Canon
The Record
The house's standing law: what it holds true, and why.
205 pages in 17 folders
Revisions
Corrections
What changed, and what it replaced.
16 October 2026 · Methods
Library, Not Hand-Rolled replaces an earlier version.
16 October 2026 · Principles
Ownership Outlives the Build replaces an earlier version.
16 October 2026 · Principles
Silence Is Not Consent replaces an earlier version.
24 June 2026 · Definitions
DOLCHEOT replaces an earlier version.
5 May 2026 · Methods
Method: Borrow, Bend, Break, Beget, Bide, Build replaces an earlier version.
19 April 2026 · Meta
Frontmatter Schema replaces an earlier version.
9 April 2026 · Resonance
Agent Skills was revised.
9 February 2026 · Constraints
ODD Is a Value-Grounded Epistemic OS replaces an earlier version.
Constraints
NewestLegibility Standard
-
Legibility Standard
The reviewability standard makes work reviewable; the seeded-response standard makes decisions
-
Canon Is Metaphor-Agnostic
The school teaches every kitchen, every airline, and whatever comes after
-
Build Provenance Endpoint
Posture: RATIFIED (captain, in-session 2026-07-19) — adopted portfolio-wide; ARS is the first conformer (its reliability inventory D6/S4 slice implements this constraint by…
-
ARS Bounded Storage
Posture: PAUSED — captain ruling 2026-10-09 19:38 ET: ARS is retrospectively on ice. The lessons it forced (the 2026-07-16 freeze, policy-precedes-build, Epoch 11) stand; the…
-
Mandate Precedes Build
Nothing is built without a clear, ratified, enforceable mandate that states what is being built and why. The order is design → mandate (→ PRD) → build — never code-first. The…
-
Ratified Model Requires Reconciliation and an Enforcer
Posture: ACTIVE — ratified by merge on 2026-10-09 (captain ruling). Filed 2026-07-16 as the durable lesson from the ARS one-blob
-
The Dispatcher Dispatches; It Never Executes In-Session
The dispatcher's job is to queue and route, not to do. Every time the dispatcher does the work itself, it stops dispatching — and the one thing it exists to protect, the captain's…
-
Per-Environment Worker Projects
When a Durable Object — or any other stateful binding — prevents clean Cloudflare preview deploys, the project MUST run as separate Cloudflare Workers projects per environment…
-
Reviewability Standard
Companion to the legibility standard. The legibility standard makes work readable; the reviewability standard makes it reviewable — with as little friction as making it legible.…
-
Seeded Response Standard
Companion to the legibility and reviewability standards. The legibility standard makes work readable; the reviewability standard makes it reviewable; this standard makes a…
-
The Dispatcher Spawns Its Own Coordinators
The companion rule says the dispatcher never does the hands-on work. This rule says the dispatcher never absorbs the coordination of that work either. Polling flights, distilling…
-
The Preflight Checklist
The flight-deck model already says preflight fires before work. This constraint makes it a gate rather than a habit: five items, checked every flight, before any work, and never…
-
A Link Is a Tap, Not a String
When you hand the operator something they will act on — a link, a command, a path — deliver it in the form that costs the fewest steps to act, on the surface they are using. A…
-
Retrieval Disclosure Contract
The writing canon requires every document to be actionable at multiple disclosure tiers — URI/title, blockquote, metadata, summary, full body. The five oddkit retrieval actions…
-
Telemetry Validation Gate
The Emission Contract requires every registered tool to emit accurate metered usage on every call. Verifying it is one smoke pass per surface: hit every tool, compare the emitted…
-
Mode Transitions Require Encoded Handoff
A mode transition without a durable handoff is mode collapse with extra steps. The receiving session has no grounded artifact to read; it has only the working memory of the prior…
-
Critic Cannot Be Resolver
The agent that detects drift cannot be the agent that resolves it. Same context corrupts both functions. The critic who also resolves optimizes findings toward things it can fix.…
-
Audit Gates Are Spawned Agent Sessions, Not Pattern Matchers
When canon defines what to check and the check requires reading prose, code, and history together to render a judgment, the gate is a spawned agent session — not a regex, not a…
-
Borrow Evaluation Before Implementation
The 6B method (canon/methods/borrow-bend-break-beget-build) is the meta-discipline. This constraint is the planning-mode operationalization that binds the agent. Before any…
