foolish-bandit makes Gary show its work

A UI pass reframes the assistant so every answer wears its sources on its sleeve.

chat-uisearch

The Gary fork just got a frontend nudge from foolish-bandit aimed at trust. The model and retrieval pipeline are untouched - what changed is the framing around the answers. Ask Gary now prompts users differently, assistant replies surface citation cards and warnings, the table-review chat picks up the same cues, and the Review, Draft, and Explain surfaces have had their copy rewritten to set the same expectation: answers are grounded in documents, and the sources are visible.

It's a small, self-merged polishing pass - not a re-architecture - but it nudges Gary from "AI that talks about your documents" toward "AI that shows you which lines of which documents it leaned on." For a legal user, that distinction is the whole ballgame.

So what Anyone building a legal assistant should note this: the trust upgrade here is almost entirely UI, not model - and it's the kind of change clients actually feel.

View this fork on GitHub →

Spotted something wrong? Or know the PR text has fresher detail than the writeup above?

Commits in this thread

1 commit from foolish-bandit/gary, oldest first. Source extracted verbatim from the harvested git log.

SHA Subject Author Date
b178b7ed polish cited answer expectations SONOMOS 2026-05-07 ↗ GitHub

Capture this thread into my fork

Download a single Markdown prompt that tells Claude how to port every commit above into your working tree — adapting paths and structure to match your repo. Run it via claude -p < capture-thread-264.md from inside the repo you want the changes in.

⬇ Download capture-thread-264.md