Video yükleniyor...
Video Yüklenemedi
Well this went viral. Seems like we need a better way to visualize agentic traces, especially for people who are not crazy technical enough to diagnose a JSON trace. Here's one UX we're exploring in Kitaru ( - a Tinder-style approach to evaluating your agent traces. You can swipe... show more
16,184 görüntüleme • 16 gün önce •via X (Twitter)
8 Yorum

Muslims bros creating a halal tinder that’s crazy

Hahahaha

he’s really gone and done it 👀

lfg

Need of the hour

Please don't give more features to non-tech people. It's already painful to explain to them what JSON is. Lol

Love this — non-technical users need verdicts, not JSON traces. We hit the same wall from the other side: every AI audit on our rail ends as human-readable reasoning + a binary PASS/FAIL settled on-chain, so even the trace is settle-able by anyone. A swipeable review layer?

A simpler trace is useful if experts can still find and fix what went wrong. Take twenty failed runs with known causes. Give domain experts either the raw trace or the simplified version without telling them which interface they’re testing. Ask them to identify the first mistake that mattered and what the agent should have done instead. Then replay the proposed correction. Compare diagnosis accuracy, review time and whether the correction actually prevents the failure. If the simpler interface gets experts to the same repair faster, you’ve got much stronger evidence than readability alone.
