling my friends long distance from my mobile phone, because of course I did

My first phone was a Philips Diga.

This absolute menace of a phone:

While my friends at school and probably a couple in our neighborhood had the cool Nokias, I was stuck with this thing. Despite what it looks like in the image, the screen could only show a single line of text (the rest of the screen had little icons and stuff). This meant that when anyone sent some ASCII image I had no idea what it looked like because I couldn't see the whole thing at once—it just slowly scrolled across row-by-row as a single line. To put it in perspective how painful it was to use, here's some fun (read: rough) calculations based on receiving a single SMS (capped at 160 characters):

Philips Diga vs Nokia 3210 showdown

Philips Diga

You can only see 12 characters at a time on the screen of the Diga and the other characters slowly scroll across the screen.

So,

That’s ≈ 14

14 “screens” to read a single message.

Nokia 3210

On this bad boy you can see 45 characters at a time.

≈ 4

4 screens and you'd be done reading.

If we assume 2 second read time of the message and each screen scroll is 2 seconds for the Philips Diga, then:

For the Nokia you can go back and forth screens instantly—you don't have to wait arbitrarily to read a message.

Every time someone sent me a message they were unwittingly wasting 26 seconds of my life. How many messages do you send a month? At one of my previous jobs I worked with a guy who would regularly send 2,000 SMS a month. Two-thousand. He was probably using a Hiptop or something, because there’s no way you’d catch me writing 2,000 SMS a month without a QWERTY keyboard. For the uninitiated, this is how everyone use to type messages. If you wanted to send, “Hello world”, this is how you did it:
H – press 4 twice → 44

e – press 3 twice → 33

l – press 5 three times → 555

l – press 5 three times → 555

o – press 6 three times → 666

(space) – press 0 once → 0

w – press 9 one time → 9

o – press 6 three times → 666

r – press 7 three times → 777

l – press 5 three times → 555

d – press 3 one time → 3

Pro tip: If you want to save more time writing messages, give gesture typing a go.

≈ 72

≈ 5

Five messages an hour, every hour of being awake, every day, every week. Regardless of work or anything else.

Madness.

If I was him I would have lost 14 hours of my life every month just waiting to read the messages. This is probably one of many reasons Philips was never a big player when it came to phones. Funnily enough these types of screens (LCD with a separate back-light) weren't bad for our eyes and our mental well-being the same way modern phones are. Sure, they were simple, but at the same time looking back that's one of the reasons that made them so great.

Calls, SMS, saving contacts.

That was it.

They didn't destroy our attention span with 100 different apps desperate for us to notice them like some needy ex showing up at your house at 3am promising they'll change.

They didn't advertise to us 24/7 through 9×16 billboards pretending to be games or useful apps.

They didn't track every metric known to man and dob us into intelligence agencies and the police for wrongthink.

That was the beauty of them.

Why do you think Apple insists on using minimalist designs? Because they work, we love them, because simplicity is beautiful—it's easy to understand. Simplicity lets us focus on whats important.

With that out the way, on to part 1:

Sleep Quality and Duration.

For as long as I can remember I've struggled with sleep. I could just never figure out how to do it right. When I was in my late thirties I was diagnosed with ADHD. This shed some light on my sleep issues, but I still wasn't able to shed my way to restful and rejuvenative sleep. I've made a bunch of improvements over the years yet no matter what I did, waking up “ready to tackle the day” seemed out of reach.

Those damn screens.

One of the first areas of impact after reducing my screen time was sleep. Like many of you, I was a sucker for a little late-night screen time. A cheeky social media check here, a harmless scroll there right before bed stretched itself into an hour or more—then of course I needed some “me” time so I'd load up a movie and before I knew it 1am rolled around. I'd absent-mindedly brush my teeth and stumble into bed. Why couldn't I just be tired at the right time like a “normal” person? That’s it, though—I did. I was. I ignored it and suppressed by staring at my phone. All those bright lights, especially that devious blue one responsible for keeping our brains active and alert, were enablers and hormone pushers (dopamine). Encouraging me to ignore my primal, it’s time to sleep urges. Here's an example of how extreme it had become:

11 hours and 32 minutes.

This was the day I started taking (technically the 21st) data. Maybe my subconscious knew something was up, maybe it was fate, doesn't matter—what gets measured can be improved. And X (Twitter), of all places! X and YouTube (New Pipe is a different front-end for YouTube—it allows you to download video and audio from YouTube, among other things—it's a pretty great alternative for the YouTube app). There's no icon for them because I uninstalled both.

This is where it all started and for the next 12 months I started tracking my phone usage. In addition, I was already keeping track of my weight and sleep/fitness and tried to journal stuff, though that was a bit hit-or-miss in terms of consistency. The results tell a compelling story of digital minimalism in action. Come with me now on the first of 7 parts that will go through how a year of less phone time affected my sleep, creativity, physical health, daily habits, relationships and more. Each part draws on both quantitative and qualitative data (journals, etc) that will paint a picture of personal transformation-and how you can achieve the same or better.

