docs: D-SAG-5 reflexive software cognition after #1509, OpenCog/NARS research and probes

[AdaWorldAPI]

Scope and lineage

Based on merged #1509, building from the shipped one-concept, one-hop SoftwareGraph rather than proposing another introspection engine. Consolidates the primary questions and scientific limits of merged #1496, #1497, #1498 and #1499.

Documentation only. No Rust changes, no semantics/ABI changes, no new opcode, IR, engine or registry, no optimizer execution and no implementation permission.

Artifacts

• D-SAG-2 plan: existing D-SAG-1: software-awareness graph — lance-graph's own methods, programs and evidence as Quack-queried rows #1509 capabilities versus gaps, typed self-model contract, proposed ContextSlice, source/claim/attention/receipt separation, four functional appraisal policy probes and D-SAG-2-0..10 falsification ladder.
• Claude Code handoff prompt: source-first reading order, independently pinned oracles, 11 discriminating fixtures, proof/measurement refusal and structured output contract.
• NARS/ONA/Hyperon primary extracts: Wang et al. 2018 self/mental operations, ONA event control, Hyperon self-understanding, rule-selection and Psi emotion modulators, with explicit non-claims about consciousness.
• Shannon/PLN/folds/optimizer extracts: Houlsby BALD Eq.2, Geisweiller/Goertzel pattern surprisingness, typed metagraph constructor laws, static snapshot and associativity assumptions, Jøsang evidence fission, Willsey egg rewrite-equivalence boundaries.
• Pinned sources and oracle/refusal matrix: repo SHAs, 12 negative-control areas, questions and licensing caveats.
• Board entry, generated index and status/integration plan updates in the same commit.

Questions that matter

1. Can D-SAG-1: software-awareness graph — lance-graph's own methods, programs and evidence as Quack-queried rows #1509's prepared methods and program receipts answer local software queries while reducing measured LLM context/token cost? Do not confuse a local graph query with universally zero-token model access.
2. Can we distinguish an observation, its evidence source, an inference receipt, an attention visit, and a later relational reinterpretation without manufacturing additional confidence?
3. Does an ONA/PLN-inspired priority queue, Shannon, valid BALD EIG, Miner/JSD or Moore/Staunen policy find more useful source-backed problems at the same rule set, evidence and compute budget?
4. Can user-coined Emberglow, Steelwind, Woodwarm, Velvetpause be operationalized as optional, ablatable priority policies instead of anthropomorphic qualia claims or CE64 bit reassignment?
5. Where do MeTTa branching/typed metagraph folds require genuine additional semantics beyond a one-hop Quack query? A 4-answer nondeterministic pattern must not be collapsed to 2 row survivors.
6. Can H1 and future candidates be proposed from source-backed DAG/MDL evidence yet remain unsafe to apply until exact compiler oracle, full benchmark and separate human authorization?

Baseline limitations preserved

• D-SAG-1: software-awareness graph — lance-graph's own methods, programs and evidence as Quack-queried rows #1509's H1 shows 1.6–1.8x current tiled-over-scalar cost, not a fused SIMD win.
• D-SAG-1-4 multi-hop, D-SAG-1-6 persistence/strong source digest, and runtime self-application are not yet implemented.
• Hypothesis layer != evidence source; confidence != priority; reappraisal != new fact; operator permission != proof.
• Hyperon/MeTTa, classic AtomSpace, PLN, ECAN and ONA are distinct reference systems.
• The CE64 little-endian ALU/ISA contract and hotplug/ClassView semantic readings remain untouched.

Reviews and gates

• Exactly nine documentation files in PR; same-commit board hygiene.
• Distinguishes source-derived claims, candidate mappings and future measurements.
• Claude Code prompt contains negative controls, source pins, evidence standards and STOP conditions.
• No new behavior tests/benchmarks/oracles run under this PR.
• D-SAG-2-1..10 require later, independent approval and measurable test PRs.
• CI board/checks pending after PR creation.

What would falsify the design first?

A stale compiler fact returned as current, a copied event increasing belief confidence, a two-hop query pretending to be solved by one-hop masks, a supposed EIG result with no predictive distribution, an irrelevant high-entropy distractor winning every budget, a MeTTa multiple-result collapse, or a code change applied without authorization.

This PR should be reviewed as a research protocol, not as evidence that an autonomous or conscious software architecture has shipped.

[coderabbitai]

Warning

Review limit reached

Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab.

Next included review available in 31 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available. Your 62 included PR review attempts over the past 7 days set your current allowance at 1 review per hour.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

• Configuration used: Organization UI
• Review profile: CHILL
• Plan: Essentials
• Run ID: a0491587-1482-4bff-9b57-08e376e429dd

---

📥 Commits

Reviewing files that changed from the base of the PR and between 42fa2a3 and f81e6de.

---

📒 Files selected for processing (9)

• .arxiv/2026-10_sag1509_information_metagraph_optimization_excerpts.md
• .arxiv/2026-10_sag1509_nars_hyperon_reflexive_appraisal_excerpts.md
• .arxiv/2026-10_sag1509_source_oracles_falsifier_matrix.md
• .claude/board/INTEGRATION_PLANS.md
• .claude/board/STATUS_BOARD.md
• .claude/board/entries/2026-10-12-sag-reflexive-cognition-research.md
• .claude/board/entries/README.md
• .claude/plans/sag-reflexive-cognition-from-1509-v1.md
• .claude/prompts/sag-reflexive-cognition-from-1509.md

---

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

---

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0383e01406

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Use crate-qualified paths for the awareness fixtures

When this handoff prompt is followed from the repository root as instructed, tests/awareness.rs and examples/awareness_loop.rs do not exist, so the required source-first reading order fails or omits the primary test and benchmark evidence. Both files are under crates/lance-graph-glove-parity/; qualify those paths just as the preceding src/awareness.rs path is qualified.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-12T19:32:31.497083Z	0383e01	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[AdaWorldAPI]

Superseded by #1523. Every commit from this branch is merged there unchanged except for the alignment edits.

Changes for this PR:

• "H1 proposal" and "optimization proposal" now map to the canonical ProposedMethod.
• The closeout placeholder is filled in.
• Your "try an invalid two-hop query" probe needs a different refusal now. out(r1).out(r2) lowers since Chained hops folded: mask-risc GatherChain, Quack Filter::Chain, glove out(r1).out(r2) (D-CMH-G1-7) #1522; in/both still refuse as FanOut.

Closing in favour of the capstone.

---

Generated by Claude Code