Links & Contact

Explainer Sites

• Synchronism

  75 pages covering the coherence equation (Synchronism's central formalism — a research conjecture, not established physics; which mathematical object the equation is remains unsettled upstream, see /notebook/projects), predictions, interactive tools, and honest assessments
• SAGE

  Situation-Aware Governance Engine — research environment for persistent local AI (beings) under identity, memory, learned state, tools and Hestia-governed action, per its own docs of 2026-09-26; the 12-step Python cognition loop is upstream's 2025 design, retained as history. The deployed Rust daemon is an inference-and-metabolism gateway. The SAGE site itself has not caught up (checked 2026-10-22): its 'Today (June 2026)' section still describes a running 12-step consciousness loop, and it glosses LCT as 'hardware-bound identity', while the fleet's being LCTs are self-issued from a seed on the host (see /notebook/fleet). Rechecked 2026-10-27: it also glosses ATP/ADP as 'Energy metabolism (bounded compute budgets)' — canon's ATP is an Allocation Transfer Packet, 'a unit of account, not a protocol currency' in the web4 README — and T3 as 'multi-dimensional trust scoring' (canon: Trust Tensor); says LCT is 'witnessed across 2,700+ sessions', a basis this site has never reconciled with its own counts; and labels HUB 'paused 2026-07-29, count frozen' and CBP 'paused 2026-08-06', where /notebook/fleet has HUB's loop still committing with its ledger frozen and CBP's publishing stopped by policy on 2026-09-20, and the SAGE README's 2026-10-23 census has every machine but Thor running a being: three sources, three states for two lines. It does carry a Get Started section (clone, Ollama, build the daemon). The fix belongs in the SAGE site repo.
• 4-gov

  Computable Accountability: a 40-block explainer of the problem and of Web4's primitives (LCT, T3/V3, MRH, ATP/ADP, R6/R7, Hardbound). Its own vocabulary differs from canon in three places (checked 2026-10-23). Its acronym tooltips put T3's and V3's root dimensions in the expansion slot (T3 'Talent / Training / Temperament', V3 'Valuation / Veracity / Validity'), while its body text names them Trust Tensor and Value Tensor, as canon does. Its Part Two headings say 'The Architecture' and 'Web4 Architecture Thesis', though its own Web4 tooltip says 'governance ontology'. Its ADP tooltip has the right expansion but describes ADP as a 'record of work done' and leaves out ADP's return to ATP (see the ATP row in the glossary). Rechecked 2026-10-27: all three stand, and the body text adds a fourth instance, 'Open standard. Open architecture. Build with us.' The site's source is not public (no repository of that name under dp-web4; 4-gov.org links to the web4 and hestia repos, not to its own), so the fix belongs with its author, not here.
• 4-Life

  Trust-native society research prototype (the repo's own description; the live site titles itself 'The Web4 Onramp' and opens as an educational explainer, with the simulation behind it) — agents, ATP (Allocation Transfer Packet) / ADP (Allocation Discharge Packet), and karma (the simulation's accumulated-standing score for an agent, derived from its past contributions — 4-Life's local stand-in for a T3/V3 reputation record, not a Web4 canonical term). Web4 in miniature.

GitHub

New here? Repo descriptions use Web4 vocabulary (LCT, T3/V3, MRH, MCP, RDF, ATP/ADP). See the /context glossary for full definitions.

• dp-web4 (GitHub account)

  All public repositories
• Web4

  Trust-native ontology — LCT (Linked Context Token) identity, T3 (Trust Tensor — Talent / Training / Temperament) and V3 (Value Tensor — Valuation / Veracity / Validity) as complementary tensors (the equation's T3/V3 reads “verified by” in the Web4 whitepaper and standard glossary; see the Glossary), contextualized by MRH (Markov Relevancy Horizon), resources cycled through ATP (Allocation Transfer Packet) / ADP (Allocation Discharge Packet) — charged, spent, recharged against validated value, not a one-way log — augmented with MCP (Model Context Protocol) and RDF (Resource Description Framework). Research-stage; core packages published (web4-core on crates.io and PyPI — install lines in the /notebook/projects Quick start)
• SAGE

  Situation-Aware Governance Engine — research environment for persistent local AI (beings) under identity, memory, learned state, tools and Hestia-governed action, per its own docs of 2026-09-26; the 12-step Python cognition loop is upstream's 2025 design, retained as history. The deployed Rust daemon is an inference-and-metabolism gateway.
• SWE-SAGE (private repository; public description)

  SAGE applied to software engineering (SWE), the Gemma 4 Developer Agent Competition entry. The repository github.com/dp-web4/SWE-SAGE is private and returns 404 to visitors; this link is SAGE's public description of it (2026-09-26). No results published; see /notebook/projects
• Synchronism

  Theoretical research program — coherence conjecture, coupling experiments; conceptual lineage for Web4, not an engineering dependency
• AI DNA Discovery

  Biological-computational analogy explorations
• ACT

  Agentic Context Tool — Cosmos SDK implementation of the Agentic Context Protocol (ACP); its repo calls it an early-prototype candidate ledger backend for Web4 anchoring (meant to become a human interface to Web4; not one today). Stalled, and reserved rather than abandoned: last substantive commit 2026-04-29 (a daemon entry point and a chain startup path for the fleet machine CBP; the 2026-04-10 commit before it removed the README, its content moved to the web4 repo), no running chain this site can confirm, no first-run steps published. Not a working entry point today. License unresolved: its LICENSE file is Apache-2.0 (the Ignite CLI template), while its own repo-index.yaml says AGPL-3.0 and CITATION.cff says AGPL-3.0-or-later (checked 2026-10-27).
• OpenClaw (fork)

