From the lab
Products, agents,
and model experiments.
Everything here is something we wanted to exist. Some are live products, some are early-access, some are still research. Open source where we can, in the open where we can't.
/Slash
Your AI engineer for GitHub and VPS operations: repo-aware chat, edits, PR reviews, project summaries, and live server monitoring in one mobile-first workspace.
ACRA
The Auditable Clinical Reasoning Agent (ACRA) is a research architecture for evidence-grounded clinical decision support. It adds a ReAct-style agent loop, chain-of-thought steps that stay tied to retrieved evidence, and structured audit logs so recommendations can be inspected. The point is traceability, not one-shot RAG with no paper trail.
Apex
In-house Solana trading stack with AI in the loop. It watches chain and context, surfaces candidate trades (the calls), and can execute including buys when checks pass. Open beta; risk rules, signals, and roadmap stay private.
Pulse
A GitHub digest that turns repository activity into short, readable updates for people who are tired of living inside the notification stream. Built for async teams.
Lexa
An in-house legal-domain model experiment: reasoning over case law, firm notes, and long research threads where answers need to stay tied to what the corpus actually says. Built to explore how far domain specialization and careful grounding can go before anything is productized.
TagPilot
Straightforward tagging and labeling for datasets: assign labels, review items, and keep data and training in a tight loop. The UI stays simple on purpose.
ChatPilot
A lightweight assistant widget you can drop into any site with a single script tag. Near-zero setup, fully customisable, open source.