Skip to content

ERRATA.md

Latest commit

 

History

History
126 lines (83 loc) · 13.9 KB

ERRATA.md

File metadata and controls

126 lines (83 loc) · 13.9 KB
Raw

ERRATA — Cross-Project Discrepancy Log

Purpose (a top priority of this effort): catalogue every contradiction — major, minor, or subtle/framing — that has crept in across the 24 projects (and across documents within a single project) as they were written and edited at different times over the past year. We surface these rather than silently resolving them: several are John's to decide.

  • Compiled: 2026-06-12, from the read-only harvest sweep. Each entry cites the conflicting sources.
  • Severity: 🔴 needs a John decision (load-bearing) · 🟠 should reconcile (naming/framing) · 🟡 minor/within-project drift.
  • How to use: when a single repo's self-description disagrees with this reference, check here first. Resolved items get struck through with the resolution noted.

🔴 E-01 — The corporate parent has three competing identities

The top of the corporate tree is named three different ways, and the newest version explicitly retires the oldest:

Source Top entity Subsidiary / product co. Regional consultancies
John's global CLAUDE.md + in-repo aixodev-aixocode/CLAUDE.md ExoDev.AI, Inc. "AIXO.Dev Platform LLC" (author field) "ExoDev.Pro - Dallas", "ExoDev.Pro - Los Angeles"
aixodev-aixocode/_REFERENCE/PRODUCT_AND_NAMING.md ExoDev.AI Corporation AIXO.Dev Platforms LLC (plural) "An ExoDev.AI Company" "ExoDevPro Dallas/LA, Inc." (dotless)
aixodev-web Phase-D research canon (Apr 2026, marked DRAFT/internal) ExoDev.Pro, Inc. (Dallas HQ) — explicitly retires "ExoDev.AI" AIXO.Dev Platforms LLC (plural) "ExoDev.Pro Dallas/LA LLC" (dotted)

Conflicts: ExoDev.AI, Inc. vs ExoDev.AI Corporation vs ExoDev.Pro, Inc.; "Platform" vs "Platforms" LLC; "ExoDevPro" vs "ExoDev.Pro"; consultancies as Inc. vs LLC. The Phase-D research is newest but is flagged aspirational/draft; John's CLAUDE.md is the live operating instruction but uses the oldest name. Decision needed: which corporate naming is canonical going forward? The VENTURES docs lead with the Phase-D model but flag it as unconfirmed pending this decision.

🔴 E-02 — The AIXO.Dev desktop edition: two names, two stacks, two framings

  • aixodev-professional's own README says it is "AIXO.Dev Professional" (aixodev-professional), a Rust/Tauri v2 + SvelteKit desktop edition, an explicit sub-project of aixodev-web, sharing a Rust/Tauri foundation with Divia.AI Professional.
  • aixodev-web's own docs call the desktop edition "AIXO.Dev Desktop" (aixodev-desktop), Electron-first (Tauri vs Electron an open ADR), and never use "professional."
  • The MetaProject _projects/README.md sided with the aixodev-professional / Rust-Tauri version.

Direct cross-repo contradiction on name, stack, and framing. Related: the "eventual Rust-server migration of aixodev-web" is asserted by aixodev-professional's README and the MetaProject index, but is not corroborated in aixodev-web's own docs (which mention only "Rust-leaning shared libraries" via FFI, not a server rewrite). Decision needed: is the desktop edition aixodev-professional (Rust/Tauri) or aixodev-desktop (Electron)?

🔴 E-03 — Divia.Life: native Swift/Kotlin vs Flutter-first

  • divia_ai-professional/BRANDING_and_PRODUCTS.md: Divia.Life = native Swift/Kotlin.
  • divia_ai-enterprise README and divialife-flutter (the actual product repo): Flutter first → native Swift/Kotlin editions later.

Resolution (high confidence): divialife-flutter is authoritative — Flutter is the first build; native Kotlin/Swift are later (SOMEDAY) editions once the data model stabilizes (the native dirs are confirmed empty). The "native Swift/Kotlin" claim is almost certainly a misquote of the table cell "Flutter (then native Swift/Kotlin)." Action: correct BRANDING_and_PRODUCTS.md when repos are next edited.

🔴 E-04 — divia_cards is a finished app, not an "early prototype"

The _projects/README.md index calls divia_cards an "early prototype — PLAN + PLAN-v2." It is actually a fully-built, working, tested application (built Nov 2025 by Factory.ai's "droid" CLI, Gemini-3-Pro-reviewed): working Flask backend, JWT auth, SocketIO, SvelteKit frontend, Svelte→Web Components compilation, 25 passing backend tests, and working Vue 3 + React 18 demo apps. Status correction needed. Compounding issue → E-05.

