@kaicontextdev

Find and fix bugs beyond the diff. An independent code reviewer that traces your changes through a semantic map of your codebase. → https://xitter.cf/t.co/ZHIbzXHA5V

kaicontext.com
Joined March 2026
  • Tweets15
  • Following3
  • Followers4
1Media
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
Kai Context retweeted
Jacob
@jakecodesOct 6
We wrote up how to read Martian’s bench → kaicontext.com/blog/how-to-r…

How to read Martian's Code Review Bench

More than one AI code review tool says it's number one on Martian's Code Review Bench. Here's how the test works and how to read its scores.

kaicontext.com
1
1
64
Kai Context retweeted
Jacob
@jakecodesOct 6
In the next version of Kai… it flags bugs the diff never touched.
1
1
135
Kai Context retweeted
Jacob
@jakecodesSep 29
Code review in 2026: ✅ Diff looks great ✅ Tests pass ✅ 3 AI bots approved ❌ The new code can never actually run Review the code around the change, not just the change.
1
1
3
155
Replying to @saranormous
👋
Replying to @PremiumGoblin
Do you also see it as a new challenge?
2
Replying to @bendee983
Greenfield feels low friction to get started, however a mature codebase we've cultivated requires a psychological shift. (legal compliance aside)
2
Replying to @MatthiasWagner
Agree!
1
3
Replying to @andrewchen
Definitely an issue that's being worked on. In the meantime an easy solution is to fire off prompts to the cloud to open PRs and then have an agent merge them in.
2
Replying to @Prathkum
Agree. Things are moving in this space!
2
Replying to @boshen_c
Have had good results with Cursor, Claude, Codex. Any specific needs/constraints?
3
Replying to @LyalinDotCom
I guess it depends 😅 Device? Prompt needs/complexity?
1
Replying to @icanvardar
You don't have to wait for them to finish. You can run them in a worktree, in the cloud, a second machine, or maybe increase the scope of the work to get more done and go watch a movie 🙂
3
Replying to @Tech_girl
Maybe for the same reason folks have subscriptions to Netflix, Apple TV, Disney, etc... My guess: optionality, variety, affinity, and to avoid lock-in. 🍿
2
🚀 We've been heads down. Kai's gone from 0.11 → 0.35: - Proper benchmark harness (roughly doubled our SWE-- bench pass rate) - GitHub CI integration - Lots of new features, lots of issues resolved More to share soon.
2
1
33
Load more