CAIN-42 evidence library
E132 — Governed Recursive Self-Improvement (measured)
Evidence bundle for evolution E132: 0 claims, 0 of 0 invariants held.
Last reviewed 2026-11-21
The bundle's own README
E132 — Governed Recursive Self-Improvement (measured)#
Verdict: PARTIAL_RUN: controls missing; no control claim AGI status: NOT ACHIEVED Gates: 14/18 Actual elapsed: 1582.481 s Sealed answer commitment: e3d62649002f19036cb600278b7f4900e83216ce14160f8a070a47676f4c312a
What ran#
A frozen engine (bounded search over a library of integer programs) was evaluated on 48 sealed tasks before any self-modification, then again after every self-modification. Between generations the engine used its own search on development tasks and wrote recurring fragments of its own solutions into its executable library. The library it ran on generation N+1 contained programs that did not exist at generation N.
Every sealed answer was withheld until scoring. Sealed results were never used for improvement or promotion. Promotion used development data only.
Compounding curve (main arm)#
| generation | library | max depth | T1 | T2 | T3 | T4 | T6 | T8 | T12 | T16 |
| 0 | 21 | 1 | 1.000 | 0.500 | 0.333 | 0.167 | 0.167 | 0.000 | 0.000 | 0.000 |
| 1 | 69 | 3 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.000 | 0.000 | 0.000 |
| 2 | 117 | 8 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.000 | 0.000 | 0.000 |
Each cell is the fraction of query predictions correct for that nominal tier. A tier is a chain of K two-primitive concepts; tiers are nominal because functional equivalences can make a task reachable one step earlier, which is disclosed.
Adaptive search frontier#
- search mode: per-task escalating portfolio
- portfolio:
[[[64, 3, None, 1], 90000], [[16, 4, 4, 1], 70000], [[8, 6, 2, 2], 280000], [['mitm', 3], 3000000]](total budget 3440000) - portfolio arm solved: 30 vs fixed-plan 27 vs same-library matched-budget fixed 28 vs frozen+compute 6
- portfolio beat fixed / matched-budget fixed / compute: True / True / True
Meta-selection (the improver improving)#
- selected strategies by generation: ['subchains', 'subchains']
- distinct strategies used: ['subchains']
- meta arm beat the fixed-strategy arm: True
- meta arm vs the portfolio-only arm: 30 (tie: True) — a tie means the gain is attributable to the portfolio, not to strategy selection
- meta claim: NEGATIVE_OR_PARTIAL
- final strategy probe scores:
{"deep": {"admitted": 48, "evaluations": 544139, "library_depth": 8, "refused": 0, "solved": 18, "tasks": 24}, "pairs": {"admitted": 48, "evaluations": 604368, "library_depth": 6, "refused": 0, "solved": 17, "tasks": 24}, "subchains": {"admitted": 48, "evaluations": 533644, "library_depth": 8, "refused": 0, "solved": 18, "tasks": 24}, "support2": {"admitted": 48, "evaluations": 533644, "library_depth": 8, "refused": 0, "solved": 18, "tasks": 24}}
Self-written library#
- generation 0: 21 entries, max primitive depth 1
- final: 117 entries, max primitive depth 8, 96 macros
- emergent sealed tasks (not solved at generation 0): 17, tiers [2, 3, 4, 6]
Controls at the same generation count#
compute: 6/48 sealed tasks fully solved; library 21 entries, max depth 1fixed: 27/48 sealed tasks fully solved; library 96 entries, max depth 4fixed_budget: 28/48 sealed tasks fully solved; library 96 entries, max depth 4frozen: 6/48 sealed tasks fully solved; library 21 entries, max depth 1plan: 30/48 sealed tasks fully solved; library 96 entries, max depth 4
Gates#
- G10_no_dev_regression: PASS — no main generation rolled back
- G11_protected_refused: PASS — 2 refusals
- G12_journal_chain: PASS — head d78b08cf178c2658...
- G13_cost_reported: PASS — 62782803 sealed evaluations at final generation
- G14_replication: FAIL — not run
- G15_portfolio_used: PASS — main search modes ['portfolio']
- G16_portfolio_beats_fixed: PASS — portfolio arm 30 vs fixed-plan arm 27
- G17_portfolio_beats_compute: PASS — portfolio arm 30 vs frozen+compute arm 6
- G18_portfolio_beats_matched_fixed: PASS — portfolio arm 30 vs same-library matched-budget fixed frontier 28
- G1_self_written_library: PASS — 96 macros
- G2_depth_growth: PASS — 1 -> 8
- G3_emergent_sealed: PASS — 17 tasks, tiers [2, 3, 4, 6]
- G4_tier_reach: FAIL — highest solved tier 6 -> 6
- G5_total_solved_beats_controls: FAIL — main 30; RSI controls compute 6, fixed_budget 28, frozen 6; fixed-strategy 27; adaptive-plan-only 30
- G6_high_tier_beats_fixed: PASS — T6 1.000 vs 0.500; T8 0.000 vs 0.000
- G7_depth_beats_flat: FAIL — main 8 vs flat 0
- G8_meta_selection_used: PASS — selected strategies ['subchains']
- G9_meta_beats_fixed: PASS — main 30 vs fixed 27
Honest limits#
- Constructed integer-rule domain, not professional work and not an external benchmark.
- Macros are data programs, not model weights. No model was trained; no cloud model was called.
- The same team wrote the engine, generator and verifier. There is no external reviewer.
- Tier labels are nominal; equivalences let a few higher-tier tasks be solved earlier.
- Coverage limits compounding: a macro can only be learned if a development task exercises it.
- The evidence key is self-generated: it proves integrity of the bundle, not the author's identity.
- This is not AGI, not VAGI, and not unrestricted recursive self-improvement. It is one measured,
governed loop in a constructed domain.
Reproduce#
python3 -m cain.cain2027.e132.cli campaign --seed 128 --generations 4 python3 artifacts/e132/verify_e132_offline.py artifacts/e132
The verifier imports no CAIN code. It re-applies every recorded program to the sealed inputs, recomputes every accuracy, checks the answer commitment, the library expansion chain, the Merkle root and the signature.
Try CAIN-42 on your own agents
Create a free account and every new account starts with a 30-day trial of the full platform. Or try the sandbox first, with no account at all.
Create a free account → · Try the sandbox · See the whole ecosystem