🔴 E-05 — "DiviaCard" means two unrelated things

The word DiviaCard is overloaded with no link between the two meanings:

  • In divia_cards (Nov 2025): a local UI widget — a Flask Card row of type outline | nlp_input | event, rendered as a framework-agnostic Web Component.
  • In the ecosystem (divia_ai-professional BRANDING + LegendaryMoney v2 vision, 2026): a cross-app data-interchange / registry type — e.g. DiviaCard::LegendaryMoney::transaction(...) published to a "global DiviaCards registry."

divia_cards predates the ecosystem framing, contains no registry/namespacing, and isn't referenced by any sibling repo's _REFERENCE/_EXTERNAL/. Best current read: divia_cards is the original rendering layer and name-source; the ecosystem later repurposed the term as a data concept; the two have never been reconciled or wired together. Decision needed: is divia_cards the rendering layer for the ecosystem DiviaCard, or a separately-named thing?

🔴 E-06 — Three "standalone" projects deny an ecosystem they belong to

The Divia.Network membership is asserted one-directionally:

  • tastypantry CLAUDE.md: "standalone … not part of any larger platform … no shared dependencies." Yet it carries _REFERENCE/_EXTERNAL/ symlinks to sattvasichealth, legendarymoney-web, and diviahome-web, and its NL food-log model is the canonical Divia.Network "I ate two quesadillas" capture. The baseline's lineage (tastypantry = seed/ancestor) and the food→SattvasicHealth macro feed are not stated in-repo.
  • sattvasichealth README/CLAUDE.md: "a standalone project … not part of any larger platform," integrating only with TastyPantry. LegendaryMoney's docs name Sattvasic Health as an ecosystem sibling; Sattvasic Health does not reciprocate.
  • spicemaster3000: fully standalone — zero Divia/TastyPal references anywhere; its only tie is the AIXO.Dev workflow methodology and John personally. (This one is correctly standalone; flagged so the VENTURES/PRODUCTS docs don't over-claim it as a Divia.Network app.)

Decision needed: are TastyPantry and Sattvasic Health in the Divia.Network ecosystem (update their docs) or genuinely standalone (update the index/this reference)? The evidence says they're meant to be in it; their docs lag.

🟠 E-07 — Brand-spelling drift: dots, underscores, and casing

A pervasive, low-stakes-but-everywhere inconsistency:

  • Product brand vs repo name: "Divia.AI" / "Divia.Network" / "Divia.Life" / "Divia.Foundation" (dot, capital) in prose, vs repo/dir names divia_ai-*, umbrella "DiviaLife", divialife-* (underscore/no-dot).
  • .dvai format has three expansions in active use: "DVAI Document Format" / "Divia Document Format (DDF)" / "DVAI format" — sometimes within a single repo.
  • KingStratVC: firm "Kingmaker Strategic"; product styled "KingStrat.vc" in-repo vs "KingStratVC" (GitHub org / env prefix) vs the LATER-002 product name "KingStratVC Knowledgebase."
  • iOS dir casing: divialife-iOS / diviacontacts-iOS (capital-OS) vs docs' divialife-ios (lowercase).

Action: pick canonical spellings in a naming appendix; not blocking, but it's the kind of skew this reference exists to end.

🟠 E-08 — The Enterprise ↔ Swarm relationship is asserted from one side only

  • divia_ai-agentswarms frames Divia.AI Enterprise as "a PKMS + Asana-style task/project server" that co-deploys with Swarm so Swarm can power Enterprise's AI features.
  • divia_ai-enterprise's own README frames itself as a collaboration/sync server and never mentions Swarm or an "AI backbone" at all.
  • divia_ai-professional's BRANDING frames Enterprise as the full collaboration server whose desktop client is Pro (no Asana framing, no Swarm).

Three partly-overlapping pictures of the same server; the Swarm co-deployment exists only in Swarm's docs. Action: write one canonical Enterprise definition (the VENTURES/PRODUCTS docs attempt this) and reconcile.

🟠 E-09 — The bootstrap-lineage "root" is wrong (and self-reports disagree)

  • The _projects/README.md index says aixodev-collabs is the root of the shared _workflows/ system. The git history says aixodev-collabs is a middle link: the chain is aixodev-codemap → aixodev-projects → aixodev-collabs → aixodev-workgroups, and the _workflows/ corpus itself predates all four (it traces back toward aixodev-web / the original workflow set).
  • Self-reported chains disagree across repos: LegendaryMoney's docs say DiviaHome → Sattvasic → TastyPantry → aixodev-collabs; Sattvasic Health says it came directly from TastyPantry; DiviaHome says it came from Sattvasic Health. The "who-cloned-from-whom" story is inconsistent.

