We build AI products and help businesses cut through the hype to ship software that actually matters. Our work spans products, consulting, and community — all grounded in the belief that AI should help people think better, not think less.
Alongside our product and consulting work, we publish open-source tools that reflect how we like to build: local-first, inspectable, and grounded in real workflows.
- arcnem-vision — Turn images into searchable, meaningful data. A platform combining mobile camera capture with AI agent orchestration via LangGraph and MCP. Flutter client, Bun/Hono API, React dashboard, and Go agent services.
- autoarc — Benchmark-guided code improvement for local repositories. An open-source Gleam service that runs measured code experiments in isolated worktrees, asks models to propose changes, and records the results in SQLite.
- picvec — Local-first image similarity search. A Go CLI that runs vision models (CLIP, SigLIP2) locally via ONNX Runtime and stores embeddings in a local libsql database.
- texvec — Local-first text similarity search. A Go CLI that summarizes and embeds documents locally with ONNX models, stores vectors in a local libsql database, and ranks matches with cosine distance.
- omnivec — API-first asset librarian for local-first document and image similarity analysis. A Python service that combines
texvecandpicvecto analyze uploaded corpora and return structured JSON plus a markdown report.