Recently stumbled upon a stat: 55% of companies that laid off people because of AI now regret it. And a METR study showed something weird - developers think they work 20% faster with AI, but actually turned out to be 19% slower.
Then I read Hinton saying that in a couple years AI will do in minutes what takes an engineer a month. And AWS CEO responds calling the refusal to hire juniors “one of the dumbest things I’ve heard.”
So who’s right?
My experience says - both are wrong. AI doesn’t replace people. And it doesn’t slow work down. It just changes what we do.
What I’ve Given to AI
Honestly - almost all analytics. Draft diagrams, text editing, searching and analyzing information. I’ve practically stopped coding by hand - always work with an agent now. Need some expertise - AI is my first stop.
Sounds like I’ve given everything away. But no.
What I Won’t Give Up
Health. This is clear for me. I don’t mind if a doctor uses AI - actually, that’s good. But it should be a doctor. With education. With experience. AI as a tool - yes. AI instead of a doctor - no.
Same with a psychologist. I like working with a real person, and I don’t see how that can be replaced. Same with a coach.
Everything related to my health, consciousness, awareness - only to professionals. Who can use AI in their work, but professionals nonetheless.
About This “Productivity Paradox”
That METR study got me thinking. People think they’re faster, but actually slower. A paradox, they say.
I disagree.
How did it used to work? You could try one or two implementation options. Picked something, went ahead. Hit problems - made a third version. Then fourth. And it all piled up. First version lying around somewhere, second, third - and you’re already on the fifth, dragging legacy behind.
Now I try a bunch of options at once in the same time. Yes, each specific option takes more time. But I’m not dragging three failed attempts behind me. I pick the best option before I start burying it in production.
That’s not slowing down. That’s a different way of working.
Juniors
AWS CEO put it well: “How’s that going to work ten years from now when nobody’s learned anything?”
Completely agree. There has to be knowledge transfer. Generational turnover of specialists.
Yes, with AI you can make websites and telegram bots without much education. But the industry isn’t just websites and bots. There’s tons of stuff that needs higher math. That needs computer science. Juniors are needed. And they need to be trained.
How My Role Has Changed
Writing code by hand - no point anymore. Now I think about architecture. Design systems. Do reviews. Validate what came out.
And the design itself is also with AI - you talk to the agent, break down the task. Then hand it off for implementation and check the result.
Basically, the role has shifted. Before - executor. Now - designer and validator.
What’s Become More Important
Systematic approach. These stories about the “hero developer” who fixed everything at night, or the “firefighter” who’s everywhere at once - that’s going away. People who solve problems systematically are valued.
And another skill has emerged: building your own system for working with AI. Just dictating a prompt - doesn’t work. You need an actual system. Your own prompts, rules, skills, agents.
I spend time not just on work tasks, but on this system. It’s my asset. My intellectual property, if you will.
What’s the Point
AI doesn’t replace specialists. It changes what they do.
55% of companies regret layoffs not because AI doesn’t work. But because they didn’t understand - people’s role has changed, not disappeared.
Less execution, more design. Less heroics, more system. Less hand-coding, more work on your own tools.
That’s how it looks right now. At least for me.