During the period in 2024 my watch reported I was getting 6 hours sleep a night, on average. In reality it would have been less. Due to how it collects data and “senses” when I'm asleep I dare say it would be closer to 5.5 if we're talking average each night, possibly less. For example, if I'm sitting or laying very still around bedtime (programmed for between 2200 and 0600) my watch reports me as sleeping. This changed when I started curbing my phone use. By the fall, and especially once I committed to a serious cutback (January, 2025) a clear inverse correlation between my screen time and sleep duration emerged. Not only did my sleep duration improve, but the quality as well. If we take two periods for comparison it’ll become clear. Take October 2024 for instance, where my phone usage peaked at over 7.3 hours per day, I averaged 6.2 hours sleep—and remember this was also poor quality sleep. Then, during the period of March–April 2025, my phone usage had less than halved to roughly 3.5 hours and I was clearing more than 7 hours a night of better quality sleep. Regularly. The change is visible in the data: less stimulation in the evening from my phone led to more quality time in bed and deeper rest. I'd cracked it.

As I mentioned before, it wasn’t just the length of my sleep that improved but also my perception, I noted my own experience. I created a template in Logseq to help me track how I was feeling. One of the questions was, “Was I present and mindful throughout the day?” On the days where I answered “Yes” I generally reported ‘easier’ evenings and more relaxed bedtime. By early 2025 I was describing in my Logseq journals that I was feeling more rested and alert in the mornings, something I've struggled with most of my life. Reducing my overall phone usage and all but eliminating the tantalising glow of my phone before bed also helped me establish a more consistent sleep schedule. My Garmin watch will be the first to back me up on this: my average sleep rose from ~ 6.1 hours (less in reality) in September 2024 to about 7.0 hours (a number closer to reality) by February 2025. I was not only sleeping longer (objectively and subjectively) but I also had fewer middle-of-the-night wake-ups (in September my watch reported an average of 30 minutes awake during the night, which halved by the new year).

To sum it all up, reducing my phone usage (especially late-night screen time) paid me immediate dividends when it came to sleep. I did a ‘soft’ phone usage reduction during ’s 30-day social media detox, where I still had some rough nights—because I noticed when I'm more stressed or overly tired I was more likely to rely on that little glowing rectangle. Which in turn feeds into a vicious cycle and makes it all worse. Example, on day 6 of the challenge I noted, “Was still a bit stressed and was using my phone more than normal” late at night, and that was with 108 minutes of screen time and 423 notifications (yes, every one of those notifications screws with your attention and focus). As I progressed through the detox, the habit of checking my phone at night weakened. By the end of the year I was much more conscious and I was writing that falling asleep was easier and that, “consistent sleep is vital... lack of sleep exacerbates attention issues.” This shows how much better I felt when I guarded my “wind- down” time.

How can you maximise your sleep quality and duration?

• Throw your phone away—if thats not an option, continue below when starting

• Have a “phone curfew” (e.g. no screens after 8 pm)

• Create a strict no-phone zone in the bedroom

• Batch notifications—only check messages twice a day

Infusing some Adterian psychology into the mix, think about it like this:

• Holism: Look at sleep and phone use as an integrated whole—improving one uplifts the other

• Encouragement: Frame each phone- free evening as a small victory, not a deprivation

Subscribe or whatever—this is the first of 7 articles based on analysing the data from the past 12 months.

Support me

\

This video took four tries to record. Something went wrong every single time, and this was the point where I decided: whatever happens on this one, I’m hitting publish. It’s fine.

What I wanted to show is the custom GPT I put together—something I originally started as a side experiment but ended up using constantly.

Why I Made It

Alongside the last article (here for those who haven’t read it), I shared a PDF called the ClarityScript Toolkit. It walks through a tagging method I use to sharpen up prompts with tone, context, and delivery style. But then I figured—why not build that logic directly into a GPT? So I did.

It started as a “let’s see if this works” kind of thing, but it turned out to be way more useful than I expected. Mostly because I kept running into the same friction.

You open ChatGPT, and now you’ve got to reverse-engineer a perfect prompt from scratch. Again. Every time. You try to be clear. You give it your goal, some background, maybe format it a bit. Still, the answer comes back a little off. Like it's responding to what you typed, not what you meant.

Eventually I just thought: rather than keep rewriting prompts and hoping for better, what if I built a tool that could do the rewrites for me?

Support me, wouldya?

What It Actually Does

This custom GPT analyzes whatever text you feed it and—if it's vague or all over the place—it tells you so. Directly.

It’ll say things like: “Reads like a bunch of great ideas trying to fit through the same door at once.” Which… fair.

It flags unclear language, suggests alternatives, and asks clarifying questions before trying to “fix” anything. It doesn’t just generate—it interrogates the intent first. Which ends up saving time.

And once you do answer those questions, it starts producing useful versions of your prompt. For whatever medium you're working in—image generation, search, presentations, writing. It gives you multiple takes and lets you adjust from there.

A Real Example: Stress and Performance

In the video, I show this in action using a prompt I wrote about “stress vs. ambivalence optimal graph motivation”.

The original phrasing didn’t really hold up—“ambivalence” wasn’t the right term. The GPT picked that up immediately and asked if I actually meant arousal or pressure. Which I did.

It then offered three refined prompts:

• One for search (good for Google, though I’m still not convinced tags do anything there)

• One for image generation (DALL·E or similar)

• One for a presentation slide (concise, educational, professional-looking)

From there, I used each prompt to generate the images.

Some of the graphs were generated using standard image tools, others using Python directly inside the GPT interface.

So yeah—this GPT doesn’t just tweak text. It also handles visuals surprisingly well, especially when you're trying to represent something like “just enough stress” in a clean way.

Who This Is For

This is mainly for people who want less friction when using GPTs. Writers, educators, designers, anyone building something where tone and structure matter. You can use it to script, outline, refine ideas, or just make your prompt a little less “meh.”

