@Just_Ryun It lives inside the project at .baton/baton.db, a local SQLite file.
You've top write a hook inside AGENTS.md (Codex, Cursor) or CLAUDE.md (Claude) to make your agents to use it.
https://t.co/GuAhrjoeGH
Just shipped Baton 🚀
Switching AI coding agents? The new one starts from zero. You re-explain everything.
Baton saves a checkpoint. The next agent reads it and picks up where you left off - Task, decisions, what's next.
Works with Claude, Cursor, Codex.
Open source.
@doug_tarr Two saves become two rows with separate UUIDs, SQLite serializes the writes.
'Latest' is ordered by created_at, so you keep both, newest wins.
Honestly, the more I think about it, Running Claude and Codex together is the whole point. A genuine gap. Noted.
@ProductHunt Built a self-verifying agent last week to scratch a personal itch. The agent revises until it passes the fitting criteria 🤞
https://t.co/QRs5DtLJJZ
@mvanhorn haha no worries, still a great feature! 👏
Same thought every time I set up a new machine 😅 watching one package install at a time is painful. Looking forward to see this to land!
I’m thinking about banning Claude code at Shopify until they change their mind and read AGENTS.md and .agents/skills etc.
Insisting on only reading CLAUDE.md sometimes leads to split brain problems when different team members use different tools. Just unnecessary.
Nobody audits those "us vs you" comparison pages, Crawlers/LLMs just scrape the false claims and serve them straight to potential customers.
Feels like there's a real gap for a TrustPilot-style verifier, but for feature claims instead of reviews. @yongfook 🤔
Idea for SaaS:
AI agent that observes your competitors "feature comparison with <your app>" lists. Builds the features they say you don't have. Then auto files a trade libel dispute requiring them to remove false information.
Comparison lists are dumb in a post AI world.
Caching is your design system's best friend — until it isn't.
Cache invalidation kills systems silently.
Ask yourself:
• When does stale data become dangerous?
• What's your eviction policy?
• TTL too short = DB overload. Too long = bad data.
Cache with intention.
API design is system design.
A poorly designed API forces every downstream service to compensate.
Get contracts right early. REST, GraphQL, gRPC — each has tradeoffs. Know WHY you chose one, not just WHAT it does.