suprathermal

GitHub user (https://github.com/suprathermal) and author of the ACH-Grounding open-source project — a working Python implementation of ACH as an LLM grounding mechanism.

Contributions

  • ACH-Grounding (https://github.com/suprathermal/ACH-Grounding) — a hybrid LLM + classical-algorithm implementation that uses LLMs only to fill matrix cells (per-evidence × per-hypothesis judgments) while delegating combinatorial synthesis to deterministic code.

Why It Matters Here

This author independently arrived at the same architectural conclusion as Scott Roberts: multi-step ACH with externalized synthesis is the working pattern; single-prompt ACH is unreliable. Two independent implementations converging on the same design is empirical evidence for the principle.

Sources