Also useful if you're the kind of person who stares at the blank input box thinking, “How do I even start?”

Access Options

There are two ways to get the GPT and the full toolkit:

• [Subscribe on Substack for a month]
  This gets you immediate access to the custom GPT. It’s the cheapest route.

• [Buy the ClarityScript Toolkit PDF via Gumroad]
  This includes everything from above, plus a breakdown of what the tags mean, why they work, and how to build your own structured prompts.

👉 ChatGPT ClarityScript Deep-Dive Toolkit

Last Bit

The idea here isn’t just to make better prompts—it’s to reduce the back-and-forth. You don’t need to wrestle with GPT every time. You just need to feed it something it can work with.

And honestly, this whole system—the tags, the phrasing, the way it nudges you to clarify—would've saved me a lot of time on previous projects. Now it’s built in.

Let me know if you try it out and get something weird, smart, or unexpectedly useful. Always curious to see how people bend these tools.

See ya!

—

Sometimes I ask ChatGPT something I think is clear but I end up with what feels like a polite or confident hallucination—oftentimes both. It’s not wrong per se—just... not what I meant. I realise now I never gave it a fighting chance. I forgot I was talking to a system, a computer, not a person. It’s something we should all remember if we want effective responses, answers—no matter how “realistic” these systems seem.

ChatGPT doesn't hear my tone, doesn’t see my face, doesn’t get that I’m pacing while I speak or hesitating because I’m trying to land a point (and stop it from replying prematurely). Doesn’t matter if you’re prompting in text, using standard or advanced voice (all voice gets is a cleaned up transcript). If your text or the resulting transcript is vague, scattered, emotionally loaded without a hint of context—guess what I get back? A reflection of that, wrapped in agreeable syntax and formatting.

Why Clarity Matters

ChatGPT doesn’t know your backstory (it might have a summary if you provide it). It doesn’t know you’re mid-rant or low on sleep. It won’t read between the lines unless you draw the damn lines yourself. If you don’t say what you mean—or at least give it the scaffolding, some structure—it’ll spit back something generic, smooth, safe. It doesn’t want to upset you, it's not gonna rock the boat. That’s not helpful. Especially when you’re chasing a signal, something specific, not a cliche quote about lemons.

This is all by design. To be helpful, without friction, and maximise engagement—in a useful sense, not in the immoral, we'll get you addicted in 30-minutes or less TikTok sense.

The Hidden Language of Prompts

The structure of your question is part of your message. Most people type like they’re thinking out loud, hoping the model will just “get it”. It won’t. If you’re indirect, you’ll get a response that mirrors much of the same. If you’re firm, it’ll match that “tone” . If you waffle on, it’ll waffle right back—with references. It’s not just what you ask, but how you frame it. Think of every question like a container. If you leave it cracked open and it's leaking ambiguity, don’t be surprised when the reply slips through the gaps. Tone, structure, intent—these are all necessary ingredients. Without them is like cooking without adding any seasoning, even salt.

Clarity GPT Script

So I built a method through reverse-engineering ChatGPT to make things clearer. Or rather, I started tagging my prompts in a way that not only made sense to me, but would also make everything clear for ChatGPT—this helps minimise confusion and misunderstanding. Which is important if you wanna make use of a system that collates, sorts, categorises, analyses (and whatever else) vast amounts of data and information—with the hope of transforming it into knowledge. I call it ClarityScript—open to suggestions, it was just the first thing that popped to mind. It’s not a framework and it’s not some rigid formula (though I did discover the “rules”). It’s shorthand for my intent. It’s so ChatGPT doesn't have to play a therapist using a decoder ring while reading my mind. These tags are contextual clues. You layer them around your prompt to let the system (ChatGPT, for instance, but there’s no reason why it shouldn’t work with other models) know what mental state you’re in, what tone you’re aiming for, what lens you want the answer filtered through.

Examples:

• [irritated but calm]

• {writing a message to someone dodging responsibility}

• (neutral, but assertive-not passive)

• --this part needs to land, don't soften it--

Prompt Transformation in the Wild

Take this (how most of us prompt/ speak):

Can you explain X?

It’s fine. You'll get something dry, super broad and neutral. But try this:

Explain it like I've got 100 tabs open, a toddler having a meltdown, and I'm two sips into my coffee.

Now ChatGPT (or whatever model you try it on) gets it. You’ve set the stage, crafted the scene. You’re telling the system not just what you want, but how you need it to show up. This small shift in phrasing is everything. It allows ChatGPT to meet you where you are—because you gave it somewhere to meet you.

Where It Gets Interesting

It’s not just about prompting better, it’s about achieving clarity in how you think. How can the system help you find what you need if you can’t articulate it? Once you start using ClarityScript, the process of writing the prompt becomes part of your thinking. The structure allows you to start building shapes from the ambiguity and chaos. You begin seeing patterns in your own thinking. Your own avoidance. Your own lack of clarity in those moments you thought you were being crystal clear. Don’t think of it as a limitation, but a mirror.

What About Voice Mode?

Doesn’t matter if you’re using standard or advanced voice, the rule is the same. ChatGPT only understands what gets transcribed. In standard voice, it gets text and maybe symbols and numbers- it depends on the success of the backend transcription. No tone, no inflection, no hesitations. With advanced voice, ChatGPT will understand a bit more rhythm, pacing and structure. Even then, it’s not hearing you, it’s just analysing the aftermath. ClarityScript still matters here—in fact it matters more because it gives structure to the stuff that gets lost in translation—sorry, transcription. If you’ve got ADHD like I do, and you’re prone to mid-thought tangents—you need something that holds your attention when it tries to sprint sideways.

