Загрузка видео...
Не удалось загрузить видео
Skill atrophy with coding agents
150,415 просмотров • 6 месяцев назад •via X (Twitter)
Комментарии: 28

Relevant paper from Anthropic: How AI Impacts Skill Formation (Feb 3, 2026)

Yep, but shoveling skills is not that relevant when you can use excavator.

Anthropic should release a feature for us to maintain our skills

Would be great, actually

Your programming skills definitely do get worse. But you should be focusing on improving your problem solving and critical thinking.

great visualization

Is it remotion?

Yes

I’m damn sure its not one prompt? Its lot of iterations?

Yep, lots of iterations, throwing away certain ideas, many different chats - around 4 hours of uninterrupted work

Ohh, i wish model can one shot this kinda of stuffs so that i can just spent few minutes or hours and create launch video like this for

Jokes on them. I had no skills to begin with anyway.

Recently wrote about something similar.

yeah this is real. i noticed my syntax recall getting worse even on languages i was solid in. what i dont buy is that all skill atrophy is bad. some of what is atrophying was always busywork

Yep, it's not about syntax recall. The problem is that it's easy to stop acquiring new skills and mental models when working with agents

yeah this is the real one. the skills i fight to keep sharp are the ones agents still cant do reliably. cross tool debugging when state is split between prompt llm and infra. that one atrophying hurts fast

For learning purposes hand-code at 1x, for production, use agents to code at 10x

It's just different skills

Haha yes! Tell em Alex, and I- wait. DID U VIBE CODE THIS.......!! 🤭

It all depends on how you use. If you let ai make all decisions and not design the system yourself you gonna lose the design thinking.

Who cares when you can just snapshot skills to be re-used like abilities in a game whenever you want? Its the next step of the abstraction of knowledge.

it's so cursor-coded animation

It’s active versus passive learning active learning. We actually hit a wall and we work our heads on it until we understand and fix it and passive learning is like watching YouTube videos. We just get a brief understanding of the explanation and we just accepted ourselves and we don’t challenge ourselves so just active versus passive pretty much.

I know how computers work, i don't need to know the specifics of the framework of the year anymore, and I was always a waste of my time

PAR solves this, for free.

What is happening?

Valid concern. The fix is using agents as accelerators, not crutches - delegate tasks but stay involved in architecture decisions and code reviews.

I thought people learn to code to change the world or make money, didn't know they just want that skill.

