Start here
Local verification engine Private alpha · external receipt open
A completely local confidence and verification layer. It samples independent answers, derives confidence from agreement, routes uncertain or consequential work to separate verifier roles, requires confirmation for corrections, and refuses when the local stack cannot establish an answer. The runtime and 30 test-bound trust invariants are working; calibration and independently governed held-out validation remain open.
Governed scale-up program In development
The larger local system being assembled around the verification engine and the research engine. Models, tools, memory, and long-running research stay subordinate to permissions, budgets, checkpoints, provenance, independent checks, human escalation, and hard stops. It is designed to scale from the current 24 GB GPU to a planned high-capacity unified-memory workstation without relaxing those controls. A development program, not a safety certification.
Lattice Working software
Local OpenAI-compatible execution gateway (~13,300 LOC, 133 tests): tiered model routing, async task management, session/provenance records, and a containment validator that flagged every adversarial proposal across 200 internal trials with no false positives. That is a self-built adversarial set scored against its own detection question, and the research on this site argues a detection rate reported without the question that produced it is not interpretable. Read it as a smoke test, not a benchmark.
lattice-commit Alpha · PyPI
A git-native, test-gated wrapper for LLM code-repair loops, packaged on PyPI, restores reliable multi-file repair where a bare loop silently reverts good fixes. Benchmarked on 22 purpose-built workspaces.
Source →
Governor Open source · MIT
A keyword router that classifies each request at the boundary and routes to the cheapest adequate model tier. About 49% measured cost reduction on mixed coding work, from 35 hand-rated turns against simulated pricing, rated by one person with no second rater. No learned classifier was ever tested against it, so nothing here shows keywords beating a learned router. Indicative, not a validated benchmark. Open source (MIT).
Source →
Mynd Research substrate
The packaged Python substrate the program's control experiments run on, including the internal-model baselines the boundary mechanisms are measured against.