Built CATS this week at @GauntletAI: a multi-agent adversarial-evaluation platform that continuously probes an AI healthcare co-pilot for vulnerabilities. The fun part: watching AI try to break another AI
Week 2 of @GauntletAI shipped a v2 clinical copilot for OpenEMR: 2 LangGraph pipelines, hybrid RAG across 4 guideline corpora, polymorphic citation verifier per source, React dashboard coexisting with the legacy PHP. PR gate runs evals on every change.
Shipped my Week 1 @GauntletAI project: a Clinical Co-Pilot for OpenEMR. LangGraph + Anthropic agent that builds a pre-visit briefing from real chart data, with a deterministic verifier that rejects any claim it can't trace back to a source record. No hallucinated labs allowed.
Not really active here or FB any more. If you want to keep up with me now, head over to @[email protected] in your #fediverse client of choice (mastodon etc.) #BYE
@martinisoft Heartbleed! That takes me back, didn't realize it had been so long - I remember you hunched over the computer working furiously - while meanwhile I was googling "what is a TLS heartbeat"
@KanikaTolver We are hiring @bradsdeals! Product managers, designers, developers, dev ops engineers. Remote-friendly, an exciting product roadmap, at a small, independent, non-VC company: https://t.co/AAwwLNqdwD