Action: record the actual lineage once (a STATUS/VENTURES appendix) and stop trusting per-repo self-reports.

🟠 E-10 — Model-discipline deviations vs John's "always Opus, never Sonnet" rule

  • aixodev-openhands/CLAUDE.md pins Opus 4.6, warns that model:"opus" resolves to Opus 4.7, and explicitly permits Sonnet subagents for "simple mechanical tasks." This contradicts John's standing global rule (always Opus; never Sonnet) — flagged 2026-06-12 when John re-affirmed the rule. Action: bring this repo's model policy into line.
  • diviacontacts-gmail's entire ~69k-word research corpus was produced by Claude Fable 5 (per John's explicit 2026-06-11 directive, doubling as a Fable-class capability test) — not Opus. Not a violation (John directed it), but recorded so the provenance is clear.

🟠 E-11 — Capability documented ahead of reality

Several repos describe features as if present that depend on unbuilt work:

  • aixocode AgentEngine mode is a core concept in the entity/PROLOGUE docs but is not in the implemented feature list — it depends entirely on the unbuilt server integration (Sprint 05).
  • divia_ai-professional ADR-002 (WYSIWYG/Typora) is cited as the live decision in CLAUDE.md/AGENTS.md, while a withdrawn copy of the same ADR sits in _REFERENCE/ARCHITECTURE/ and an active TipTap→ProseMirror migration is already questioning the "settled" editor.
  • Several Phase-00-pending repos (legendarymoney-web especially) assert "extensive deep research already gathered," but their _research/ directories contain only .gitkeep — the research is claimed, not present in-repo.

🟡 E-12 — LATER-002 forward-decisions not yet propagated to the repos (expected)

Canonical decisions captured in the MetaProject's _backlog_TODOs/LATER-002 have no footprint in the product repos yet — expected, since LATER-002 is the capture point, but listed so they get pushed down later:

  • .dvai "LiveDocuments" + Divia.AI Enterprise "Research Projects" — absent from divia_ai-professional/-enterprise/-swarm.
  • Typed Swarm workflow steps (deterministic-vs-probabilistic + Haiku→Fable tiers) — absent from divia_ai-agentswarms.
  • DiviaHome nightly consumption-driven replenishment + kitchen-counter voice device — absent from diviahome-web (it lists only a future "DiviaHome devices" repo; no nightly review, grocery, or replenishment logic).
  • "KingStratVC Knowledgebase" product name — absent from kingstratvc-web (in-repo it's "KingStrat.vc").

🟡 E-13 — Heritage & symlink integrity

  • diviacontacts-gmail's README/CLAUDE.md claim it was "carried over with full git history from divia-gmail." The git log starts at "Initial commit: project placeholder" and is self-contained; no divia-gmail directory exists anywhere. Looks renamed-in-place / freshly initialized, not history-grafted. (The _projects/README.md repeats the "full history" claim.)
  • Broken reference symlinks: diviacontacts-gmail/_REFERENCE/_EXTERNAL/diviahome-web → …/DiviaAI/diviahome-web (does not exist; the real one is …/DiviaHome/diviahome-web). aixodev-web's external symlinks point to /Users/jstanforth/… (macOS paths, dangling on this Linux host).

🟡 E-14 — Within-project numeric/label drift (non-blocking)

Captured for completeness; none load-bearing:

  • aixocode test counts: ROADMAP/in-repo CLAUDE.md say 1732/1682/50; the global CLAUDE.md says 1714/1693/21; the Users Guide cites "1241+"; AGENTS.md is staler still.
  • aixocode Users Guide says Codex/Gemini parsers are "Phase 7" while the ROADMAP places them in Phase 8 (Phase 7 was Workflow Orchestration, already closed).
  • aixodev-web table count: "25 tables" (CLAUDE.md) vs "27" (status report) vs "~47/~75" (entity vision).
  • divia_ai-professional test counts: 37 vs 39 Rust across docs; repo-structure blocks still say divia_ai-professional-codex/.
  • spicemaster3000 data counts: 23,690 pairings / 595 subjects (authoritative, verified on disk) vs a report-card table's 23,378 / 584.
  • divia_ai-professional copyright runs "© 1996–2026" while the brand history says "Divia.AI (2020–present)."

Cross-cutting pattern

Almost every 🔴/🟠 item is the same root cause: the ecosystem-level story was written into many repos at different times, and the repos drifted — newer strategic decisions (the FDSE pivot, the corporate rename, the DiviaContacts repositioning, the LATER-002 agent decisions) landed in some docs and not others. That is exactly the skew this _REFERENCE/ is built to absorb: once each repo points back here and sheds its non-technical vision text, there's one place to keep correct.