אל תעשה את זה עם כלי ה-AI של הקידוד שלך!

Last year two studies came out that seemed to completely contradict each other. One said AI made developers way faster. The other said experienced developers actually got slower when they used AI tools in real codebases. So what's going on?

I think both studies are right. And the gap between them explains why so many frontend projects that start with AI end up taking longer than if you'd just written the code yourself. It's not that AI is bad at generating code. It's that most workflows around AI are backwards.

In this video I walk through five specific ways this goes wrong, especially on the frontend where problems show up fast. I'm talking about the stuff that looks fine in the editor but breaks when real users hit it. The kind of code that passes your linter but creates a maintenance nightmare six months later.

If you've ever generated a component that looked great until you actually tried to use it, or spent more time fixing AI output than it would've taken to write it yourself, this one's for you.

Check out my last video:
https://youtu.be/VKk-l97CNU8?si=imkinaIyGaCGqMJs

Links:
https://metr.org/blog/2025-07-10-early-2025-ai-exper...
https://github.blog/news-insights/research/research-...
https://kiro.dev

Follow me on socials!
X: https://x.com/ErikCH
LinkedIn: https://www.linkedin.com/in/erikhanchett
Bluesky:https://blueskydirectory.com/profiles/erikch.bsky.so...

BTW: I am a developer advocate for AWS, anything expressed in this video is my opinion and my opinion only.

0:00 Introduction
01:14 Tip 1
02:20 Tip 2
03:51 Tip 3
05:01 Tip 4
06:15 Tip 5
07:04 Modern tooling
07:25 Conclusion