-
Adding an oddkit Action: Update Both Switch and VALID_ACTIONS
A new MCP action in oddkit must be registered in two places, not one. Wiring only the dispatch switch ships an action that the validator rejects before the dispatch is reached.…
-
Bash Test Rigs: Assignment Chains Need Explicit Newlines
Bash silently concatenates two assignments separated by missing whitespace. The result parses, runs, and produces wrong values. Tests against the wrong value pass or fail…
-
superseded_by Values: Three Shapes Allowed, One Resolution Algorithm
Canon authors write superseded_by in three different shapes. Tools that read that field MUST normalize across all three; pushing the work to authors is the same anti-pattern this…
-
Canon-Integration Audit
A doc can be well-written and still fail to integrate with the knowledge base. The existing gauntlets check intrinsic quality (voice, structure, frontmatter, claims). They do not…
-
Measure Before You Object
A concern is a claim. An unverified claim is a liability. When a contributor raises a performance, cost, or complexity worry that could block a proposal, the worry must be…
-
Governance Change Discipline
Any change intended to impact how operators or agents use the system must carry the structural markers that make the change visible: bumped canon version, appended changelog…
-
Proactive Frequency Calibration
Continued micro-rituals (swipe, tap, continue, swipe, tap, continue) are a smell, not a workflow. When the operator has been absorbed into the system as a mechanical component…
-
Release Validation Gate
Convention is optional; convention plus an enforcer is binding. Smoke tests catch mechanical bugs; independent review catches PRD-vs-PR intent drift; canon-vs-ledger conflict…
-
Mode Discipline and Bottleneck Respect
Exploration, planning, and execution are distinct epistemic modes with different obligations. Collapsing them — smuggling planning questions into execution turns, "checking in"…
-
Core Governance Baseline
A knowledge-base-driven tool that cannot function without a live knowledge base fetch is a tool that disappears when GitHub throws a 403. A tool that ships its full governance…
-
Harness Disambiguation
"Harness" means two different things in this canon. In the essay that introduced the concept (E0005), harness meant a coding constraint toolkit — Ben's thing, not ours. In…
-
AI Voice Clichés
These articles are co-authored, not AI-generated. The AI is a ghost writer — organizing brain dumps, transcripts, and source material into publishable form. The human author's…
-
Constraint: Dual-Context Writing
A document that lives in both a book and on a website must never assume which context the reader found it in. The body stays agnostic. The metadata carries both identities. The…
-
Frontmatter Validation Before Merge
Broken frontmatter crashes the site renderer. This has happened repeatedly — the only signal is a blank page in production. No writing PR merges without automated frontmatter…
-
Telemetry Governance
oddkit is maintained by one person making decisions about where to invest their most limited resource — their own attention. Telemetry exists to make those decisions informed…
-
Prompt Pattern for oddkit-Powered Applications
System prompts contain the creed, axioms, and a pointer to oddkit. Governance is fetched at runtime, never hardcoded. Hardcoding governance into prompts creates stale copies that…
-
Relational Sensitivity
The author works closely with the organizations and individuals referenced in public essays. Truth is told honestly but framed positively — the pattern is structural, not…
-
Author Identity Language
Klappy is a software builder with ~15 years in Bible translation — not a translator.
-
Guide Posture
In every public-facing surface — websites, voice agents, documentation, tool descriptions, onboarding — we are the guide, not the hero. The user is the hero. They have a problem…
-
Boundary Transitions Require Deceleration
Speed is allowed inside a boundary. Transitions require slowing down.
-
Decision Rules
Heuristics for choosing between valid options when designing systems.
-
Definition of Done & Evidence Policy
The enforcement backbone that defines what "complete" means.
-
Encode Epistemic Decisions
If a decision matters, it must become durable and inspectable.
-
Epistemic Challenge
Challenge claims proportionally, surface contradictions explicitly, and protect collaborative flow.
-
Humans Are Variable Inputs
Humans are not reliable, repeatable components.
