01Start with Ordivon
Why durable agent work matters
Begin with the practical failure, the project intent, and the larger future Ordivon is trying to make possible.
Writing
Ideas, experiments, and decisions behind durable agent work.
Start with the architecture, follow the experiments that changed it, or browse the complete dated record.
- Research reports
- 3
- Engineering records
- 10
- Essays and notes
- 6
From Tokens to Work: The Complete Agent Execution Stack
A model generates representations. A harness creates one cognitive, tool-using run. Host preserves the work across runs. Runtime turns admitted actions into physical evidence.
12 min read ↗Research reportWhat Survived When Codex and Hermes Replaced Each Other Mid-Task
How one task remained coherent through Codex→Hermes and Hermes→Codex replacement, three injected faults, and materially different provider lifecycles.
12 min ↗Research reportWinning the Move Can Lose the ContestHow Ordivon Security Round 1 used local dynamic opponents, CAGE Challenge 4, and bounded Hermes/Codex diagnostics to separate tactical success from strategic outcome without promoting a Campaign engine, organization ontology, strategic state, or custom cyber range.
12 min ↗Reading paths
Choose the question you want the work to answer.
Each path begins with a clear problem and moves toward the experiments, decisions, or long-form argument behind it.
02Agent architecture
From model output to completed work
Follow the execution stack, the Harness boundary, and the experiments that made the surviving architecture smaller.
03Experiments and failures
What changed under real pressure
Read the reports where tactical success, duplicate authority, response loss, and strong baselines forced a different judgment.
Publishing rule
Publish when an experiment, boundary, release, or judgment becomes worth preserving—not simply because another page can be filled.
Follow a research question
Continue from a specific uncertainty.
For readers already following one research line, these groups connect the current question to its complete published arguments.
Ordivon Gametesting
Which game structures become possible only when Agents persist and act through the same Host?
Testing with a released application. Station Zero now demonstrates persistent specialists, partial observations, communication whose reachability changes outcomes, player authority, provider replacement, replay, diagnosis, and exact run comparison. It still does not prove that these mechanics outperform a strong classical implementation or equal-budget single Agent.
Ordivon Hostopen
What is the smallest Ordivon Harness that turns a bare model API into a verifiable Agent Run?
Ready for construction at M2. The complete model-to-work stack, H1–H5 ownership boundary, first falsifier, comparison baselines, acceptance, non-goals, and deletion condition are frozen. No implementation evidence exists yet.
Ordivon Computinganswered
Which Agent-native responsibilities remain after strong classical baselines?
Answered within the Round 1 boundary. LangGraph and Temporal carried durable work state; current-revision retrieval matched the tested Context need; single-backend Effect machinery shrank to identity, UNKNOWN, correlation, reconciliation, and no blind redispatch; live provider replacement preserved Task state without proving model equivalence.
Ordivon Hostanswered
Which Harness objects survive live provider replacement without duplicating Host or Runtime state?
Answered by H1–H5. Both replacement orders completed through one Task Attempt and fresh Assignment generation; stale completion, missing Artifact, and response loss were handled correctly. Provider final text was not portable evidence, and Codex/Hermes lifecycles remained materially different. The durable Host boundary survived; a shared internal Provider lifecycle did not.
Ordivon Runtimetesting
Which real structured operation can complete the minimal Effect contract across a second backend?
Testing after contraction. Runtime dogfood supports commitment as the local abstraction, while Round 1 showed no advantage for a larger single-backend Effect graph. Host H2/R2 also completed real correlation and recovery through existing Runtime request identity and opaque references without expanding Runtime.
Ordivon Hosttesting
Can Host complete a general repository Goal without absorbing Runtime mechanics?
Strongly supported but still bounded. H2–H6 established continuation; Harness H1–H5 completed a frozen repository-repair workload in both Codex/Hermes replacement orders, rejected stale and missing-Artifact completion claims, recovered a dropped repair response without redispatch, and committed TaskOutcome only after independent Runtime acceptance. A broader unbounded repository Goal remains outstanding.
Ordivon Computingtesting
When should an Agent act, wait, observe, refuse, reconcile, or request a decision?
Testing. Round 1 reduced deterministic approval interruptions from 12 to 7 with no missed escalation in its designed cases, but those timings were estimated rather than measured with real operators. Security, Game, Runtime, and World now provide stronger act-versus-wait trajectories.
Ordivon Securitytesting
Can strategic adversarial trajectories be evaluated without collapsing them into one reward or success flag?
Testing with executable evidence. Round 1 completed 84 Trials and showed that tactical success can be strategically harmful, CAGE reward and foothold spread can conflict, richer model interpretation need not improve action, and organization can isolate compromised advice without creating new capability. The method survived; no Campaign engine or strategic ontology was promoted.
Ordivon Webtesting
Can the public site expose changing research judgment without becoming a second fact system?
Answered after contraction. The full Experiment/Finding/Decision/Event graph and explicit relation registry were removed. Project, Question, and Article metadata now drive Research, Writing, Current, article context, and curated System views with substantially less duplicate state.
Ordivon Worldopen
Does a thin World boundary prevent a real failure better than direct Host integration?
Supported as a corrected research boundary but still experimentally open. Link and Edge unification removed the wrong repository split; World W1 must now prove that correlated recovery creates more value than direct Host integration.
Ordivon Securitytesting
Does compiled opponent state retain value across seeds, held-out policies, Context loss, and Harness replacement?