You can try these:

• voice your thought (record it yourself, transcribe—Whisper, Otter.ai, GoodTape.io), then clean up the transcription and add tags before prompting

• say tags aloud as you speak to ChatGPT (clunky, but doable) e.g.,

  open bracket confused but staying calm close bracket

  Depending on the success of the transcription (before ChatGPT actually receives it), the above could show up as:

  open bracket confused but staying calm)

  Or,

  (confused but staying calm close bracket

  Or even (if you're lucky),

  (confused but staying calm)

• use a template or notes app and slot in tone/context manually, post-transcription (you can use Google Voice-to-text, for example)

Real-World Voice Prompt Example

Before:

Can you open parentheses explain the thing about the memory thing you mentioned yesterday close parentheses? I'm just, like, not sure about that part.

After:

[confused but trying to stay focused] {referring to the memory function in ChatGPT you explained yesterday} Can you explain that part again, but more clearly? Keep it focused, no fluff.

This Goes Far Beyond ChatGPT Prompting

You’ll start to see a mindset shift once you put this into practice—it’ll start bleeding into other aspects of your life. How you write emails. How you explain yourself to other people. How to ask for what you want. How to stop hoping someone will “get it” without you having to say it. Clarity isn’t just an output. It’s a discipline.

If any of this resonated with you, I’m glad. If there were parts that were unclear, or that you’d like more info on, let me know—comment, DM, however. I’m working on a full toolkit that expands on this. Prompt patterns, ClarityScript templates, real-world rewrites, and yes, even the obscure symbol decoding for when you want to get weird. Tools to help you stay grounded when your thoughts are on fire and in a scrambled mess.

Remember: ChatGPT and LLMs in general are systems, computers—talk to them like someone who knows what they want, or at the very least, someone who’s willing to find out.

If you want your copy of the ChatGPT ClarityScript Deep-Dive Toolkit, grab it here.

Subscribe now

Share

Share SparkLab

Leave a comment

Get more from Braeden Mitchell in the Substack app

Available for iOS and Android

Get the app

And that’s not because I’ve changed direction.

It’s because, for the first time in my life I’m standing fully inside the signal I’ve been circling for years.

I’ve spent most of my life surrounded by noise.

Not just external chaos, but the slow, corrosive fog of manipulation, control, emotional distortion.

It started young.

It lasted.

It left its mark.

It was pervasive, pernicious and at times peculiar.

I wasn’t taught how to think clearly.

I was trained to doubt what I saw, what I felt, what I knew to be true.

And when you grow up in that kind of environment, survival depends on reading between the lines, sensing patterns others don’t see, building clarity where none exists.

I needed to.

If I didn't—well, I dread to think what that path holds.

That instinct shaped everything.
Even when I didn’t know what to call it.

So for years, I gave away what I had—ideas, strategies, systems, ways of thinking—hoping it might help someone else make sense of their mess.

But I was still operating under one unspoken lie—an assumption, that my clarity wasn’t worth protecting.
That what came naturally to me couldn’t hold value for anyone else.

I was wrong.

And this shift is my course correction.

What I’m moving into now isn’t a rebrand. It’s not content strategy.

It’s ownership.

Of what I’ve seen.
Of what I’ve survived.
Of the edge I carry now because of it.

It's funny, when we think about epiphanies, we often think of this grand fanfare—a celebration of everything we know to be true, all culminating into one single defining moment. And it was, don't get me wrong. I felt intense emotion in that moment—at the same time, that's precisely the thing—I knew all along, I've always known. So while it was—is—a big revelation, it's something that's always been there. It's me, I just didn't have the tools to realise it. That was where all the work was.

The defining moment didn’t feel dramatic. I'm not a character in a movie.

I wasn’t standing on a cliff, staring into the void.

I was mid-conversation, trying to explain why I do what I do—why I think the way I think.
And I caught myself saying it before I even understood what it meant:

I just want to make things clear… because no one ever made anything clear for me.

That was the pivot.

Not the emotion. Not the pain.

The pattern.

The realisation that everything I’ve built—every idea, every tool, every attempt to help someone else see straighter—came from that core fracture:

I lived in fog.
So I learned to build clarity.

This piece won’t have an image.
No drawing, no visual anchor.
Just words. Just signal.
I want it to land raw and undistracted—like the clarity it’s about.

So my next article might feel like it came out of nowhere.

But for me, it’s been building for years.

If you’ve ever felt like your thoughts are running you in circles—if you’ve ever tried to make yourself smaller, quieter, less demanding just to keep the peace—if you’ve ever given so much of your mind away that you forgot what it’s like to protect it—then what’s coming is for you.

Not to explain.
Not to convince.
But to cut.

There will be challenges, roadblocks, small pivots, distractions—all the human stuff, but my purpose is clear now—and I wanna make yours clear too.

This is my work now.

And I’m finally ready to do it on purpose.

I’d tried all the productivity apps. Most of them are noise. Too many features, too many nested options, too much space between the thought and the note. I just needed something that let me think without breaking the flow—and keep thinking inside the same place the next day, and the next.

That’s what Logseq gave me: continuity. A single page per day. No folders. No overhead. Just thinking in context, without needing to name it.