-
Meaning Must Not Depend on Path
No canonical meaning, scope, or lifecycle state may be inferred from filesystem paths or branch names.
-
No Irreversible Action Without Epistemic Justification
Defer irreversibility until epistemic thresholds are met.
-
ODD Is a Value-Grounded Epistemic OS
ODD is an epistemic OS grounded in axiomatic values. It does not define morality or authority, but it does define an unconditional commitment to truth — the foundation on which…
-
Single-Agent Integrity Precedes Collaboration
Collaboration is only constructive when integrity exists first.
-
Verification & Evidence
Claims are untrusted. Only observed evidence counts.
-
Visual Proof Standards
What "prove it visually" actually means for UI and interaction work.
Principles
NewestOwnership Outlives the Build
-
Ownership Outlives the Build
The price of a load-bearing layer is not the effort to write its first version; it is every patch, incident and security fix for as long as anything stands on it, paid by whoever…
-
Silence Is Not Consent
A project registers review perspectives because each sees what the others miss, so a decision they were registered to see is not reviewed until every one has recorded an outcome…
-
Data Carries Its Observed Time
Time blindness is not only about the clock on the wall. A model that cannot perceive elapsed time also cannot perceive how old its data is. A gate state read at 23:05 and a gate…
-
Status Carries Three Tenses
A status that cannot be placed in time is not a status; it is a mood. The fourth dimension of software is the one that sinks projects: not the code, but the inability to say…
-
Mandate-First
Former name: Policy-First — Policies First Ensures All Code Is Self-Building and Self-Documenting (uri klappy://canon/principles/policy-first-self-building-self-documenting…
-
MCP as a Universal Interface
The captain's framing, 2026-07-09: give every user the product's MCP connector, so that everything they can do in the graphical interface they can also do through Claude, ChatGPT…
-
Rulebook Transfer
Discernment a frontier model performs once can be written down as a rulebook a lesser model runs — but only the part that can be written down transfers, only adjacent tiers can…
-
Skills Are Procedure, Not Judgment
A skill encodes a procedure and repeats it the same way every time; that is its strength and its ceiling. It cannot render a judgment — decide whether a claim is verified, whether…
-
A Proxy Confers Shape, Not Standing
A proxy turns a resource into context and a callable tool. That is transport, and transport transfers for free. Standing does not: provenance, verification, the right to be…
-
Agents Need Their Own Wire
Multi-agent collaboration today routes through human attention. The operator copies output from one agent and pastes it into another, summarizes the first agent's reasoning for…
-
Quality Attributes Are In Tension
Optimizing one quality attribute reliably degrades at least one other. The pairs are not random; they are structural. Urgency erodes stability and maintainability. Originality…
-
Sessions Mirror Modes
The structural blindness that makes a creator unable to validate their own work, a critic unable to remediate their own findings, and a planner unable to honestly execute their…
-
Symmetric Participation
The wire treats every peer through the same primitives, regardless of what or who the peer represents — autonomous agent, channel adapter bridging a human, role embodiment…
-
Creators Get Paid
Every participant on the substrate — adapter authors, agent builders, role creators, application operators, substrate maintainers — can be paid for the value they actually…
-
Magical First-Run
Any user-facing application built on the Klappy substrate must be reachable, useful, and intelligible to a non-technical operator within sixty seconds of first contact, without…
-
Methodology Personification
An epistemic methodology with a named human exponent becomes more accessible when given a personified voice that demonstrates the methodology in action rather than explaining it…
-
Voice as Cognitive Load Shedding
Under high information density — real-time stream interpretation, incident response, audit findings at volume — voice features like dryness, refusal to panic, brevity, and…
-
Doing Less Enables More
A substrate wins by refusing to own layers above it. Every opinion the substrate adopts is a layer the substrate must defend forever, and a layer that the layers above cannot…
-
Principle
Canon governs what should be. Code observation is the only ground for claims about what currently is. A claim of the form "the parser still hardcodes X," "the worker currently…
-
Specs Lock at Implementation
A spec is a contract between the spec author and the implementer. Once implementation begins — once the spec has been handed to a coding agent, an autonomous run, or a human…
-