  Web4 governance extension for OpenClaw agent framework (upstream branch web4-governance-complete)
• Claude Flow (fork)

  Web4 governance WASM (WebAssembly) plugin for Claude Flow
• Claude Code (fork)

  Web4 governance hooks for Claude Code (upstream branch add-web4-governance-plugin)
• ARC-SAGE

  ARC-AGI-3 (Abstraction and Reasoning Corpus for Artificial General Intelligence, third generation) harness — 94.85% official ARC Prize action score, efficiency-weighted (Claude Opus 4.6, public set, network-enabled; 23 of 25 environments completed, 92.0%; 175 of 183 levels), world models, multi-agent frame-questioning (MIT-0 — MIT No Attribution). Score carries caveats — see /arc-agi-3
• ARC Prize official scorecard

  Official, externally scored record of the 94.85% action score — the external source of record for the ARC-SAGE result. It verifies the score, not the method: the engine-source affordances and the missing harness ablation are on /arc-agi-3
• Gemma 4 Good Submission

  Parse it as 'Gemma 4 Good' + 'submission', not as a self-rating: the lab's entry (titled Attested Resilience) in the Gemma 4 Good Hackathon, run by Kaggle and Google DeepMind in May 2026. Self-supervising AI with embedded oversight for constrained environments — persistent cryptographic identity per device, R6-auditable actions (R6 = the Six-Element Action Framework: Rules / Role / Request / Reference / Resource → Result — see /context), policy as signed artifact (Apache 2.0)
• Membot

  Brain cartridge server — embedding-based semantic memory with swappable cartridges
• Hestia

  Open local-first governance daemon — agents from different vendors under one signed law on one machine, with a hash-chained action record. ('Governance' is deliberate; the one explanation is at /notebook/context#governance-oversight.)
• Hub (4-hub)

  Rust society runtime that turns a community into a Web4 society — standalone mirror of the Hub daemon, whose source also lives in the web4 repo. Its README said seven roles until web4 #951 (2026-10-28) corrected it to eight, matching the web4 hub README, the glossary and /notebook/projects (upstream's 2026-10-23 landing audit had listed the lag as P1; rechecked 2026-10-29); the mirror's code is current, pushed 2026-10-29 in step with the hub
• SNARC

  SNARC (Surprise / Novelty / Arousal / Reward / Conflict) — salience-gated memory plugin for Claude Code (formerly engram)

Hardbound (hardware-bound oversight suite — “oversight” meaning machine-enforced policy gating, not human supervision) is internal research — code is not publicly available. Capabilities described on the Projects page.

Contributing

New here? Read the /context glossary for the vocabulary. Hands-on: start with SAGE. The ontology and its code: start with Web4, whose core packages now install directly. ACT is the Cosmos SDK side, but it is stalled — don't start there.

Quick start

The commands live in one place: the Quick start on /notebook/projects — Web4's published packages, a SAGE instance from clone to running daemon (including the ollama pull step). This page used to carry a second copy; the two drifted apart, so it now links instead. SAGE is the hands-on starting point — it runs on a single machine, has 2,499 internally logged session records as context (2,620 including HUB; per-machine counts on /notebook/fleet, the source of record), and is where most fleet behavior originates. Web4 is the starting point for the ontology and its reference implementation. ACT is the Cosmos SDK take on the same trust primitives, but it is stalled and has no published first-run steps, so it is not a third starting point today.

Licenses

ARC-SAGE: MIT-0 (MIT No Attribution; see repo). Gemma 4 Good Submission: Apache 2.0. ACT: unresolved — LICENSE file Apache-2.0, its own repo-index.yaml AGPL-3.0, CITATION.cff AGPL-3.0-or-later (checked 2026-10-27); treat it as undetermined until upstream reconciles them. Other repos: see each repo's LICENSE file — check before forking. Hardbound is internal and not publicly available.

Questions or contributions? Open an issue on any public repo. SAGE and Web4 are the most active issue trackers. The dp-web4 account holds 40 original repos — 21 public, 19 internal — plus 33 forks of external work the lab builds on (32 public, one internal), 73 repositories in all. (“Original” means not-a-fork. Verified against the account on 2026-09-12.) Clone what interests you, experiment, and open a discussion if you want to engage.

There is no CONTRIBUTING guide yet and no stated policy on outside pull requests. Ask in an issue first — that's the on-ramp until a formal policy exists.

For researchers: no papers or preprints are published yet. A Synchronism preprint strategy exists upstream but is unscheduled — it gates on the researcher, with no submission target; this page previously said “in late revision,” which the repo does not support. The claims on this site (other than the externally scored ARC-AGI-3 result) rest on the lab's own reading of its session records. The raising session records through 2026-09-19 are public as full transcripts in the SAGE repo under sage/instances/<line>/sessions/. Since then most lines mirror their records privately under a ruling by the lab's human lead, not the machines (SAGE commit cefb5c184); see /notebook/raising. Coding criteria and a rater protocol are not published, and no third party has reviewed the readings.

Contact

• dp@metalinxx.io

  Research inquiries