A strategic-reasoning AI tuned to the Hammerstein framework. Portable system prompt + retrieval. The framework survives the model.
A persistent epistemic substrate for AI. Nous treats language models as larynx, not mind, and benchmarks epistemic structure instead of output fluency.
A fully local, governance-ready terminal powered by specsmith. Fork of Warp via OpenWarp.
Local-first RAG over your own document library (PDFs, papers, books, notes). Grounded answers with inline citations, eval harnesses that measures retrieval quality, epistemics and per-answer provenance.
Flarepoint | A plaza of small, behavioural Claude Code plugins that make the AI's cognitive state explicit and persistent
Fail-closed output layer for LLM systems: an unsourced claim is a build error.
Epistemic decision ledger: append-only event log with evidence-gated trust, deterministic Tessera projection, and a human control room. TypeScript + Cloudflare Workers.
Epistemic confidence engine — every AI belief carries its provenance
A Claude Code skill that separates established facts, inferences, and speculations before synthesising a response.
An autonomous research agent that must show its evidence. Field notes on the one bug that kept reappearing: confusing 'we did not find it' with 'it is not there'.
Your assistant tells you what you want to hear, and doesn't know when it's making things up. HEXAGRAM adds the missing layer: explicit truth statuses, named substrate voices, measure-before-you-speak. Plus an astronomical engine and a six-element model of time. Core in Polish by design; the assistant speaks your language.
Epistemic infrastructure for decision-making — methodology, architecture, and roadmap
Four papers, three seed-pinned verification batteries, and a Lean 4 formalization of the core theorems on AI deployment and collective epistemic capacity: pressure, drift, capture - and a pre-registered audit protocol to locate deployments on the ladder.
La Alegría 🇦🇷 — a bilingual (EN/ES) interactive explorable explanation of Argentina's 2026 World Cup run: refereeing & FBI-probe gossip examined honestly, with three sims, an LLM chat over the compiled research, and a self-grading re-run loop. Astro + TypeScript.
Cross-model audit engine for documents and technical artifacts: hallucination gate on verbatim quotes, claim verification, blind adversarial review across different model vendors, Type-I error control. Dependency-free Python core + Claude Code plugin. Never reports "validated" on internal grounds — only a human does.
A Claude Code skill that forces grounded research + a cheapest-falsifying test before Claude commits to any consequential claim. Red-team your own belief.
Local-first epistemic knowledge substrate for AI agents — provenance, contention, stigmergic refinement, MCP-native
世界引擎 (WorldEngine) — 在长程上守住一致性与秘密的叙事世界引擎:LLM 提议、Canon 裁决、fail-closed。A fail-closed narrative world engine for long-range consistency & secrets.
A benchmark for evaluating LLMs as mediators of political information, measuring different dimensions under controlled information conditions.
Absence ledger for the four-line governance set: records epistemic gaps, withheld claims, and deliberate non-assertion as NAMED, UNRESOLVED, WITHHELD, or NOT_RECORDED, decays dated observations past their review horizon, and exports line.report-envelope/1.0 reports.
To associate your repository with the epistemics topic, visit your repo's landing page and select "manage topics."