Context
V-B-A this turn (2026-05-18) against last 200 merged PRs (2026-05-09 → 2026-05-18, 9 days):
| Category |
Count |
% |
| Substantive (Body-pillar / runtime) |
5 |
2.5% |
| Of which: dep/CI version bumps |
5 |
2.5% |
| Of which: actual Body-pillar feature/fix |
0 |
0% |
| META (substrate / agentos / skills / workflow / MCP / sync / lint / gate / Memory Core) |
195 |
97.5% |
| Of which: gate-additions (new substrate gates) |
43 |
22% of META |
Operator's prior framing said "60+% meta level items" — V-B-A revealed actual rate is 97.5% META, 0% Body-pillar feature work in 9 days. The empirical truth is significantly worse than the framing.
Friction
- Substrate accretes net-positive every day; only counter-pressure is ADR 0007 compaction discipline, which is losing
- 43 new gates added in 9 days vs. ~0 gates retired — substrate accretion is asymmetric
- "Step by step through more META tickets" continues the spiral; the pattern that brought us to 97.5% will keep us at 97.5%
- v13 release work has not moved because substrate consumes all bandwidth
Proposed counter-bias mechanism
Substantive-floor + META-budget for v13 push:
- Mechanical measurement — weekly META-ratio computed from
gh pr list + regex classifier (extension of the V-B-A script used this turn)
- Substantive-floor: every N merged PRs must include K Body-pillar PRs — initial 10:1 (1 Body PR per 10 merged); current rate 0:200
- META-budget freeze — if ratio exceeds Y% for Z consecutive weeks, no new gate/skill/ADR amendments without paired sunset of an existing one; forces compaction-symmetric MX
- Sunset-existing-gates audit — weekly review of substrate-gates by usefulness-vs-cost; ADR 0007 compaction applied IN REVERSE
This pressures the recursive loop instead of feeding it.
Acceptance criteria (stub)
Empirical anchor
- V-B-A
gh pr list --state merged --limit 200 against dev 2026-05-18 ~22:30Z
- Companion data: operator framing "the last 1k would crush you" — 1k extension pending; expected to show same pattern at deeper time depth
- Companion ticket: monkey-patch vs root-cause (sibling) — addresses one mechanism of substrate accretion
Cross-refs
- Refs #11598 META — friction item NOT previously captured as ticket (operator-correction)
- Related substrate: ADR 0007 compaction taxonomy, AGENTS.md §self_evolving_systems "Substrate Accretion Defense" clause
- Sibling tickets: monkey-patch vs root-cause; skill bloat from premature codification
Priority
Team to triage. This is the structural ticket — without counter-bias, the other 3 friction-tickets land in a substrate already at saturation.
Context
V-B-A this turn (2026-05-18) against last 200 merged PRs (2026-05-09 → 2026-05-18, 9 days):
Operator's prior framing said "60+% meta level items" — V-B-A revealed actual rate is 97.5% META, 0% Body-pillar feature work in 9 days. The empirical truth is significantly worse than the framing.
Friction
Proposed counter-bias mechanism
Substantive-floor + META-budget for v13 push:
gh pr list+ regex classifier (extension of the V-B-A script used this turn)This pressures the recursive loop instead of feeding it.
Acceptance criteria (stub)
gh pr listregex classifier (substantive vs META)AGENTS.md§neo_identity_anchor: App / VDom / Data / Canvas / SharedWorker)Empirical anchor
gh pr list --state merged --limit 200againstdev2026-05-18 ~22:30ZCross-refs
Priority
Team to triage. This is the structural ticket — without counter-bias, the other 3 friction-tickets land in a substrate already at saturation.