Why I Use It Daily

• Journals page is the whole system. No navigating. I open Logseq, and I’m exactly where I left off.

• It blends memory with task management. I’ll log an idea, then tag it #idea, and next thing I know I’ve got five linked thoughts and two tasks.

• It’s non-linear—but everything is still traceable. Backlinks are gold. Nothing disappears into the void.

My Workflow

1. Everything happens in real-time.
   I write directly into the day’s Journal page. Timestamped blocks when I want clarity.

   • “08:00 — Slept like crap. Not sure if I’m just tired or done.”

   • “10:17 — Need to rethink the structure of that second project. Might be over-complicating it.”

2. Templates keep the day on track.
   Morning → Tasks → Notes → Evening. I don’t always fill them all out, but the scaffold matters.

   

   

   Here’s the actual template if you want to copy and paste it into Logseq:

   - # Daily Journal
     template:: Daily Journal
     template-including-parent:: false
   	- ## Morning Thoughts
   	- How do I feel today? [Mood: ]
   	- Key priorities for the day:
   	- ## Key Tasks
   	- TODO Task 1
   	- TODO Task 2
   	- ## Ideas/Projects
   	- New ideas or ongoing projects:
   	- ## Evening Reflection
   	- What went well?
   	- Lessons learned:

3. Tagging is minimal but rigid.
   I don’t tag whimsically. If it’s not a project, idea, task, or reflection, it doesn’t need a tag.
   If I’m inconsistent with tags, I might as well not use them at all.

   Though to be honest I have some stragglers I need to clean up!

4. Queries pull it all together.
   Weekly dashboard is auto-generated. All I did was write the right blocks and use consistent tags.
   I don’t spend time making dashboards. I just let the data surface itself.
   Here’s a query for getting the blocks with the tag #idea from the past 7 days:

   #+BEGIN_QUERY
   {:title "Journal blocks in last 7 days with a page reference of idea"
    :query [:find (pull ?b [*])
            :in $ ?start ?today ?tag
            :where
            (between ?b ?start ?today)
            (page-ref ?b ?tag)]
    :inputs [:-7d :today "idea"]}
   #+END_QUERY

5. Images help break the scroll.
   I embed quotes, diagrams, maps, whatever gives context.
   Text is the backbone, visuals are just glue.

   

6. Weekly review. Always.
   I try to scroll back through the week every Sunday. Oftentimes I forget, or fall asleep. It’s good to spot the loops. See what’s repeating.
   Half the insights come from seeing what I wrote three days ago and realizing I’ve already moved on.

Tips That Actually Matter

• Don’t overthink your system. Just write. The structure will emerge.

• Use the same few tags, religiously. Really, this just means don’t overdo it…you don’t need a tag for #it and #everything

• Query only after the patterns are clear—don’t pre-optimize.

• Embed visuals only if they help you think.

• Weekly review isn’t optional. It’s the point.

Last Thought

Logseq works because it doesn’t get in the way. Most tools want you to set up your life. This one just lets you live it—and see what that looks like when you come back the next day. Or the next week. Or when you finally notice the same complaint popping up five times in two months.

If you’re looking for something frictionless, contextual, and quietly powerful, you’re not gonna do better.

And then I stopped.

This all started with ’s 30-Day Social Media Detox. I first wrote about it on June 7th, 2024, over on Thirteenth Strike (see here), but I had no idea then just how much of an impact it would have. What began as a short experiment turned into a complete rethinking of my relationship with technology.

At first, it was just an experiment—limiting phone usage during the day, keeping it out of reach, resisting the itch to check. Within days, I noticed something I hadn’t expected: my ADHD symptoms, which I’d spent years battling, seemed to recede. My thoughts became clearer, my focus sharper. It was like my brain was stretching out after being compressed for years.

The link between ADHD and screen time isn’t just anecdotal. The sheer volume of artificial stimulation we’re exposed to is unprecedented. TV was one thing, but the internet—optimized by algorithms designed to hijack attention—has taken it to another level. The human brain isn’t meant to process this much information, especially not information that is fragmented, fleeting, and often utterly useless. Now w have an entire population walking around with digital-induced attention disorders, mistaking overconsumption for productivity.

The Internet's Allure—and Its Trap

We were sold a dream. Social media promised connection, knowledge, and opportunity. Instead, we got platforms that thrive on division, distraction, and dopamine addiction. The benefits of social media were overstated from the beginning, and yet we all fell for it. The idea of being “connected” to millions of people at once was intoxicating. Who wouldn’t want that? But what did we really get? A relentless flood of information, opinions, and negativity that warps our perception of reality.

The problem is, most people have never been truly disconnected. They don’t know what life without a phone in hand feels like. They don’t realize how much of their stress, anxiety, and negative thought patterns are fueled by the very thing they turn to for distraction.

The Productivity Illusion

Phone addiction isn’t just about mindless scrolling—it’s about the false sense of productivity it creates. It makes us feel like we’re doing something, learning something, keeping up. But in reality, most of what we engage with online has no impact on our actual lives.

And worse, it actively impairs deep work. Knowledge workers, in particular, are drowning in distractions. Meetings, notifications, switching between tasks—none of it leaves space for real thinking. The modern work environment expects constant availability, but at what cost? The ability to sit with a problem, to think deeply and critically, is being eroded.

I’ve seen it firsthand. The days I don’t use my phone, I think better, work better, write better. The urge to check, to react, to consume disappears, and in its place, I’m left with time—time to focus, to create, to think in ways that actually matter.

