9267506a42
Opens the holistic macro→micro assessment of CORE's cognitive-cycle design and implementation fulfillment. Read-only; no runtime behavior changes. Method is governed by docs/conceptualizing_engineering_mastery.md (committed here because the assessment cites it as its method): Semantic Rigor, Mechanical Sympathy, Third Door; delete before optimizing; automate last. Phase 0 findings: - 0-A CORE carries FIVE unreconciled macro articulations of its own cognitive cycle (AGENTS.md north star; AGENTS.md live path; the 2026-05-12 draft three-physics-layer blueprint; ratified ADR-0252's five paradigm stages; the system map's 10-band L0-L11 layering). The blueprint's allocation/attention layer (Salience/Attention/Inhibition) exists in no other articulation. - 0-B .system-map/ is the best existing macro→micro artifact and is 48 days stale, gitignored/local-only, and carries four zero-subsystem zones — three of which sit on the serving path. - 0-C ADR-0252 §5 — the experiment gating the governing paradigm's build — has never returned a verdict (two unmerged rnd/ worktrees). The 34 surface organs the ADR diagnoses as the error keep serving. L10 blocks all 7 OPEN W-* items; field + T1 vault are discarded on exit, so "one continuous life" is today many short lives sharing a checkpoint. - 0-D No live gap register: docs/gaps.md is 26/26 closed, the liveness ratchet is L10-blocked and stale, docs/analysis/ is a chronological archive. - 0-E 333 ADRs, flat numbering, status not reliably machine-extractable; INDEX-by-domain is 21 files behind; the Blueprint reserved 0254-0261 for intents never materialised while those numbers were minted for other work. - 0-F Measured: CGA is ~73% of turn time via cga_inner/geometric_product (~34k calls/turn), NOT versor_condition (0.22%). Rust backend off by default, parity question open and blocked. MLX in no runtime path. - 0-G The 2026-07-25 verification precedent falsified ~1/3 of an external blueprint's items by reading code — the bar for every downstream phase. Phase 1 receives the system map's 17-field proto schema and 7-value liveness vocabulary, plus the four dimensions it lacks (evidence pointer, capacity, design-vs-built, fitness judgment) and five decisions to make explicitly. PR #138's fabrication findings are recorded as measured-and-pinned, held for ADR + ratification — never re-discovered, never fixed here.
78 lines
6.4 KiB
Markdown
78 lines
6.4 KiB
Markdown
# CORE Holistic Assessment — Scope and Method
|
|
|
|
**Status:** Ratified approach (Shay, 2026-07-27). Phase 0 complete.
|
|
**Branch:** `docs/holistic-assessment` (worktree `core-wt-assess`, based on `forgejo/main` @ `8927c563`).
|
|
**Nature:** Read-only investigation. This assessment changes no runtime behavior, fixes no defect, and decides nothing. It produces evidence and judgments for ruling.
|
|
|
|
---
|
|
|
|
## 1. The question being answered
|
|
|
|
Four questions, in order of dependency:
|
|
|
|
1. **Where does CORE actually stand** on its cognitive cycle — design articulated versus implementation fulfilled — from the telos down to individual components?
|
|
2. **Is the layer model itself complete?** Are there layers, sublayers, or components missing from the *design*, not merely from the implementation — things an AGI/ASI-grade system requires that nothing in the current architecture accounts for?
|
|
3. **What is the metadata for every layer, sublayer, and component?** Philosophical intent, functional contract, design shape, implementation status, evidence, capacity, and role — such that any future dive begins with a clear target rather than a grep.
|
|
4. **What is hindering us?** Implemented ADRs or designs that are the wrong solution for their underlying problem; responsibilities lodged in the wrong subsystem; trade-offs tuned rather than dissolved.
|
|
|
|
Question 4 is not a courtesy pass. It is the question with the highest expected value, because a wrong component that works is more expensive than a missing component that is known to be missing.
|
|
|
|
---
|
|
|
|
## 2. Governing method
|
|
|
|
The assessment is conducted under `docs/conceptualizing_engineering_mastery.md`, applied to the assessment itself and not only to its subject.
|
|
|
|
**Pillar I — Semantic Rigor.** Completeness criteria and the metadata schema are defined *before* any component is judged (Phase 1), so that "missing" and "complete" have fixed meanings rather than per-component ones. Every `implemented` verdict requires an evidence pointer: a test, an eval lane, a pinned SHA, or an acceptance packet. "The module exists" is not evidence that it executes; "the flag is threaded" is not evidence that it changes behavior.