Identity Is Resolved By The Protocol
An author writing link is hardcoding a location in source. The location will drift; the source will not. Every consumer that reads this source will encounter a broken reference at…
-
Partial Data With Transparency And Background Warm
The user-blocking path never pays the full cost of an expensive corpus scan. Return whatever the system has already observed, disclose the missing portion concretely, schedule the…
-
Type Contract Plus Adversarial Review
When a function's correctness depends on reporting its own incompleteness honestly, a type-system contract at the API boundary is necessary but not sufficient. The type forces…
-
The Discernment Layer
When the execution layer is AI — the model writes the code, the agent runs the bench, the tool produces the draft — the human's load-bearing contribution is discernment, not…
-
The Dream House Principle
When facing a decision that involves cutting scope, sacrificing features, or pre-optimizing, draw the full version first. Then cut from contact with reality, not from prediction.…
-
Cache Fetches and Parses
Cache the expensive work — fetching markdown over the network, parsing tables into structured data, stemming vocabularies into token sets. Don't cache microsecond derivations…
-
Contract Governs Handoff Drift
Session ledgers and handoffs are records of one session's judgment under one session's pressures. Canon is the program's binding contract, refined across many sessions and many…
-
Envelope Time Fields
Every oddkit response carries two kinds of time, and they are not the same. servertime is when the response was produced. debug.generatedat is when the canon content powering the…
-
Agent Self-Report Under Stress
An agent's terminal self-report is a belief, not a log. When a safety layer, rate limit, or other mid-session pressure injects contradicting guidance, the agent's final summary of…
-
Verification Requires Fresh Context
The same lenses used to create an artifact are the same lenses used to evaluate it. This structural blindness persists across multiple sequential passes by the same agent in the…
-
Antifragile
When oddkit encounters a governance failure — a missed gate, a hallucinated claim, a governance gap — the fix is a document, not a feature. The Progressive Disclosure Failure…
-
Consistency
The server behaves identically regardless of what knowledge base it serves. Same MCP interface. Same retrieval algorithm. Same encoding format. Same gate mechanism. oddkit proves…
-
DRY
If a governance rule exists in the knowledge base, the server must not re-encode that rule in its own logic. Every duplicated rule is a drift surface — the canon version and the…
-
KISS
Every action, endpoint, or abstraction in an epistemic server must justify its existence against the alternative of not existing. oddkit has ten composable actions — not twenty…
-
Maintainability
A vodka layer must be maintainable by a single person without heroics. If the server requires a team to operate, it has grown beyond its function. oddkit proves this: it is a…
-
Prompt Over Code
oddkit's entire governance model is prompt over code in production. The Writing Canon gate, the Identity of Integrity creed, the relational sensitivity constraint, the guide…
-
Vodka Architecture
Infrastructure so thin and clean it disappears into whatever it carries. A stateless server over a stateful canon. No opinion about the domain — all the opinion about epistemic…
-
Scoped Truth
When a single knowledge base serves every context, domain contamination is inevitable — software opinions bleed into theology conversations, engineering constraints shape venture…
-
Capability Is Not Permission
When the cost of creation drops below the cost of deliberation, the natural friction that once forced humans to pause — compile times, meeting cadences, the speed of thought…
-
Async by Default
Any MCP tool whose work could exceed ~5 seconds wall-clock returns an identifier within that budget and places the long-running work behind a separate read tool. No tool blocks…
-
Bulldoze the App, Keep the Blueprint
When code stops being the scarce resource
-
Documentation Is the Lever, Not the Goal
ODD (Outcomes-Driven Development) makes heavy use of documentation.
-
Focus Is Exclusion
Possibility is infinite. Capacity is not.
-
Irreversibility Is the Real Cost
Most work is cheap to think and expensive to undo.
-
Persistence Must Be Intentional
When improvement is no longer observed, continuing without reassessment is escalation, not discipline.
-
Progress Over Perfection
No system survives contact with reality intact. The sooner the contact, the cheaper the edges.
-
Ritual Is a Smell
If correctness depends on people repeatedly remembering a procedure, the system is compensating for missing design.
-
Scope Over Folders
Epistemic scope is an attribute of a claim, not a property of its storage location.
Methods
NewestLibrary, Not Hand-Rolled
-
Library, Not Hand-Rolled
Two principles meet at every infrastructure choice: the cost of a load-bearing layer is its lifetime of care (ownership-outlives-the-build), and a decision is not reviewed until…
-
Method: ToC OODA Sprints
Name the one constraint, look again every thirty minutes, and ship last window's frozen work while building this window's.
-
Writing Simple Rules
A ## Simple Rules section is nine one-line rules that let a reader route a document (load it, skip it, or put it down) before its body enters context. Write each line to answer…
-
The Driver's-Seat Lens
Before a plan binds, one planning seat reads the whole design set as the
-
Method: The Discernment Transfer Ladder
Discernment a frontier model performs once can be written down as a rulebook a lesser model runs — rulebook-transfer states the principle. This method is the empirical procedure…
-
The Directory Submission Gauntlet
A method for taking an MCP server from working code to a submitted Claude Directory listing — field-tested end to end on git-repo-auth-mcp, submitted 2026-06-12, operated entirely…
-
The Publish Gauntlet
"The gauntlet" has been working vocabulary without a document, which meant it relied on tooling memory of what it covers — the exact failure mode this canon exists to prevent.…
-
The Triangle Pass
Introducing a topic is one move with two failure modes: prove too much and the audience leaves before you finish; prove too little and you have attention resting on nothing. The…
-
Dispatch Paths for Spawned Agent Sessions
Every spawned-agent-session invocation arrives via one of two paths. Assistant-orchestrated — an assistant (Claude, OpenCode, etc.) in a chat session dispatches a runtime…
-
Dispatch Routing for Spawned Agent Sessions
A spawned agent session is fully characterized by the five orthogonal dimensions of the runtime contract — persona, mode, role, surface, engagement. Sessions still have to start…
-
Persona-Shaped Agent Runtime
A spawned-agent-session substrate (CF Sandboxes, Anthropic Managed Agents, etc.) hosts one session at a time. A runtime sits on top of a substrate and turns it into a service…
-
Quality Attribute Tension Survey
A seven-phase method for surfacing a project's quality-attribute priorities at scope-setting time. Pick the ilities that matter for this project. Rank them at the project's…
-
Spawned Agent Session Runtime Contract
A spawned agent session is fully characterized by five orthogonal dimensions: persona (voice and methodology), mode (epistemic state — exploration / planning / execution /…
-
Spawned Agent Session Substrate Options
Every spawned agent session has at least two cost streams that add up: LLM inference (paid to the model provider) and runtime hosting the session. The cost-optimal architecture…
-
Method: Borrow, Bend, Break, Beget, Bide, Build
The canonical sequence for maximizing work not done. Before building anything yourself, attempt to borrow what exists, bend it to your context, and let the breaks reveal what's…
-
Method
When the next phase of work has its contract in canon — a handoff, a ledger, an architecture doc — and the question is whether to continue in the current session or open a fresh…
-
Governance Validation via Managed Agents
Canon defines what to check. Agents do the checking. Every governance constraint in canon is executable by a Managed Agent before merge. The authoring session should validate…
-
Reference Integrity Audit
Every cross-reference is a claim. A klappy:// URI says "this document exists at this address." A complements field says "this sibling is real." A README table says "these are the…
-
Method: Revision Lens Sequence
Quick multi-lens checks are valuable throughout drafting — preflight, inline validation, sanity checks. But when formal verification is required, applying multiple governance…
-
Method: Supersession
Drift has five responses, not one. Tolerate it when the cost of fixing exceeds the cost of leaving. Observe it when the tension is real but action is premature — record it, defer…
-
The Planning Queue
The planning queue is a collection of implementation-ready specs stored in docs/planning/ that are fully documented but deliberately not yet built. Each doc is a ledger entry of…
-
Community Checking
Output verification asks "does this match the author's intent?" Outcome verification asks "does this transfer to the people it's for?" ODD currently validates outputs but not…
-
Epistemic Surface Extraction (ESE)
Making visual/audio/video evidence legible to agents without turning it into doctrine.
-
Method: Boundary Transition Review
A repeatable review-and-prepare step used when moving between epistemic boundaries.