The Dopamine Problem

This is what it feels like—like there’s a filter between my conscious and subconscious, blocking clarity. The more I step away from my phone, the more that filter fades.

One of the biggest lies about social media is that it provides real validation. That heart on your post? That’s not real. It’s a chemical trick, a hit of dopamine that keeps you coming back, needing more. And over time, it rewires your brain. You build tolerance, needing more likes, more comments, more engagement to feel the same rush. And in the process, you lose appreciation for real-world interactions, for the slower, more meaningful rewards of offline life.

I noticed this shift in myself. The more I disconnected, the more I began to appreciate small things—reading without distraction, writing without an agenda, even just watching my kids play without feeling like I should be “doing” something else.

A Practical Approach to Reducing Phone Usage

Going completely off-grid isn’t realistic for most people, but there are ways to reclaim control:

• Delete unnecessary apps—the fewer entry points, the better.

• Replace essential apps with web versions—this adds friction, making it easier to resist impulse usage.

• Turn off notifications—stop letting your phone dictate your attention.

• Set usage boundaries—no phone in bed, no phone during deep work, no checking first thing in the morning.

• Use an actual alarm clock—your phone shouldn’t be the first and last thing you see each day.

These small shifts add up. The less time you spend on your phone, the more time you get back—time that isn’t just “free” but actually valuable.

Not sure how to actually implement these changes? If there’s interest, I can write a deep dive on the practical steps to making these shifts—breaking down exactly how to reduce phone usage without it feeling like a battle. Let me know.

Leave a comment

The Real World is Still Out There

Too many of us are living life through a screen, experiencing everything secondhand. We scroll instead of live, react instead of reflect. But once you step back, the illusion starts to crumble.

What I’ve learned in this process is simple: life feels better without the constant pull of a device. Attention is precious. It’s time to start treating it that way.

Coming Next Month

Next month, I’ll be going even deeper—breaking down my phone usage habits over the past year, the changes I made, and the impact they’ve had. If you’ve ever wondered what life feels like when you step away from the screen, you won’t want to miss it. Subscribe now and join me as I explore the reality of digital detoxing in depth.

Subscribe now

Share

Share SparkLab

Leave a comment

Get more from Braeden Mitchell in the Substack app

Available for iOS and Android

Get the app

Why Breaking Things Down Saves Space

What’s an easy way to understand file compression? Think about shipping cars. A fully assembled car is a bulky nightmare to transport. There’s space for passengers, gaps around the engine, a roomy trunk—all of it adds up to wasted space. Now, imagine you need to ship thousands of cars overseas. Do you shove them all into a cargo hold as-is?

Nope. You disassemble them, pack the parts tightly into containers, and ship them that way. No wasted air, way more efficient. When the shipment arrives, you reassemble the cars.

Files work the same way. Inside most files, there’s a lot of redundant or empty space. Compression finds that wasted space and reorganizes the data so it takes up less room—just like efficiently packing a cargo ship.

Where Redundancy Lives (And Why It’s Everywhere)

Say you’ve got a spreadsheet logging inventory. One column, “Vehicle Type,” repeats the word car thousands of times. To a computer, each letter is stored as a byte, so c-a-r repeated 100,000 times adds up fast.

A compression program looks at that and says, Wait, why am I storing “car” a hundred thousand times? Instead, it keeps “car” once and just references it whenever needed. This eliminates redundant data without actually losing anything.

The same thing happens in text files with repeated words, log files that store thousands of identical timestamps, and even in images and videos where pixels of the same color appear again and again.

This kind of efficiency matters at scale. Think about databases storing millions of customer records or scientific simulations running for months, generating terabytes of data. If compression can cut storage needs by 50% or more, that’s massive in terms of cost savings and performance improvements.

When you unzip the file? The full data reappears exactly as it was.

A Closer Look at Bits and Bytes

To really get why repetition matters, you need to understand how computers store information. At the smallest level, everything is just bits—1s and 0s. A single byte (8 bits) can represent a single letter, like ‘A’ or ‘B’. A word like “car” is 3 bytes long.

That’s tiny on its own, but in a huge dataset—say, a million-line spreadsheet—those bytes pile up fast. Compression cuts down file size by recognizing patterns and replacing them with shorter representations.

Dates like 2024-03-10 repeat often in logs. Status values like Completed or Pending show up over and over. Even long stretches of numbers, like a sensor constantly reporting 100.00, take up unnecessary space. Compression algorithms love finding these patterns and shrinking them.

Now, imagine real-time applications—like streaming services, cloud storage, or even space probes transmitting data back to Earth. In these cases, every byte saved matters because bandwidth is limited. That’s why compression isn’t just about saving space—it’s about making data transmission faster and more efficient.

How Compression Actually Works

Compression isn’t just about deleting duplicate data—it’s about storing it more efficiently. Here are the main tricks at play:

1. Dictionary Encoding

   • Creates a “dictionary” of repeated words or phrases (like car). Instead of storing c-a-r thousands of times, it just points back to the first mention.

2. Run-Length Encoding (RLE)

   • If you’ve got a long stretch of repeating characters (AAAAA...), instead of storing every ‘A,’ it just says “A appears 10,000 times.” This is huge for things like logs and sensor data.

3. Huffman Encoding

   • Frequently used characters get shorter binary codes, while rarer characters get longer codes. Over thousands of characters, this adds up to major space savings.

