Host preserves the Task, Harness runs a replaceable Agent episode, and Runtime commits physical local execution.
Build Agent systems that can survive contact with time, tools, capital, people, and Reality.
Ordivon is a family of independently owned systems and research lines. Host, Harness, and Runtime preserve durable work; Game, Finance, World, Human, Security, and Media pressure those ideas in materially different domains; Computing tests what should become reusable knowledge and what should be deleted.
What exists today
Capability, not roadmap languageProvides schema-v5 Host Journal/CAS authority, revision-fenced TaskProjection state, external-continuity WorkingCheckpoints with exact response-loss replay, bounded handoff/inspection, opaque extension-state durability, Doctor/backup/restore/deployment receipts, and a deliberately small six-Tool MCP surface for adopt, checkpoint, resume, list, observe, and Host status.
Operational owner-trusted local engineering component; pre-1.0 public interface↗usable nowOrdivon RuntimeOwns exact Workspaces, durable Jobs and Attempts, immutable-input execution on local Linux and configured limited Windows targets, provider/source commitments, digest-bound image observation, request/target affordance discovery, Windows-native Job Object execution, cancellation, restart recovery, bounded Artifacts, exact replay, reclamation, and reconciliation of uncertain local commits.
Operational owner-trusted local infrastructure; public interface remains pre-1.0 and Windows-native is experimental opt-in↗playable nowOrdivon GameRuns the registered Station Zero v2 product while retaining v3 and other treatments as executable research pressure. Station Zero v3 now has owner-native Product Value lanes with cheap objective/profile slices, currentness-fenced retained evidence, browser/playable proof, and a full promotion matrix; recent natural dogfood used those pressures to reject two plausible fixes and correct a real Core extraction defect without making Human approval a default inner-loop gate.
Registered playable v2 product plus active pre-G0 product-value/direction research; no next product selected↗implemented prototypeOrdivon FinanceProvides Finance Foundations v1, an Owner Constitution, point-in-time and aggregate multi-venue portfolio evidence, FinancialResource and exposure semantics, exact held-state/valuation-time separation, CapitalStateView and replayable exposure graphs, owner-capital research context, capital-rationality and Decision/Proposal machinery, a process-separated quantitative research lab, exact effect/reconciliation contracts, and Agent Context v21 / Primary Agent Surface v14 with seven semantic operations.
Implemented Agent-first capital system with real-capital observation, decision, effect, reconciliation, and performance semantics; live external submission remains disabled↗researchOrdivon MediaMedia now owns a technology-neutral mediation contract with SourceBinding, ObserverScope, ProjectionEnvelope, RepresentationBody, Transformation/ProvenanceTrace, and ActionExposure/Disclosure roles plus eleven tested non-lifting invariants. Studio remains the active production plane for revision-bound Claims, editable cross-medium sources, exact selected assets, rendering/QC, review preparation, and human-response calibration when a decision genuinely depends on it.
Phase 1 closed: stable Media owner, frozen MF0–MF9 research substrate, OMPC-v0 conformance baseline, active Studio production plane↗researchOrdivon ComputingRuns owner-preserving cross-project falsification, historical and live dogfood, world-model revision, SCD and Computational Possibility research cores, reusable Knowledge synthesis, responsibility placement/deletion tests, and selected protocol conformance. Mature external results are imported into the informed baseline rather than re-proved for Ordivon ownership.
Mature cross-project synthesis, Knowledge, method, and conformance root; new research remains admission-gated↗Project map
The family is organized by what it owns—not by one universal platform diagram.
Core boundaries carry durable work. Domain owners preserve their own truth. Research and Knowledge decide which distinctions transfer—and which should disappear.Game, World, Finance, and Media pressure the system through play, external reality, capital effects, and structured human-facing mediation; Studio remains Media’s production plane.
Computing tests cross-project laws, Human studies practical trajectories, and Security runs bounded adversarial experiments.
Why the boundaries exist
12 August 2026 · retained evidenceWe made the explanation richer. The decisions did not improve.
Across five frozen surfaces and 1,326 accepted Agent decisions, compact causal prose matched causal cards, typed relations, and explicit question grammars on exact action while using fewer Provider tokens. The result supports a smaller default explanation—not a claim that diagrams or structure are never useful.
Current frontier
When should a finite intelligence expand its interface to Reality instead of only reasoning harder?
Testing under a contracted theory. External prior art falsifies the claim that action/interface expansion is absent, but it also weakens any absolute transcendence claim: callable, representational, affordance, and executor expansions can be new relative to one closure while remaining compositional inside a deeper supplied substrate. No universal basis hierarchy or generic basis-escape object is admitted.Wait for a natural owner-level failure that satisfies an already-frozen strict burden: a real certificate/language or grounded-meta-closure case where the current closure is prospectively explicit and the repair cannot be reconstructed as ordinary composition inside the admitted substrate. Do not manufacture E1b/E2 merely because apparatus exists.
Writing
Dated arguments preserve what changed—and what later became historical.
Articles interpret evidence at a date. Current project pages summarize today's boundary and link back to the repository that owns the facts.The Decision Was Valid. The Portfolio Changed.
Why Decision validity and commitment currentness are different facts, using Ordivon Finance's exact portfolio-baseline fence for state-changing proposals.
Ordivon Finance↗The Evidence Is Real. It Is Still Too Old.
Why authentic evidence can become stale without becoming false, how Security EC1 and World experiments tested current applicability, and why TTLs or generation numbers are weaker than exact dependency revalidation for the tested cases.
Ordivon Computing / Security↗A Better-Looking Result Can Still Be Noise
What Studio R4–R6 learned about richer perception, grounded meaning, search inflation, real Web encounters, and why no single score deserves creative authority.
Ordivon Studio / Web↗Choose the depth that matches your question.