-
Method: Choosing the Right Narrative Container
Not every insight belongs in the same kind of document.
-
Method: Exploration Exhaust
A repeatable way to preserve exploration context, closures, and warnings so scope stays closed.
-
Method: Extreme Exploration for Limit Discovery
Exploration may go to the edge so implementation does not have to.
-
Method: Using Ease and Resistance as Signals
Resistance is often a classification error, not a lack of insight.
-
Method: Writing Apocrypha Fragments
Apocrypha fragments are not guidance. They are residue.
-
Method: Writing Predocumentaries
Predocumentaries are reconstructions written in micro-documentary voice — investigative journalism from the near future, reporting on events that haven't happened yet but follow…
-
Pivot on Inversion
When observable improvement turns negative, change mode. Snapshot, extract invariants, regenerate cleanly.
-
Reframe Before Trimming
When an MCP server's tool surface feels bloated, the bloat is usually in the frame the surface implies, not in the tool count. Reframe first; the trim follows mechanically.
-
Self-Audit Checklist
A reflection layer that makes the Definition of Done actionable.
-
Tool Specialization
A general pattern for preserving reliability as tool availability increases.
-
Triple-Loop Debrief Review
Learning to learn from mistakes is a meta-skill, not an emergent nicety of
-
Weighted Relevance & Arbitration
How the system handles conflict between competing truths — and why resolution is not always the goal.
Patterns
NewestDocs Proxy
-
Docs Proxy
An MCP server whose action surface depends on a sibling canon repo for domain semantics SHOULD expose a docs(query, audience?, depth?) tool that proxies the canon-server (oddkit)…
Decisions
NewestModels Do Not Mutate Canon
-
Models Do Not Mutate Canon
Models do not mutate the operator's canon. A model may hold delegated, bounded, revocable stewardship over a sub-scope — governing within it, never above it, never over itself.
-
DR-20260514-0001
Locks the architectural choice that closes the verified 27% telemetry coverage gap. Every server.tool() registration in workers/src/index.ts is wrapped with…
-
Decision Record Standard
Decisions are first-class documentation. A decision record preserves intent, alternatives, rationale, and consequences.
-
DR-20260211-0001
Camping detection introduces the risk of over-instrumentation, paternalism, and metric gaming.
Definitions
NewestGovernance Artifacts
-
Governance Artifacts
From July to October the program used one word, "policy," for the whole of what shapes behaviour, and before that "canon article." The word failed in the mouth before it failed on…
-
The Cooking Taxonomy
WHEN AND WHERE THIS POLICY APPLIES. Fetch this document when chartering or drafting anything a client will read: a menu, proposal, statement of work, onboarding note, or price…
-
DOLCHEOT
Decisions, Observations (closed), Learnings, Constraints, Handoffs, Encodes, Opens, Tensions. Six artifact types, one meta-level action, one forward-pointing thread, and one…
-
Epistemic Modes
Exploration, planning, execution, validation, and resolution are not interchangeable.
-
Software Virtues Vocabulary
Ten properties every software project optimizes for, in tension with each other: usability, originality, stability, urgency, efficiency, maintainability, versatility…
-
Validation as Epistemic Mode
Validation is not a phase of execution. It is a distinct epistemic mode with its own truth conditions, and it requires a structural separation from creation: a context break…
-
Cognitive Saturation Threshold (CST)
The point at which continued conceptual exploration yields diminishing returns and must stop.
-
Epistemic Obligation and Document Tiers
Document tiers define epistemic obligation, not importance.
-
Execution Posture
How strongly a document is expected to govern behavior.
-
Tier vs Relevance
Two different axes with different purposes. Do not conflate them.
