Hacker Newsnew | past | comments | ask | show | jobs | submit | jukkat's commentslogin

Put every detail into CLAUDE.md and after a while CC starts to forget/ignore what it’s been told.

I’d like to see dynamic task-specific context building. Write a prompt and the model starts to collect relevant instructions.

Also a review loop to check that instructions were followed.


Why these browser extensions cannot live in a guarded sandbox? Extensions are given full access to whatever is available on any page. I had legacy React developer tools and Redux DevTools installed for years. What a great attack vector.


Would be cool to use server actions in other clients, too. Maybe the most powerful feature in Next.js.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: