Tell HN: If working with agents means this, robots, please take my job

“We found that the phenomenon described in these posts—cognitive exhaustion from intensive oversight of AI agents—is both real and significant. We call it “AI brain fry,” which we define as mental fatigue from excessive use or oversight of AI tools beyond one’s cognitive capacity.”

https://hbr.org/2026/03/when-using-ai-leads-to-brain-fry

That's something I haven't seen discussed much, what is the "annoyance-cost" of using the tool?

Say I complete a task 30% faster but what was the annoyance-cost of the model constantly getting it wrong and having to keep correcting it

AI is to Software what CNC was to machinists. It's turning it into something more commoditized and less tilted in favor of skilled labor. The more you actually use these AI setups, the more training data you're giving these companies, the faster everyone gets obsoleted.

And make no mistake, it's entirely possible at some point they are storing outputs for being trained on to replace your Org's business too. Selling the inference agent is what's paying the bills in theory right now, that may change, and the next logical step is cannibalizing customers.

And eventually they will raise the price, do all the things that other companies have done.

I figured computers that can run AI models locally are the new workstations in the near future. Otherwise pay a premium to rent AI.

There is no secret sauce in what most companies do and could easily be duplicated from the business process angle by any other company with sufficient resources.

I’m not saying any company could actually overcome incumbency advantages.

I don’t believe Antrophic or OpenAI would be dumb enough to train on a company’s proprietary data if you had an enterprise agreement or even if you explicitly opted out.

I wouldn’t trust Meta or Grok though…

I think you feel that when you fell in love with the process and not the outcome. feel bad for devs who enjoyed coding, though the ones who prefered shipping products over coding are living their best life right now, not for long though.

> but the mental load is huge and they don't feel able to be honest with leadership about it

There will get used to it very fast + UIs are going to become much better very soon.

I feel the same. After February when we got our team access to claude I 100'xd my work, but felt what I'm now calling agent fatigue.

Honest question what were you doing that could be 100xd

That did the companys revenue / profit significantly change in some way?

Even at a large company if a person 100xd their productivity that would meaningfully move the needle.

I'm going to keep pointing this out because insane claims require insanely strong evidence.

If you 100x your work since the start of the year you would have produced a lifetimes amount of output. Have you made 7 successful SaaSes in the last month? Or whatever field you work in?

Even 10x in the last month you would have produced a year's worth of work.

Have you?

Or have you just done what you would have normally done in a month with a few side projects no-one's going to use and will be serious tech debt in 6 months time?

I’m thinking about getting into Robotics given the opportunity. But it’s a vast gap so not sure if it’s going to work. We will see.

My coworkers and I have an informal agreement to let a few production issues show up until leadership changes their toon.

Congratulations they'll just replace you with another human before they replace them with AI.

You've added a step.

Hey, if you want me to be an AI's manager, I can only give you a manager's level of certainty that a task is done well :) if I could fire the AI that created the bug I would, the best I can do is Ctrl+c. But either way, by the time I'm fired the codebase will be an incomprehensible mess and any new engineer won't be able to provide a good review for months. Hopefully they don't decide to fire everyone all at once.