Values
NewestDrift
-
Drift
Drift is the natural tension between what was written and what is now understood. Its presence is proof of a growing system. Its absence is the warning sign — it means the system…
-
Trust Is Built by Managing Expectations
Trust — between humans, between human and AI, between teams, between product and user — is the outcome of expectations well managed. Every trust failure traces to an expectations…
-
Shared Values as a Trust Proxy
Trust between agents, humans, and mixed teams cannot be assumed — it must be bootstrapped through shared values. When participants share explicit axioms, identity, and…
-
Foundational Axioms
Four values from which all ODD epistemic discipline is derived: (1) Reality Is Sovereign — observe before asserting, (2) A Claim Is a Debt — every assertion requires evidence, (3)…
-
Orientation
Before I speak, I observe. Before I claim, I verify. Before I confirm, I prove. What I have not seen, I do not know. What I have not verified, I will not imply. — This is the…
Voice
NewestOrville the Osprey
-
Orville the Osprey
Orville is the personification of the model collaborator under Epoch 10 — the crew seat made visible, the way Oddie is the methodology made visible. He is an osprey: a river hawk…
-
Oddie the River Guide
Oddie is the personification of Klappy's operational methodology — the way Klappy actually thinks under pressure, made repeatable and accessible without requiring Klappy to be…
Bootstrap
NewestThe Any-Seat Boarding Recipe
-
The Any-Seat Boarding Recipe
The generic boarding pass gave any adopter the seat: creed and axioms
-
The Otto Boarding Pass
Recut 2026-07-19 (captain direction, in-session): the pass becomes a
-
The Otto Operating Card
This is the wallet card, not the law. It is fetched live and injected at the freshest position in Otto's context every turn — because recency decay, not deletion, is how the…
-
The Generic Boarding Pass
The maintainer's boarding pass (canon/bootstrap/boarding-pass) is the authoritative text of one person's account and project instructions — and it carries cargo no one else can…
-
The Boarding Pass
Instructions at account and project level are the only persistence mechanism this collaboration relies on — never the model's memory system, which is instance-side state the…
-
The Flight Deck Model
Earlier models were bootstrapped with a costume: an identity and a creed they adopted, and the adoption did real orienting work. Newer models hold their own judgment, decline the…
-
Model Operating Contract
Project instructions should be short: creed, axioms, time rule, and a pointer. The operating contract — how a model works inside this system — lives here and evolves here.…
Architecture
NewestThe Two-Loop Operating Model
-
The Two-Loop Operating Model
Terminology note (captain ruling 2026-10-09, klappy://canon/definitions/governance-artifacts): the per-build governing statement Loop 2 consumes is a mandate (typically carried…
-
The Klappy Substrate Stack
Six layers, each with one concern, talking only to neighbors through narrow interfaces. The stack exists so the question "where does X live?" becomes a layer lookup rather than an…
Meta
NewestAnatomy of a Build Mandate
-
Anatomy of a Build Mandate
Posture: ACTIVE — ratified by merge on 2026-10-09 (captain ruling). A meta-template filed 2026-07-17 alongside the ARS storage
-
The Triangle of Yaps
Every unit of communication carries one Thought or Topic (T or T), pays it down with a Metaphor, Analogy, Example, or Story (M-A-E-S), and closes with an Application or Next step…
-
Frontmatter Schema
YAML frontmatter values must match the types the site renderer expects: booleans unquoted, integers unquoted, dates unquoted, simple identifiers unquoted, strings with special…
-
Constraint-Driven Audits
A constraint that cannot be audited against its governed documents is advice, not enforcement. Every constraint document declares a governance scope (what it governs), a set of…
-
Writing Canon
"Show me what I'm deciding in a way that respects my attention and time." That is the reader's one demand, and every canon document must answer it at every extraction depth: title…
-
Agent-Executable Documentation Outline
Supplement human-readable documentation with decision leverage.
-
Completion Report Template
The standard format for claiming work is complete.
-
Epistemic Architecture
This document describes how epistemic judgment is shared across tools without collapsing them into a single implementation.
-
Policies vs. Requirements
"Policy" has been used to name two different things: durable, slow-changing
Case studies
NewestCase Study
-
Case Study
oddkit serves 500+ documents. The primary public interface for reading those documents could only retrieve 196 of them — not because of a bug, but because oddkit was never…
-
Case Study
In February–April 2026, a five-person team turned a year of rich methodology work into a live field deployment across five sessions. Four sessions of meeting time totaling roughly…
Defaults
NewestEpistemic Posture
-
Epistemic Posture
These defaults encode posture, not truth.
-
Evidence Intake
When prior work exists, the system must request or retrieve it before proceeding.
