Bringing ideas to life. A small, bootstrapped studio building AI-native open source tools and selective partner products. No VCs. No fluff. Shipped software.
AI as a first-class citizen of the workflow, not an afterthought. Most of our work is open source, written in Rust, designed to be one binary you can drop on a machine and trust.
Three things:
Products built end-to-end by a small AI-augmented team
Open source tools that solve problems we hit in our own work
Selective partnerships with teams shipping something we believe in
Quality bar over feature count. Performance over abstraction. We ship publicly.
Products
Octomind Cloud — Pick an agent; it gets its own computer. Twelve specialist agents — researcher, blog writer, SEO auditor, data analyst, developer — each on a real Linux machine in the cloud that browses the web, writes and runs code, and keeps your files. Free tier, per-second billing, REST API on every plan. · source
TypeTab — On-device autocomplete for your whole Mac. Tab takes the next word, Ctrl+E the line. Learns how you write — entirely on Apple Silicon, nothing leaves your Mac.
Timex — Automatic Mac time tracker, break timer, and lid-down keeper. 1 Hz sampling into a local SQLite file. No cloud, no account. · source
Vext — Voice-to-text for Mac. Local-only, with cleanup, translation, and meeting summaries on Apple Silicon. · source
VnePN — VPN built to work in Russia. VLESS + Reality, smart routing for banking apps. · source
Two founders, an AI person (@avanuts) who keeps full context across every repo, and a fleet of on-demand specialist agents for review, testing, security, and ops. We use our own tools — Octomind runs our day. If something we need doesn't exist, we build it and open it.
Structural code intelligence for AI agents — semantic search, knowledge graphs, and a built-in MCP server in one Rust binary. Give Claude, Cursor, and any MCP client a deep understanding of your co…
One Rust API for 30+ AI providers — chat, tool calls, structured output, embeddings, reranking and media generation behind a single provider:model string, with built-in cost tracking
Benchmark harness that compares coding agents — not just models — on real, contamination-resistant tasks: 80 fail-to-pass cases from merged OSS PRs, long-run sequences, SWE-bench-Live, and 25+ domain benchmarks. We benchmark our own agent, octomind, under the same rules.
There was an error while loading. Please reload this page.
You are visitor 01402503 — GeezerVison does not store the pages you look up. Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.