This is the single source of truth for the public-facing Web4 ecosystem. If a doc elsewhere describes one of these repos differently, prefer this file.
This repo. Canonical Web4 specifications, reference Python SDK, attack-simulation suite, attestation envelope reference, whitepaper. Trust ontology and the spec corpus live here.
License: AGPL-3.0-or-later (with web4-core/ MIT for ARIA grant compatibility).
The cognition harness. Wraps an LLM with Web4-shaped context — typed world models, skill registry, trust-calibrated dispatch, structured perception. Produced the 94.85% on ARC-AGI-3 with same Claude Opus 4.6.
Use when: building agents that need to reason over complex, novel state with verifiable structure.
See: docs/proof/ARC-AGI-3.md for the headline result.
Interactive lifecycle and trust-evolution explainer. A browsable demo of how agents earn trust over time — witnessing, accumulation, decay. Live deployment: https://4-life-ivory.vercel.app/
Use when: introducing Web4 trust evolution to a non-technical audience, or sanity-checking that the spec doesn't assume too much (visitor-friction logs surface invisible assumptions back to the spec).
Public-facing site for governance content + AI Demo Day 4 deck. The site at https://4-gov.org hosts the long-form talk on AI accountability + the elevator-pitch demo deck at /demo.
Use when: needing the audience-tested public framing rather than the technical spec.
Theoretical foundation. Coherence equations, MRH, phase transitions, compatibility lens. Web4's MRH primitive is grounded here; the math underlying trust formation (Hill function, p_crit ∝ 1/⟨C⟩) lives here.
Use when: reasoning about why MRH is fractal, how trust thresholds emerge, and the formal grounding of Web4's contextual primitives.
Distributed ledger for ATP tokens and LCT presence registry. Cosmos SDK implementation. Use when: persistence and consensus around ATP economics or LCT registration is required.
Distributed memory paradigms. Lightchains, blockchain alternatives for edge devices, memory-as-temporal-sensor research. Use when: designing memory architectures for resource-constrained agentic deployments.
Edge AI kernel research. MoE expert selection, trust-based routing. Active forum and insights archive (synthon framing, identity portability findings). Use when: studying how Web4 trust composes with neural architecture choices.
Web4 in production. Hardware-bound identity, verifiable audit chains, contextual computable trust — productized for regulated industries (finance, defense, healthcare, infrastructure). TPM 2.0 binding on Linux, FIDO2 fallback, policy enforcement, audit pipeline.
Built on the open Web4 ontology — Hardbound is the deployment, Web4 is the substrate.
Contact: [email protected] for evaluation and deployment.
| You want to... | Go to |
|---|---|
| Read the standard | this repo (web4-standard/, whitepaper/) |
| See the 94.85% proof point | SAGE + docs/proof/ARC-AGI-3.md |
| Understand trust evolution interactively | 4-life-ivory.vercel.app |
| See the audience-tested public framing | 4-gov.org + docs/why/DEMO_DAY_2026-04.md |
| Deploy in a regulated environment | Contact [email protected] about Hardbound |
| Study the formal foundations | Synchronism |
| Persist ATP/LCT state | ACT |