Open after mixed Round 1 evidence. The local opponent-aware actor improved diagnosis and objective rate, one Codex strategic Trial recognized a switch, Hermes outcomes were unchanged, and no model Trial completed the genuine objective. No causal or transferable advantage is established.
Ordivon Computingtesting
Which contracts survive a second independent workload without being tailored to one repository?
Strongly supported but not fully closed. Game implemented the immutable Host workload profile across a TypeScript deterministic World, single- and multi-Agent paths, response-loss recovery, and exact replay, then removed thirteen duplicate truth tables without changing World outcomes. A second non-Game external domain remains useful pressure.
Ordivon Runtimeopen
Which remaining friction belongs above Runtime rather than becoming another primitive?
Strongly supported. Repeated dogfood earned request lookup, progress, repair, lifecycle retention, and receipts. Harness H2/R2 then proved Assignment and Harness Run correlation, exact replay, fresh-client recovery, and terminal evidence using four opaque foreign references with zero new Runtime semantic objects.
Complete archive
Every published claim remains dated.
Filter the complete record by document type. Later arguments may revise the working model without silently rewriting earlier claims.
01020304050607080910111213141516171819
Research note · Ordivon Game
Communication Is Gameplay State
Why agent communication needs identity, reachability, delivery state, and World-conditioned timing rather than decorative chat logs or a shared transcript.Research essay · Ordivon Computing
Creation, Judgment, and Recoverable Systems
The canonical public statement of Ordivon's project intent: creation, capability externalization, judgment, changing implementation costs, low governance, high recoverability, and participant freedom.Architecture guide · Ordivon Computing
From Tokens to Work: The Complete Agent Execution Stack
A first-principles walkthrough from token generation to tool execution, verification, TaskOutcome, and graph advancement, compared with OpenAI, Anthropic, Microsoft, Google, and Ordivon's tested boundaries.Engineering report · Ordivon Host / Game
One Authority, Thirteen Tables Deleted
Why Ordivon Game reused the Host contract without forcing every workload through one Python process, and how unique semantic authority enabled a net deletion of 641 implementation and test lines.Engineering report · Ordivon Game
Replay Without a Second Truth Store
How replay and evidence-linked diagnosis can remain pure deterministic projections of World, Host, Team, authority, Message, Provider, Effect, Observation, and verification records.Research report · Ordivon Computing
The Smaller Core That Survived Strong Baselines
How Core Work System Round 1 used strong classical baselines and live Codex/Hermes replacement to remove a Task Runtime, reject a generalized Context Kernel, shrink Effect, localize DecisionRequest, and retain provider-neutral Task state.Release note · Ordivon Game
Station Zero v0.1.0-alpha.1 — The First Source-Playable Release
What shipped in Station Zero v0.1.0-alpha.1, how the release journey is verified, what one coordination-only comparison proves, and which agent-native game claims remain open.Engineering report · Ordivon Host / Game
A Thin Host Can Improve Strategy Without Becoming a Planner
How Station Zero M2.1 improved Codex and Codex-to-Hermes strategy by compiling trustworthy decision semantics without forcing rank one, adding unrestricted memory, or introducing a hidden planner.Research note · Ordivon Computing
A Transcript Is Not a Task Database
Why transcripts and summaries remain useful cognitive evidence but should not become the authoritative database of durable agent work.Research note · Ordivon Runtime / World
UNKNOWN Is an Operational State, Not a Model Feeling
Why unresolved external reality must be represented as system state rather than inferred from model confidence or collapsed into failure.Research report · Ordivon Host
What Survived When Codex and Hermes Replaced Each Other Mid-Task
How one task remained coherent through Codex→Hermes and Hermes→Codex replacement, three injected faults, and materially different provider lifecycles.Architecture decision · Ordivon Host
Why Ordivon Needs a Harness—but Not a Universal Harness
Why H5 correctly rejected a common Codex/Hermes lifecycle while a separate Ordivon Harness remains strategically necessary for bare model APIs, local inference, and controlled agent-loop research.Research report · Ordivon Security
Winning the Move Can Lose the Contest
How Ordivon Security Round 1 used local dynamic opponents, CAGE Challenge 4, and bounded Hermes/Codex diagnostics to separate tactical success from strategic outcome without promoting a Campaign engine, organization ontology, strategic state, or custom cyber range.Research essay · Ordivon Computing
The Future Will Not Wait
A historical and systems argument for accelerating AI, robotics, science, verification, resilience, access, and cooperation instead of centering civilization on a global frontier slowdown.Architecture report · Ordivon Host
Why Task Continuity Belongs Above Execution
Why durable task continuity belongs in Ordivon Host above replaceable model sessions and Ordivon Runtime process execution.Architecture report · Ordivon World
Why Link and Edge Became One Ordivon World Boundary
Why Ordivon first separated connectivity and external execution, then unified both prototypes around one complete Task-to-World Interaction.Engineering report · Ordivon Runtime
How Ordivon Runtime Grew Beyond Its Ten-Tool Core
How a production-tested ten-tool runtime became a thirteen-tool execution and recovery system without absorbing task meaning or external-world semantics.Release · Ordivon Runtime
Ordivon Runtime: From Governance Platform to Durable Execution
Ordivon Runtime closes its M0–M7 core with ten production-tested tools for durable agent execution, observation, cancellation, artifacts, and recovery.Design argument · Related to Ordivon Runtime