|
|
|
|
**Pillar II — Mechanical Sympathy.** Components are judged against CORE's own doctrine — deterministic decoding, exact recall, field-as-substrate with intelligence in the wiring, replay-gated learning, `wrong=0`-or-refuse — and not against a generic AGI checklist. A capability only counts as a gap if CORE's own telos requires it. Importing an external architecture's expectations would manufacture false gaps.
|
|
|
|
**Pillar III — The Third Door.** The hindrance audit's explicit charter: find the places where a bad trade-off was split rather than dissolved, and the places where deletion beats addition. Per the execution algorithm, deletion (step 2) precedes optimization (step 3); a component that should not exist is never a performance problem.
|
|
|
|
**Two standing discipline rules carried in from prior work:**
|
|
|
|
- **The sabotage test.** For every claim that a mechanism is live and load-bearing, ask what the measurement would look like with the mechanism removed. If it would look identical, the claim is decoration and is recorded as such. This repository has produced exactly that failure before — a rate reported as evidence of a reader that was `0.0` throughout.
|
|
- **Identity, not value.** When measuring whether two things are the same thing, measure identity rather than equal-looking values. Source-scanning metrics can move the wrong way on success.
|
|
|
|
---
|
|
|
|
## 3. Phases and division of labor
|
|
|
|
| Phase | Deliverable | Executor |
|
|
|---|---|---|
|
|
| **0 — Ground truth** | Canonical-document ingestion; ADR triage; system-map recovery; raw material and open tensions for the taxonomy | Opus 5 — **complete** |
|
|
| **1 — Taxonomy & schema** | The macro→micro layer taxonomy and the metadata card schema every card must fill | Fable 5 |
|
|
| **2 — Macro layer cards** | One card per top-level layer; layer-level verdicts; cross-cutting concerns | Opus 5 |
|
|
| **3 — Micro component cards** | Per-subsystem descent, depth allocated by load-bearing-ness | Fable 5 |
|
|
| **4 — Gap register + hindrance audit** | Two separate registers; evidence-carrying | Opus 5 |
|
|
| **5 — Synthesis** | Executive assessment; ranked gaps; ranked hindrances; recommended R&D attack order | Opus 5 |
|
|
|
|
Phase 1 is the keystone. A wrong taxonomy miscategorizes everything downstream, and it is the cheapest phase to correct.
|
|
|
|
---
|
|
|
|
## 4. Deliverables
|
|
|
|
```
|
|
docs/assessment/
|
|
00-scope-and-method.md # this file
|
|
01-phase0-ground-truth.md # Phase 0 findings + Phase 1 handoff
|
|
02-layer-taxonomy.md # Phase 1
|
|
03-card-schema.md # Phase 1
|
|
10-layer-cards/ # Phase 2
|
|
20-component-cards/ # Phase 3
|
|
30-gap-register.md # Phase 4
|
|
31-hindrance-audit.md # Phase 4
|
|
40-assessment.md # Phase 5
|
|
```
|
|
|
|
---
|
|
|
|
## 5. Rules of engagement
|
|
|
|
1. **Read-only.** No runtime code is modified. No defect is fixed. Where a fix is obvious, it is recorded as a finding with its evidence, not applied.
|
|
2. **Verify against code, not against documents.** Documentation in this repository has been measurably wrong about the repository before — most recently `docs/research/architecture-assessment-verification-2026-07-25.md` falsified roughly a third of an external blueprint's work items by reading the implicated code. A claim sourced only from a document is labeled as such.
|
|
3. **Settled rulings are constraints, not subjects.** The deduction pivot, the scripture-content deferral, the no-merge-automation rule, and ratified ADRs enter as given. An ADR is reopened only on evidence of hindrance, and only as a flag for ruling — never as a unilateral recommendation to reverse.
|
|
4. **Known-and-held findings are recorded as such.** The PR #138 fabrication findings (`every dog is a mammal` → `member(every_dog, mammal)`; `Given: furthermore; p implies q; p.` reaching served output) are **measured and pinned, not fixed**. The fixes are known — two of the 13 mutations — and are deliberately held out pending ADR and ratification because they change what CORE comprehends from user input, which is serving-path truth behavior. They enter the gap register pre-labeled and are never re-presented as newly discovered.
|
|
5. **No timelines.** Scope size, phase, and priority only.
|
|
6. **All wrinkles surfaced.** Technical truths are volunteered unprompted, including ones that complicate the picture or reflect badly on prior work.
|