Reset-Oriented Development

12 Sep '26 llm

Other posts

Or: your life isn’t just about LLMs


the xkcd 'compiling' comic but with 'I ran out resets' instead of 'my code is compiling

It is getting rather frightening that a lot of my friends working in Software Engineering legitimately organize their entire work day around AI usage resets.

Start working at 9. Burn through the 5-hour limit in 3. Have lunch or play Genshin Impact (or whatever the kids play nowadays) for 2 hours. Burn through the 5-hour limit again. Pop a reset. Burn through the 5-hour limit again. Switch between Claude Code and Codex. Limit again. Grab dinner. Scroll through Twitter to see if Tibo is about to reset limits again. If he does, start up some random bullshit slop side project with max thinking effort and fast mode to burn through the entire weekly limit in one go and wait for the reset to drop. Rinse and repeat.

As a fairly traditional coder with very high standards for code quality (and as a cheapskate), I just don’t get the mindset behind this. It’s all fucked up in so many different ways.

It’s like going to a Pizza Hut salad bar and then stacking up an actual tower of salad pieces on your plate because you’re only allowed one bowl per visit, and then planning your day around when you can go to the salad bar again. If the salad bar tells you to piss off because you’ve had too much, then you go to a different restaurant with a different salad bar. Is the salad bar really this important to you? How about doing or eating literally anything else?

I get that a lot of vibecoders are ready to sacrifice their own convenience and sanity if it means they get to produce more slop. Like trying to learn Chinese since prompting in it allegedly saves a lot more tokens than in English (spoilers: it doesn’t), or reorganizing LLM usage to avoid peak pricing when using DeepSeek V4 Pro. But if anything, this kind of self-inflicted flagellation only impacted the vibecoder themself; we otherwise ordinary folk don’t get to suffer because of it.

But now things are rather different. I now have to face the real possibility of friends not going out for dinner because they’re trying to squeeze the last bit of tokens out of their OpenAI or Claude subscription. Employers have to tolerate employees slacking off because they can’t possibly code without LLMs in an entirely vibed codebase. Entire departments at the mercy of AI corporations. Our own use of time and free will bound to the whims of one man on Twitter. Many don’t even have a choice as some “innovative” companies mandate using LLMs at work, so they’re trapped in this cycle without a way out (aka actually writing code).

Maybe then the actual form of cybersubjugation humanity will endure would not be an outright malicious stick à la Skynet, but a seemingly innocuous carrot that dulls and deprives us of the ability to create, to work, to breathe life; our skills atrophied to such a point, that we are no more than simpletons bound to a chain, consuming mass-fabricated media unquestioningly to find a semblance of the life driven by thought and creation that we used to lead.

But perhaps that is overly pessimistic. After all, any spin on the myth of infinite growth is bound to be just that—a myth. Soon, someone is bound to find out what the real cost of behind all of this really amounts to.

In the mean time, don’t let your mind decay.

To vibe, or not to vibe, that is the question:
Whether ‘tis nobler in the mind to suffer
The slings and arrows of outrageous prices,
Or to take arms against a sea of errors
And by handcoding end them? To stop, to sleep
No more; and by a sleep to say we end
The heart-ache and the thousand unnatural slop
That AI is heir to, ‘tis a consummation
Devoutly to be wish’d.

 — Hamlet, Act III, Scene I (modernized 2026)