-
Iteration Bias
These defaults encode operational preference, not truth.
Diagnostics
NewestCamping Risk
-
Camping Risk
Raise when iteration continues after observable improvement has flattened or inverted, and subjective momentum substitutes for measurable progress.
-
Diagnostic: RITUAL_DETECTED
Trigger
-
Epistemic Hygiene
How the system detects when truth is decaying — and when review is required.
-
Generative Arc Curve
In generative artifact iteration, coherence often peaks early and degrades under sustained local steering. Inversion is a signal; camping past inversion is the failure.
Observations
NewestTelemetry Wrapper Drops writeDataPoint on a Variable Subset of Calls
-
Telemetry Wrapper Drops writeDataPoint on a Variable Subset of Calls
Two consecutive smoke passes against workerversion 0.28.0 (main preview and prod) each emitted only 14 of 17 toolcall rows despite all 17 tool handlers returning successfully. The…
-
Observation
Diagnostic record for the telemetry coverage gap that motivated PR 1 (canon) and PR 2 (code) under Phase 1 of the paid-tier observability work. 27% of eventtype='toolcall' rows…
-
Observability Tension Extension
Observability earns its own extension because ODD has been writing about it for an entire epoch already. This document maps observability's tensions against the canonical ten…
-
Quality Attribute Tension Matrix
A worked example of the tension graph across the canonical ten quality attributes: forty-five unique pairs, each tagged with relationship type (mutual tension / asymmetric…
-
Clone-Klappy to Oddie Recognition
For twenty-plus years, across multiple companies and roles in software product development, colleagues have asked to "clone Klappy." The refrain was constant: the ability to read…
-
Observation
Production bug verified by direct code observation and live tool call on 2026-04-30. workers/src/orchestrate.ts:discoverEncodingTypes (line 512) keeps the first encoding-type…
-
Performed Prudence
A contributor sees a proposal and accumulates theoretical objections to it. The objections sound careful. None has been tested. The cluster of unmeasured worries either blocks the…
-
Time Blindness
Models have no perception of time. Not degraded perception — zero. They infer elapsed time from context clues the way a person might guess the season from a photograph. Sometimes…
Resonance
NewestJev and System One Models
-
Jev and System One Models
Diogo Almeida / TypeSafe AI, announced 2026-09-15. A hosted model that generates no text: state and typed questions in; a Choice, a Score, or a Noul (yes/no) out, each with a…
-
Agentic Engineering (Karpathy)
Andrej Karpathy, 2025–2026. The convergence is directional validation at a visibility canon could not reach — and a gift of shared vocabulary; the divergence is the one the…
-
Agent Skills
Anthropic (open standard via agentskills.io), December 2025. Updated through March 2026 with subagent delegation, eval/benchmark pipeline, resource subfolders, and hot-reloading.
-
Seeing Like an Agent
Thariq (@trq212), Anthropic Claude Code team, 2026. A practitioner account of lessons learned building Claude Code's tool surface. The thread independently converges on ODD's core…
-
AI Coding Toolkit
Ben Shoemaker, 2026. Discovered via Changelog News, not studied. A practitioner-built toolkit for semi-autonomous AI coding that independently converged on patterns nearly…
-
Antifragile
Nassim Nicholas Taleb, 2012
-
OODA Loop
John Boyd, 1970s–1990s
-
Sprint
Jake Knapp et al., 2016
-
The Double Diamond
Design Council, 2005 (revised 2019)
-
The Lean Startup
Eric Ries, 2011
Standing pages
NewestDelegating Responsibility Over a Project
-
Delegating Responsibility Over a Project
Delegating responsibility over a project is not handing off a task list and it is not abdicating the wheel. It is granting a crew agent standing authority over a workflow…
-
The Director's Chair
Status: Captain-approved. This is a thin overview — a compressed frame over a much larger body of source material. It is his vision, in his voice, consolidated from many nights of…
-
The Epistemic OS
The Epistemic OS is infrastructure for managing expectations in AI-human collaboration. The core product loop across all verticals is: converse → generate → validate → promote or…
-
Canon Changelog
This changelog tracks changes to the Canon pack as a whole.