@0xmrtn@cjzafir Because context bloat is a thing. A better place to have your "context cached" is an md file you always have control over your context in that way.
A fresh chat is 99 times better then continuing everything in the same one.
one more tip: review completed work in a ๐ณ๐ฟ๐ฒ๐๐ต ๐๐ผ๐ฑ๐ฒ๐ ๐๐ฒ๐๐๐ถ๐ผ๐ป.
It didn't take part in the implementation, so it doesn't carry the original session's ๐ฟ๐ฒ๐ฎ๐๐ผ๐ป๐ถ๐ป๐ด ๐ถ๐ป๐ฒ๐ฟ๐๐ถ๐ฎ.
Before closing the original session, have it create review-brief.md in the repo root.
The brief contains only:
โ your original request, verbatim
โ requirements or scope changes you approved later
โ the diff or commit range to review
Leave out the ๐ถ๐บ๐ฝ๐น๐ฒ๐บ๐ฒ๐ป๐๐ฎ๐๐ถ๐ผ๐ป ๐๐๐บ๐บ๐ฎ๐ฟ๐.
In the fresh session, send:
Read review-brief.md. Use the repo, docs, diff, and tests as evidence. Mark every requirement complete, partial, or missing. Look for requirements that never made it into the implementation. Fix confirmed issues, rerun the tests, then delete review-brief.md.
I take the lazy route: handoff writes the context file. A fresh session reads it, then runs ce-code-review.
both skills below ๐
New way to hire developers in the AI world:
1. Skip the "are you proficient in TS?" questions. Spend 2-3 minutes on their genuine interest in AI's future.
2. Give them a broken piece of code and an AI tool. If they canโt use the AI to diagnose and fix the root cause in 10 minutes, theyโre behind the curve.
3. Reverse-engineer their logic. Donโt just look at the output. Ask: "What instructions did you give the AI to get here, and why did you iterate on them?"
4. They should be able to explain the "why" behind every design choice. Technical depth matters more now, not less.
5. Use the interview to see if they can use AI to map out a system, rather than just writing individual functions.
Project based learning is your friend. Forget learning the concepts first. Go straight to building something, when you get stuck somewhere, then learn that part, rinse repeat.
Choosing a project to build is much more intresting and fun than choosing what to learn. And the end of it, youโve learned a lot and have something to show for it
@sidborderwala Iteration 6 looks much better. The final one just looks off. Also looks like youโre not following design principles. Because both the font and symbol have unequal inner padding/margin vertically it just looks off. Choose either the font or the symbols having both looks not good