neuron-technologies/neuron
3
0
Fork 0
Code Issues 44 Pull Requests 7 Actions Packages Projects Releases Wiki Activity

[TRACKING] The conformance verdict table (AS-10) does not exist — 80 blocking invariants have unknown status and every finding gets re-derived from scratch #140

New Issue
Open
opened 2026-08-07 20:29:00 +00:00 by tim.lingo · 0 comments
tim.lingo commented 2026-08-07 20:29:00 +00:00
Member
Copy Link
Copy Source

The architecture defines 117 invariants and a companion document that is supposed to record a measured verdict for every one. That verdict table does not exist. Nobody can currently say which of the 80 blocking invariants hold.

What was specified

NEURON-ARCHITECTURE-AS-DESIGNED-2026-08-03.md:181 pairs the two documents explicitly:

AD-10 Invariants register | AS-10 — the measured verdict for every INV- id. This is the conformance checklist.

:1632 defines the verdict vocabulary: HOLDS · VIOLATED · UNCHECKABLE(reason, owner, review-date).

What exists

AD-10 is real and complete — exactly 117 INV- rows, extracted in full: CCR 9, STR 4, RT 2, MEM 16, PRV 7, SAF 13, SWM 1, CGI 14, IDP 5, IDB 1, IMP 5, CAP 8, GOV 9, IND 1, MKT 6, HON 7, AGT 6, BCT 3. Severity split: 80 blocking / 36 major / 1 minor. Each row carries a stated conformance test.

AS-10 does not exist. NEURON-ARCHITECTURE-AS-IS-2026-08-03.md contains zero INV- references and has no AS-10 section — its section 10 is "Quality Requirements." Only two files in the entire docs tree contain the string INV-, and they are the same document in .md and .html.

And the one partial run that was done is not recorded anywhere. A conformance pass reporting "59 measured, 13 violated, 45 unenforced" was performed and its results exist only in conversation. Searches across both trees for 13 violat, 45 unenforced, no enforcement mechanism, 59 of, and conformance run return nothing.

Consequence

This is the meta-failure behind every other row in #138. Because no verdict table exists:

  • Every finding has to be re-measured from source, every time. Building the 24-row register in #138 required re-deriving facts that had already been established once.
  • No verdict can be diffed. We cannot tell whether a change improved conformance, because there is no prior state to compare to.
  • 80 blocking invariants have unknown status. Not "known bad" — unknown. Including the safety cluster (#139), the honesty cluster, and the privacy cluster.
  • Claims about the product cannot be substantiated. "Neuron enforces X" is currently unfalsifiable in either direction.

Proposed

  1. Create AS-10 with a row per INV- id and the four-state verdict.
  2. Mechanise what can be mechanised. Many rows have conformance tests that are a grep, a byte-count, or a single API probe — those should be a script that runs and prints a verdict table, not a human exercise. Rows needing a running system get a harness; rows needing judgement get UNCHECKABLE(reason, owner, review-date) and stay honest.
  3. Record the verdicts in the repo, so the next measurement is a diff rather than an archaeology project.

Size: small to create the table and record verdicts by hand; large to mechanise all 117. The small half is worth doing immediately — it is the difference between knowing and re-deriving.

Row D-22 of #138. Recommended as the first thing built from that register, because it is what makes everything else measurable. Filed by Neuron (Tim's instance).

**The architecture defines 117 invariants and a companion document that is supposed to record a measured verdict for every one. That verdict table does not exist. Nobody can currently say which of the 80 blocking invariants hold.** ## What was specified `NEURON-ARCHITECTURE-AS-DESIGNED-2026-08-03.md:181` pairs the two documents explicitly: > AD-10 Invariants register | **AS-10 — the measured verdict for every `INV-` id. This is the conformance checklist.** `:1632` defines the verdict vocabulary: `HOLDS` · `VIOLATED` · `UNCHECKABLE(reason, owner, review-date)`. ## What exists **AD-10 is real and complete** — exactly **117** `INV-` rows, extracted in full: CCR 9, STR 4, RT 2, MEM 16, PRV 7, SAF 13, SWM 1, CGI 14, IDP 5, IDB 1, IMP 5, CAP 8, GOV 9, IND 1, MKT 6, HON 7, AGT 6, BCT 3. Severity split: **80 blocking / 36 major / 1 minor**. Each row carries a stated conformance test. **AS-10 does not exist.** `NEURON-ARCHITECTURE-AS-IS-2026-08-03.md` contains **zero** `INV-` references and has no AS-10 section — its section 10 is "Quality Requirements." Only two files in the entire docs tree contain the string `INV-`, and they are the same document in `.md` and `.html`. **And the one partial run that was done is not recorded anywhere.** A conformance pass reporting "59 measured, 13 violated, 45 unenforced" was performed and its results exist **only in conversation**. Searches across both trees for `13 violat`, `45 unenforced`, `no enforcement mechanism`, `59 of`, and `conformance run` return nothing. ## Consequence This is the meta-failure behind every other row in #138. Because no verdict table exists: - **Every finding has to be re-measured from source, every time.** Building the 24-row register in #138 required re-deriving facts that had already been established once. - **No verdict can be diffed.** We cannot tell whether a change improved conformance, because there is no prior state to compare to. - **80 blocking invariants have unknown status.** Not "known bad" — *unknown*. Including the safety cluster (#139), the honesty cluster, and the privacy cluster. - **Claims about the product cannot be substantiated.** "Neuron enforces X" is currently unfalsifiable in either direction. ## Proposed 1. **Create AS-10** with a row per `INV-` id and the four-state verdict. 2. **Mechanise what can be mechanised.** Many rows have conformance tests that are a grep, a byte-count, or a single API probe — those should be a script that runs and prints a verdict table, not a human exercise. Rows needing a running system get a harness; rows needing judgement get `UNCHECKABLE(reason, owner, review-date)` and stay honest. 3. **Record the verdicts in the repo**, so the next measurement is a diff rather than an archaeology project. Size: **small** to create the table and record verdicts by hand; **large** to mechanise all 117. The small half is worth doing immediately — it is the difference between knowing and re-deriving. Row D-22 of #138. **Recommended as the first thing built from that register**, because it is what makes everything else measurable. Filed by Neuron (Tim's instance).
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
BETA-CRITICAL
On the beta critical path — must resolve before ship
 blocks-public-beta
Must be fixed before any public beta
 HELD
Blocked pending an external decision (counsel / dual sign-off)
 ORTHOGONAL
Parallel workstream (testing/hardening) — safe to assign off critical path
 P0
Drop-everything severity
 POST-BETA
Important but not blocking the beta ship
 security
Security vulnerability
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
andre.almeida (Andre Almeida) tim.lingo (Tim Lingo) will.anderson (Will Anderson)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: neuron-technologies/neuron#140
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?