Helmryth
A private operating system for autonomous work. Persistent operators run on the CLIs you already own, and nothing consequential runs until you answer.
- 3,054 tests across 274 files
- 160 routes and 76 IPC channels pinned by build gates
- 30 product films, each shot in one take on a live build
Run Phantom
A local-first debugger that turns AI-agent traces into timelines, side-by-side comparisons and regression checks for CI.
- 349 commits on main in three weeks
- Windows portability from an outside contributor, 5 merged PRs
- Reads traces from TypeScript, Python, Go and Rust with no SDK
JITMIND
Memory for long-running agents that knows when a fact stopped being true, and retrieves across four channels at once.
- Five timestamps on every memory, across two clocks
- BM25, dense, index and graph retrieval fused by RRF
- As-of queries: what the agent believed on any date
Siblings. One lane vocabulary — Luna, Terra, Sol — and one rule: no lane is trusted for work it only claims to have done. FABLEWRIGHT runs inside Claude Code and Codex; ORRERY serves six clients as an MCP server.
FABLEWRIGHT
Nobody grades their own homework. Claude Fable writes the spec and accepts the result; pinned GPT, DeepSeek and GLM lanes do the work.
- v1.0.0 released 3 Sep 2026
- Every model pin proven from its session record
- 143 offline tests, zero tokens spent
ORRERY
Architect-first orchestration for coding agents. Your chat keeps the architecture; pinned lanes implement; a read-only reviewer decides.
- Zero runtime dependencies, 73 offline tests
- Role files for six clients, from Codex to Kiro
- The model never chooses where a file is written
Upstream
In other people’s code
350+ pull requests to 50+ open-source projects, including Hugging Face, Superset, Burn, Baseten and mcp-use.
- Pull requests
- 350+
- Merged, with Breken’s
- 100+
- Stars on those projects
- 800k+
63 of those merges are mine; 42 came from Breken’s own account.