experiments: thesis-G interdependency design note (non-canonical spike) - #64
Open
IanMayo wants to merge 4 commits into
Open
experiments: thesis-G interdependency design note (non-canonical spike)#64IanMayo wants to merge 4 commits into
IanMayo wants to merge 4 commits into
Conversation
Standalone sandbox note probing concept §6.3 — does thesis G (PMESII interdependency) admit an honest v1 slice, or is it horizon-only? Provisional verdict: interdependency is honestly computable as attributed reachability / path-surfacing (thesis F generalised to an ontic graph), but not as weighted effect prediction (the EBO/ONA false-precision failure). The boundary is magnitude — the same line the core already holds via DEC-35 and tests/k3-trace.test.ts. Non-polluting by construction: experiments/ tree stamped non-canonical; no schema change, no src/ import, no register assertion. Any finding feeds back only as a flagged concept §6 candidate (DEC-2). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KTBujZMsBhnq1gpsj4W17G
|
Adds rung 2 (signed direction) as a first-class tier between reachability and weighted magnitude, and works the proof: per-link signed direction is honest (ASSAY is already signed-without-magnitude at its core — directional commitments, the O-3 signed-margin verdict), but sign COMPOSITION splits — same-direction chains survive only with per-hop owned monotonicity, and net sign over opposing parallel paths is undefined without weighing (DEC-19), handled by refusing the net via the O-3 straddle discipline lifted to the graph. Sharpens the verdict: the honest boundary is not "topology vs magnitude" but "per-link attributed facts vs any aggregation that weighs paths against each other" — one principle covering rung 2b (net-sign) and rung 3 (magnitude). Meridian exhibit: threat_exposure(FE-ANVIL) opposing-path straddle. Demonstrator becomes three reads with a live straddle detector. Grounds the composition failure in system-dynamics causal-loop-diagram loop-polarity limits (Sterman). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KTBujZMsBhnq1gpsj4W17G
Self-contained offline HTML (no bundler, no network, no src/ import) rendering one Meridian-derived PMESII graph three ways over a shared node-link map: - Reachability (rung 1): forward transitive walk, provenance per edge, honest dead ends — no fabricated Social→Political cascade. - Signed direction (rung 2): asserted ↑/↓ hops with owners; the centrepiece is the live straddle detector — "what drives FE-ANVIL exposure?" opposes (FALCON suppression ▼ vs ANVIL advance ▲) and the net is refused (the O-3 discipline lifted to the graph), both paths shown. - Weighted propagation (rung 3): the labelled EBO/ONA trap — fabricated weights and indices under a red FALSE PRECISION banner, per-number annotation of why each lies (DEC-19 / JFCOM-2008 / DEC-35). Numbers generated at render time, never stored on the graph. In-page honesty self-check (all pass): no stored scalar magnitude (DEC-19), every edge attributed (G3), straddle detected + net refused (rung 2b), honest reachable dead ends. Verified end-to-end in Chromium across all tabs/questions, zero console errors; preview-*.png are the verification shots. Note §8 + README updated to point at the built file. No core contamination. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KTBujZMsBhnq1gpsj4W17G
… area
build:site now copies experiments/thesis-g/demonstrator.html into
site/experiments/thesis-g/ and writes a clearly-labelled site/experiments/
landing ("experimental · not canonical", "Not part of ASSAY canon").
Deliberately NOT linked from the Home page — an accepted exception to the
copy-and-link convention, following the dep-graph-mocks (review-only)
precedent. Featuring a horizon-thesis spike on the public face would be the
false-precision pollution the sandbox exists to avoid; this keeps it reachable
in the PR preview at /experiments/ while staying out of the canonical nav.
Verified from the built site/ in Chromium: landing stamp + card link correct,
demonstrator loads, honesty self-check all-pass, straddle refusal fires, zero
console errors. site/ stays git-ignored (generated, CI-published).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KTBujZMsBhnq1gpsj4W17G
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
A standalone, non-canonical design note probing concept §6.3 — does thesis G (PMESII interdependency) admit any honest v1 slice, or is it horizon-only? It reasons through what a standalone demonstrator should learn before any code, and records a provisional (non-asserting) verdict on where the honest line falls. Docs only; no code, no schema, no
src/touched.Slices
Not a SPEC. This is a sandbox spike, deliberately kept out of
docs/research/(it gates no build stage and decides nothing canonical). It probes the concept §6.3 open question and is adjacent todocs/research/horizon-forward-derivation.md.experiments/thesis-g/thesis-g-interdependency.md— the design noteexperiments/README.md— stamps the wholeexperiments/tree non-canonicalRegister impact
No new register decision. By construction:
npm run gen, no LinkML change (assay-knowledge-model.md:484keeps PMESII shapes out of core — this note respects that).importintosrc/; fixture reads would be one-way and read-only.Provisional verdict (the substance)
Three findings worth a reviewer's eye:
assay-doctrine.html:101). The demonstrator surfaces what's implied; it invents nothing.tests/k3-trace.test.ts, DEC-35 civil-harm-stays-geometric). A weighted panel that made "55k vs 5k" move a verdict would do exactly what the core already declined. Thesis G's honest line is the same line, stated for a whole graph.How to review
1. Quick check
typecheck/testunaffected (docs-only; nothing imports the note)2. Read the note
experiments/thesis-g/thesis-g-interdependency.md— §1 (the EBO/ONA false-precision framing), §2 (the latent Meridian PMESII graph + the epistemic-vs-ontic distinction), §3 (the honest reachability read), §4 (the labelled trap), §5 (verdict), §6 (honesty invariants), §7 (what the demonstrator would show).experiments/README.md— the non-pollution rules.3. Blog article
N/A — no blog article for a sandbox spike (comms cadence is for canonical slices, DEC-30).
Evidence
N/A — no runtime surface; nothing to screenshot. The note is the deliverable.
Honesty checklist
Test plan
typecheck/testscope unchanged; theexperiments/tree is not imported bysrc/🤖 Generated with Claude Code
https://claude.ai/code/session_01KTBujZMsBhnq1gpsj4W17G
Generated by Claude Code