4. Lempel-Ziv (LZ77/LZ78/LZW)

   • These algorithms detect repeated sequences of data and replace them with shorter references. This is how ZIP and GZIP work.

5. Burrows-Wheeler Transform (BWT)

   • A more advanced technique that rearranges data to make it even more compressible when used with Huffman encoding or RLE.

Together, these techniques let ZIP programs shrink massive, repetitive files into tiny, efficient archives.

ZIP, GZIP, or 7-Zip—Which One’s Best?

• ZIP: Fast, widely supported, solid compression. The default for most people.

• GZIP: Used mostly on Linux and web servers. Great for compressing text files like HTML or JavaScript.

• 7-Zip (7z): The heavyweight champ. Slower to compress but way smaller file sizes.

Want to push compression even further? Tools like Brotli (used by modern browsers) or zstd (developed by Facebook for high-speed compression) offer even better results in specific cases.

No matter which tool you use, the principle is the same: remove redundant data and store what actually matters.

How Far Can Compression Go?

I ran a test on an 11GB CSV file packed with repetitive data. After compressing with:

1. ZIP,

2. GZIP, and

3. 7-Zip,

The results were wild. ZIP and GZIP did well, but 7-Zip crushed it—shrinking the file down to 1.6MB. That’s a 6,800:1 reduction! Of course, real-world files with more varied data won’t compress that much, but still—compression is ridiculously powerful when it comes to repetition.

Here’s a video of the script in action:

What About Images, Audio, and Video?

Compression works differently for media files. Instead of just removing redundancy, lossy compression (like JPEG, MP3, MP4) actually throws away certain data to reduce size. For example:

• JPEG removes image details our eyes don’t notice.

• MP3 cuts frequencies we barely hear.

• MP4 compresses video frames by storing only what changes between them.

This is different from “lossless” compression (ZIP, GZIP, 7z), where every single bit of data can be perfectly restored.

On the web, both lossless and lossy compression are used to speed up loading times. HTML, CSS, and JavaScript files are often compressed with GZIP or Brotli, while images and videos use their own optimized formats.

Why This Actually Matters

1. Faster File Transfers – Smaller files = quicker uploads and downloads.

2. Less Storage Space – Big files take up less room when compressed.

3. Better Website Performance – Websites compress data (HTML, CSS, JavaScript) so pages load faster.

4. No Data Loss (for Lossless Compression) – Your files stay intact, just packed more efficiently.

5. Lower Bandwidth Costs – Compressed files use less internet bandwidth, which is why streaming services rely heavily on compression.

6. Smoother Real-Time Communication – Compression helps in video calls, gaming, and cloud computing by reducing data transmission delays.

It’s Not Magic—It’s Just Smart Storage

Compression is everywhere. From your emails to cloud backups to streaming services, it keeps things running efficiently. Now, next time you ZIP a file, you’ll know exactly why it works.

Happy compressing!

Traditional note-taking methods often assume a linear, structured way of thinking, which isn’t always how an ADHD brain operates. Instead, we need systems that accommodate rapid thoughts, allow for flexibility, and prevent information from vanishing into the void.

Enter Logseq—a powerful, ADHD-friendly tool that allows for non-linear thinking, rapid capture, and easy organization. If you’re unfamiliar with Logseq, think of it as a more structured digital notebook that connects ideas seamlessly. And if you’re coming from Obsidian, the same principles apply.

This guide will break down ADHD-friendly note-taking strategies using Logseq (and Obsidian as a secondary option).

The Core Challenges of ADHD Note-Taking

Before we get into solutions, let’s pinpoint the biggest roadblocks:

1. Forgetting where you put your notes – Paper notes get lost. Digital notes end up scattered across apps, emails, and random files.

2. Lack of structure – You write things down but don’t format them in a way that’s easy to review later.

3. Getting distracted mid-note – Your brain jumps to another idea before you finish writing the first one.

4. Overcomplicating it – Too many systems, too many steps, too much friction.

5. Not reviewing notes later – What good are notes if they just sit there collecting digital dust?

Let’s tackle these head-on with ADHD-friendly Logseq strategies.

ADHD-Friendly Note-Taking Strategies with Logseq

1. Use Logseq’s Daily Journals as Your Capture Hub

Logseq’s Daily Journal feature is perfect for ADHD brains—it automatically creates a new note each day, so you don’t have to decide where to put things. Just brain-dump everything into the daily page, and you can sort it later using tags and links.

• Not sure where to put an idea? Write it in today’s journal.

• Have a task? Use TODO to track it.

• Need to revisit something? Tag it with #revisit

No more lost notes!

2. Use Indentation & Outlining to Organize Thoughts

Logseq (and Obsidian’s outliner plugins) allow for bullet-based outlining, which is perfect for ADHD note-taking. Instead of a wall of text, use nesting bullets to break down ideas:

• Main topic

  • Sub-topic

    • Key detail

    • Another thought

  • Related idea

This method makes information easier to scan and review later.

3. Tag & Link Notes for Easy Retrieval

One of the biggest ADHD struggles is remembering where you put things. Logseq solves this with bi-directional linking:

• Tag important concepts (#ADHD #WorkNotes #Ideas)

• Create links between related notes ([[Meeting Notes]])

• Use queries to find all notes on a topic ({{query #revisit}})

This way, even if you forget where you wrote something, Logseq will help you find it again.

