No AI Fridays
TL;DR
A one-page campaign site asks developers to switch off AI assistants one day a week and write code by hand. It cites studies linking LLM use to cognitive debt, weaker critical thinking and poorer skill formation. It argues that if AI's productivity gains are real, one day without it costs little.
The article
- The site opens with four studies claiming LLM use builds cognitive debt, lowers engagement, hurts critical thinking and slows skill formation.
- Handing decisions to an AI hides the trade-offs. A weekly day without it is a chance to review the choices the AI made and check they still match your own style.
- If the productivity gains are as large as claimed, giving up one day a week to limit the downsides should be an easy trade.
- Defaulting to AI skips ordinary automation. A day without tokens can cut token spend over time. The site also appeals to the fun and flow of writing code yourself.
- The setup is to send the link to your boss, turn off assistants for the day, read the docs and think problems through. Review bots like greptile stay allowed, because you learn from their feedback and the LLM doesn't.
- The tone is tongue-in-cheek. The author claims to be "CEO of HTMX", a running htmx joke. The FAQ answers a request for a quick peek with a joke about "a pint of Codex" while trying to quit.
Why it matters
The page takes two minutes and is mostly a pitch. The useful part is its links to the four studies. One commenter checked them and found them weaker than the site claims, so read the papers before quoting the page to a manager.
The discussion
A heated, split thread. Most of it argued whether leaning on AI erodes a developer's skill, with the calculator analogy fought over in several places. One commenter's claim of 10-20x output and a $5M AI film business drew a long run of hostility. It also drew an unusually candid explanation of where that hostility comes from.
AI lets you build without learning
The largest group said AI breaks the link between making something and learning how. Students and juniors were the main worry: one commenter's recently graduated brother can't answer first-year CS questions, and another called AI "eternal tutorial hell". Hobbyists pushed back that they now finish music, CAD and statistics projects they could never attempt. Asked whether that knowledge sticks, one admitted it doesn't, and wondered whether the output matters more.
If you don't have AI then either you build what you need to build and learn how to do it along the way, or you build nothing and learn nothing. There is no option to build something but learn nothing, but that's *the default outcome* with AI.
I have found in the last year I "get more done" but I learn less. Whats even worse is if I burn though my token quota I am less likely(and less able) to "switch to manual".
Calculators, compilers and the abstraction ladder
Skeptics of the site listed past warnings that a new abstraction would rot skills, about slide rules, assembly and garbage collection. Replies said those warnings came true: GC languages produced developers who can't reason about memory, hence 1GB form apps. Others said a calculator still needs you to know what to enter, while an AI can skip the understanding entirely. A pilot anecdote got support: pilots fly manual landings regularly to keep the skill, though autopilot can land.
You don't need to "keep practicing" in order to maintain your understanding.
Big difference is when you use a calculator or other tools, you still have to interpret and know what to enter.
Flow state: lost or found
Several developers said coding agents killed the flow state. Waiting on an agent invites distraction, and flow needs the whole program in your head. Others get flow more often now, running agents in parallel and working at the planning level; one described a 12-hour session that felt like two. Flow purists answered that constant context switching is by definition not flow.
What I've realised I missed the most is feeling in the zone or in a "flow" state as some call it. It really feels amazing. Coding agents take that away from you.
With AI, I can achieve this state every weekend for personal projects, and every day I'm WFH or when I'm alone in my corner of the office.
The 10x enthusiast and the backlash
The top thread's longest subthread followed a commenter who claimed 10-20x productivity, several games, and an AI filmmaking tool with $5M ARR. Replies called his product slop, suggested a manic episode, or cast him as a crypto-style True Believer. When he asked why people treat AI users this way, one reply listed employers mandating LLM use, artists losing paid work, and many social harms. Another said his success rests on expertise built without AI, which newcomers won't get.
In my case, my employers have mandated LLM usage and all but banned actually programming, and I can see it rotting my ability to code and understand things.
Who is actually deciding
Some agreed with the site that offloading decisions hides the trade-offs. One said picking from options an agent presents isn't deciding, because you never see the option it left out. Others said agentic work is mostly decision-making and only lazy users skip it. A few preferred a daily rule to a weekly ban: code by hand when explaining is harder, and use AI as a tutor.
Choosing from a set of decisions the agent presents you might feel like you are making a decision, but you’ll never make the one that wasn’t presented.
But if you use it seriously, you're in a constant state of doing requirements engineering, weighing trade-offs, and making architecture decisions.
Consensus
Learners should build without AI. Anyone who wants to keep hands-on skill needs to practice it on purpose, the way pilots practice manual landings.
Dissent
Many rejected the premise. Some said every abstraction drew the same warning and the industry adapted. Others objected to any mandate, whichever direction it points, or said a little manual work every day beats a whole day off.