Actionist Knowledge Docs
Generated knowledge spine · knowledge/02-ASSUMPTION-LEDGER.md

02 · Assumption ledger

Observed, inferred, hypothesis, unknown and rejected claims with falsifiers.

Generated from knowledge/02-ASSUMPTION-LEDGER.md · regenerate with node site/system-map/scripts/generate-docs.mjs

First-principles assumption ledger

Vocabulary:

  • Observed: supported by a direct local receipt or measured artifact.
  • Inferred: reasoned from observations but not directly tested end to end.
  • Hypothesis: falsifiable proposition awaiting an experiment.
  • Unknown: insufficient evidence.
  • Rejected: contradicted or replaced by stronger reasoning/evidence.
IDAssumption or claimStateWhat supports or challenges itNext falsifier
A01Clients will pay for bounded workflow outcomes faster than for generic app generationhypothesisActionist solution catalogue and public demand signalsInterview/paid pilot around one workflow
A02Demand across 17 industries compresses into reusable atoms and archetypesinferred12 atoms, 10 archetypes, repeated case/portal/CRM/finance patternsBlindly classify new industries and measure fit
A03We need 17 different technical stacksrejectedShared archetypes and atomsNone; preserve only industry deltas
A04A repository is the reusable unitrejectedRepositories contain multiple capabilities and assumptionsReuse-shape analysis on representative sources
A05Every useful source should become the same block shaperejectedSISOCRM service/transplant/module evidenceThree-shape pilot
A06Most donor applications need some branding/onboarding/settings/navigation adaptationobserved/inferredSISOCRM and user observation; donor ownership patternsMeasure adaptation categories across 10 donors
A07Adaptation is always only 1–2% of the workunknownCosmetic delta may be small; auth/data/runtime changes can be largeRecord hours and changed boundaries per pilot
A08Postgres is the best default for new owned B2B transactional stateinferredEcosystem fit, local donor stack, SISOCRM reasoningWorkload-driven data-plane bake-off
A09Every block must internally use PostgresrejectedIntact donors and local/ephemeral/file/search state differNone; standardize typed data ports
A10One ORM should be part of the universal block contractrejectedAdapter-local ORM reasoningBind one data port to two implementations
A11One owner per table and migration is mandatoryobserved/principleSISOCRM donor data-layer analysisArchitecture review of any proposed shared table
A12Shared cross-service views should use events/read models rather than shared-table mutationinferredSISOCRM separate-schema/event envelope reasoningImplement one cross-system read model
A13The current Block Contract is the correct universal objectrejectedPhase 8 richness plus composition couplingThin-contract dry run
A14Capability, packaging, host binding, qualification and release should be separate recordshypothesisFirst-principles ownership separationThree-shape contract sufficiency test
A15The ISSO five-area shell is universalunprovenAutoSaaS convention onlyCompare against case/portal archetype task success
A16Generic dashboards are the best initial nicherejected as strategic pilotSupply is saturated; weak differentiationUse only as tooling/control case
A17Case/workflow or portal is a stronger framework testinferredDemand-rich, template-thin shelf resultScoped pilot comparison
A18The local 21st stores provide substantial UI precedent supplyobserved8,515 joined identities; 3,506 source-bearing legacy entriesPer-item quality and duplication sample
A19Corpus size equals reusable component supplyrejectedBundle/metadata/source differences and zero admitted blocksQualification funnel metrics
A20A 1.3M-repository local index exists and is currently queryableunknownOperator recollection; no authoritative path in current receiptsLocate owning manifest/query interface
A21The 850k/80k Mini corpus is the same assetunknownPhase 9 could not resolve path or identityOperator path or machine inventory receipt
A22GitHub stars predict production adaptation qualityunproven/weakBroad corpus work did not test source execution or adaptationCorrelate metadata with pilot effort/defects
A2317,000 matrix observations meant 1,700 complete reusable repositoriesrejectedOnly 270 complete pairs at measured baselineMaintain denominator semantics
A24The 118-surface competitor register is exhaustiverejectedExplicit expansion candidates and missing FlutterFlow correctionDeclared stopping rule and universe definition
A25The 144-feature taxonomy is stable and fully mapped to 118 surfacesunknownStrict census mostly unknown; 68-vs-144 taxonomy conflictCanonical taxonomy merge plus evidence review
A26Connector catalogues can be adopted with their storage layersrejectedOpenConnector admin token exposed global connectionsOwn tenant connection store
A27OpenConnector catalogue/OAuth/action mechanics are reusableobserved at spike level1,445 providers, 15,156 actions, auth/action/OAuth execution receiptsHost-scoped connector pilot
A28Models should compose applications directly from raw retrievalrejectedPhase 8 plan-then-fill and compatibility reasoningKeep deterministic elimination layer
A29A deterministic solver can materially reduce model errors and tokenshypothesis26 fixtures and 10 solver checks are design evidence onlyRun solver/model comparative eval
A30Cheap models can deliver the final assembly once context is constrainedhypothesis20-task falsifiable matrix; unexecutedAuthorized model eval with fixed fixtures
A31Fully automatic repo-to-block conversion is desirablerejectedShape choice and semantic adaptation require judgmentAutomate evidence, preserve named decision points
A32Existing mature services should usually remain downstream servicesinferredSISOCRM engine-shaped donor reasoningUpgrade/maintenance comparison against transplant
A33Product-defining surfaces should often be owned/transplantedinferredTwenty record-surface reasoningMeasure donor-upgrade friction and UX control
A34The Actionist host can absorb donor identity/settings/navigation cleanlyunknownLocal integrations show partial patternsHost-contract pilot with one mature donor
A35Standardized settings, onboarding and branding can be adapter-drivenhypothesisRepeated donor delta patternDefine and implement host surface ports
A36A template is a governed composition recipe, not a copied appinferredB2B shelf and composition frameworkBuild one archetype from independent assets
A37Production learning will improve retrieval and reuse decisionshypothesisAutoSaaS learning loop is procedural, not measuredPersist pilot adaptation/outcome metrics and rerank
A38More breadth research currently has higher value than depthrejected for next loopRepeated breadth exposed same unresolved seamsFreeze broad loops until pilot evidence
A39Legal/rights should drive this quality architecture discussionout of scopeOperator explicitly delegated legal to othersRetain evidence separately; do not optimize architecture around it
A40One representative pilot can validate the entire platformrejectedIt can test contracts, not every industry/runtimeUse staged pilots across reuse shapes
A41Candidate evaluation may use a bounded synthetic composition scope without granting production admissionobserved in contract pilotThe corrected three-candidate solver passes the read-only evaluation projection while strict production R-SCOPE still refuses every NOT_ADMITTED RegistryRecord; fixture host checks add no donor or client-data evidenceQualify a real donor-bound composition only after explicit runtime authorization

Assumptions that should become explicit framework invariants

  1. One authoritative owner per stateful resource.
  2. One explicit reuse shape per packaged capability.
  3. Requirements remain implementation-independent until composition.
  4. Models cannot grant themselves data, tenant or side-effect authority.
  5. Unknown compatibility returns UNDERDETERMINED; it does not become model confidence.
  6. A whole workflow must pass before an application is accepted.
  7. Every release pins exact capability and binding versions.
  8. Every production composition has a recovery or replacement path.

Canonical source remains knowledge/02-ASSUMPTION-LEDGER.md. This HTML is a generated projection; edit the source, then run generate-docs.mjs.