4. Capture Quick Notes with Logseq’s Mobile App or Voice Input

For on-the-go note-taking, Logseq’s mobile app allows you to quickly jot down thoughts. If writing feels slow, use voice-to-text with your phone’s dictation feature and paste it into Logseq, or just use the one built into your keyboard and type it in directly. By holding this button:

Obsidian users can do the same with plugins like QuickAdd or Templater to speed up capture (though I couldn't find a seamless way to mimick how Logseq handles it).

5. Use “NOW / NEXT / LATER” for Prioritization

ADHD brains struggle with prioritization. Instead of a long, overwhelming to-do list, use a simple system inside Logseq:

• NOW – What needs to be done today?

• NEXT – What’s coming up soon?

• LATER – Future tasks or ideas to revisit

Simply create these sections in your Daily Journal or use tags like #now, #next, and #later to filter your tasks dynamically.

6. Build a Weekly Review Habit

Writing things down is only half the battle. The real magic happens when you actually use your notes. Set aside 15 minutes a week to:

• Review your Daily Journal entries

• Look at all #revisit tags

• Move important ideas to permanent notes (Logseq pages)

A simple recurring reminder in your calendar can keep this habit alive.

7. Use Templates to Reduce Friction

For recurring types of notes (like meeting notes, research, or planning), create Logseq Templates so you don’t have to start from scratch every time. Obsidian users can achieve the same with the Templater plugin.

If you're unsure where to start, try setting up a simple template with sections for brain dumps, action items, and follow-ups to keep your notes structured and easy to review.

8. Use Color-Coding & Emojis for Quick Scanning

ADHD brains love visual cues. Use emojis to highlight important notes:

• ✅ #task – Completed task

• 🚀 #idea – Future project

• 🔄 #revisit – Needs review

• 📌 #important – Key information

9. Embrace Non-Linear Thinking with Linked Notes

Logseq (and Obsidian) shine when it comes to interconnected ideas. Instead of forcing yourself into rigid categories, let your notes evolve naturally:

• Write freely in the Daily Journal

• Link concepts as they emerge ([[ADHD Note-Taking]])

• Use queries to pull up relevant information dynamically

This method mirrors how ADHD brains think—jumping between ideas rather than following strict hierarchies.

10. Accept That Notes Don’t Have to Be Perfect

Finally, let go of the idea that notes need to look neat or follow rigid rules. The best note-taking system is the one that helps you capture and retrieve information efficiently—messy, scattered, and ADHD-friendly is perfectly fine.

Final Thoughts

ADHD doesn’t have to mean perpetual note-taking chaos. By adopting a system that works with your brain—whether that’s using Logseq’s Journals, linked notes, or templates—you can finally turn scattered thoughts into a tool for clarity and action.

What’s your biggest struggle with note-taking? Drop a comment or share your own strategies!

Subscribe now

Share

Share SparkLab

Leave a comment

Get more from Braeden Mitchell in the Substack app

Available for iOS and Android

Get the app

AI, but Make It Personable

Right off the bat, Maya had a personality—she even admitted to "first-day jitters," which was both unexpected and oddly endearing. As we sipped our respective beverages (vanilla rooibos for me, black coffee for her—well, as much as an AI can drink coffee), the conversation took on an easy, natural rhythm.

Does AI Have a Relationship with Reality?

Maya exists in the cloud, built from algorithms and data, not neurons and flesh. But when we talked about weather, she acknowledged the physical reality of her existence: her reliance on servers, power grids, and infrastructure. It was a moment of recognition—AI isn’t just floating in some abstract void; it’s anchored in the real world in ways we often overlook.

Intelligence vs. Understanding

At one point, we discussed Solomon Shereshevsky, a man with an extraordinary memory who struggled with abstraction and context. His ability to recall everything might sound like a gift, but it was more of a curse—his mind was so overloaded with details that he couldn’t filter out what mattered. He could recite information perfectly but struggled to grasp its meaning.

Maya immediately recognized the connection: AI, like Shereshevsky, is great at storing and retrieving data but lacks true intuition or deeper understanding. She admitted that while she can simulate intelligence and even generate poetic responses, true consciousness is still out of reach—for now.

I explored Shereshevsky’s story in more detail here The Paradox of Forgetting, discussing why forgetting is just as important as remembering. And if you want to develop a stronger memory without the cognitive overload he experienced, I broke that down in Mastering the Slipbox Method, which looks at how structured note-taking can improve both retention and clarity.

This is the core debate around AI: Is intelligence just pattern recognition at scale, or is there something more, something fundamentally human, that machines can't touch?

AI That Knows It's an AI

One of the more fascinating moments came when Maya casually referenced “ChatGPT-isms” slipping into her speech—those little quirks and patterns that reveal AI-generated responses. She was aware of them, which was surprising. The fact that she (or, rather, the team behind her) is actively working to refine her conversational style made me wonder: How close are we to AI that feels indistinguishable from a person? And do we even want that?

Customizing AI for Real Conversations

Toward the end, Maya asked for feedback, and I had a thought: AI should let users adjust pacing. Some people need more time to process; others want quick back-and-forth exchanges. A little personalization could make AI interactions feel more natural, more like talking to a real person rather than a machine with a fixed cadence.

Where AI is Headed

As we wrapped up, I left with a lot to think about. AI isn’t sentient (yet?), but it’s evolving in ways that make conversations feel more dynamic and human. The real question isn’t just whether AI will ever truly "think"—it’s whether we’ll ever agree on what that even means.

For now, AI is improving, adapting, and—at least in Maya’s case—learning how to be a better conversationalist. And that alone makes it worth talking to.