There's no point at which turning your brain off will work
TL;DR
Dan Luu argues that acting as a brain-off "meat proxy" for an LLM can't pay off for an employee. While the LLM can't do the job alone, the result is bad software. Once it can, the company has no reason to keep the proxy.
The article
- Since early 2025 Luu has seen people hand tasks to an LLM and assume the output works. Others sit in a loop and paste failures back to it. Niklas Gruhn calls this being a "meat proxy".
- Being a meat proxy works better in September 2026 than it did in early 2025. If LLMs improve enough to make it produce decent software, the company can run the loop itself and lay off the human.
- LLMs still do badly on out-of-distribution work. On modern board games like Dominion, a frontier model plays worse than a sensible human who has never played. ChatGPT's advice made a new player worse.
- Software built by people who outsourced the thinking usually doesn't work. A thought leader said programming was solved, but every example Luu checked on his GitHub was broken or weak. Gary Bernhardt reports cutting LLM diffs to 25% of their size.
- LLM fact checks of his posts have poor precision and mediocre recall. Using an LLM's praise to justify your work is a fairly strong sign the work is poor.
- Waiting out a window of corporate inertia is also a bad bet. Companies once took months or years to fire people who stopped showing up. LLMs are now their excuse to stop tolerating that.
Why it matters
The main text is four paragraphs, and almost all the substance sits in long footnotes. Read the original for Luke Burton's account of a Bazel migration that took months and needed constant supervision. It also has the board-game experiments and Luu's stories about how long companies took to fire people who stopped working.
The discussion
The thread mostly agreed with Luu, and many developers said the brain-off push comes from management, not from them. Others debated whether a human stays useful as a holder of blame, where the thinking belongs, and whether jobs disappear anyway. A long side argument over a photography analogy took up the single largest thread.
Management is forcing it
The most personal replies said engineers aren't choosing to switch off: bosses push full automation and get angry when questioned. One commenter books a meeting room to code by hand, so his lead won't see him working without Claude. Several people said they are resigning over it. Others blamed executives who bought a headcount-reduction pitch, and some saw no escape in changing jobs because investors want the same thing everywhere.
Even if I agree that you have to use your brain, when my boss is literally telling me to stop questioning whether AI automation is the right answer in team meetings/chats I'm just going to shut up and give them what they want.
What they care about is that the actual work is being done by humans, and that's out of line with their grand vision of getting rid of humans.
The human as liability sink
Several people answered Luu's question about why a company would keep the proxy with "to take the blame". They called the role "compliance ablative armor", an "AI patsy", a moral crumple zone and an accountability sink, like backup drivers in self-driving cars. Nobody called it a good job. One veteran engineer would rather retire than sign off on generated work he can't review to his own standard.
The human's purpose is to keep an eye on the LLM, occasionally intervene/interfere with its work, and most importantly, incur social/legal liability for anything that the LLM does wrong.
When the new plan is to increase throughput via delegation to these generators, I can't possibly review to my standards. So, I refuse to accept any of it as my work.
Where the thinking goes
The top thread argued about when to use your brain. Arubis said to think before the prompt, since the loop biases you toward its options, and to stop the loop when mediocre isn't good enough. xendo answered with Programming as Theory Building: you only understand software by building it. Others said from practice that most tasks need little effort, a few need real design, and telling them apart is hard.
I guess what I'm ultimately getting at is that the loop is useful _and_ a mind-killer, and using your mind is _also_ high value, so do both those things, but don't do them at the same time.
So I've been moving away from plan mode and letting the spin away on building large chunks, and back toward working in small increments and typing out the important code with my own hands. That seems to be the sweet spot.
Many shots versus one good one
dzink compared LLMs to a tourist firing off thousands of phone photos beside a photographer setting up one shot. Filtering many attempts, dzink said, beats overthinking. Most replies rejected this: volume without the eye to judge it doesn't win, and cherry-picked mass-produced movements killed a watchmaking contest. btrettel added that Luu's brain-off users take one or two shots, say LGTM, and move on.
I can take as many thousands of pictures as I want but, without taking the time to critically review them to see what works and what doesn't, I will rarely produce a prize-worthy picture.
They take a couple photos at most, say "LGTM", and move on to taking a photo of something else, even if their couple of photos of the sunset aren't great.
Will engagement even save the job?
Some took Luu's logic further than he did: if the goal is an LLM that beats humans at everything, engagement won't keep the paycheck either. Whole companies then become thin proxies their customers can cut out. One commenter cited a nearly 20% drop in employment for developers aged 22 to 25 since late 2022; another blamed the end of cheap money. Skeptics said LLMs aren't replacing people at all; managers were only sold a product they believe can.
Even if you refuse to turn off your brain, the economics of paying you to keep the lights on upstairs aren't going to work out.
Anyway, it's just a single occurrence of a hiring drop in a single economic cycle. Unfortunately, it just happens to perfectly line up with the rise of GenAI.
Consensus
Turning your brain off produces bad work. The useful human contribution is judgment: deciding what to build, catching the out-of-distribution case, and cutting the diff down. People who keep using their judgment are worth more than proxies.
Dissent
A sizable minority thought engagement won't protect anyone, because companies are cutting staff and hiring fewer juniors either way. Others said brain-off use is fine for throwaway or creative work where shoddy code doesn't matter.