CAIN-42 CAIN Studio
CAIN-42 is an AGI–RSI research organism. Governed general-intelligence and recursive-self-improvement research, with signed evidence for every governed action. AGI: NOT ACHIEVED · NOT ESTABLISHED · RSI: EXPERIMENTAL. RSI Lab · Two Labs · AGI Lab · E135 · 35/48 · E136 · tier-12 · E132 · E127 · E126B · APEX · E137 · AGI+RSI Index · Investor

AGI STATUS: NOT ESTABLISHED · TWO LABS · ONE GOVERNED RUNTIME

CAIN-42 is an AGI–RSI Organism — The Open AGI + RSI Research Laboratory

AGI: NOT ESTABLISHED · RSI: EXPERIMENTAL

Two distinct research programs share one execution and evidence fabric. Neither is AGI. The RSI lab is bounded and constructed, not unrestricted recursive self-improvement.

AGI Lab — general intelligence

E127 living research lab · MEASURED · LOCAL
100 registered identities, 8 configured, 5 peak real workers, 60 tasks, 7 missions, 144 hash-chained events, 177.87 s, 0 failed, 0 dead. Not 24/7. report · verifier
E126.b continual learning · PASS vs no-memory
0.6667 vs 0.5556 (+11.1 pts, p=0.00049); retention 0.667 vs 0.500; poison 1/1 refused. Ties retrieval/episodic. evidence
E126.b novel environments · PARTIAL
0.895 vs retrieval 0.860 (p=0.210, H1 not met); vs E125 0.820 (p=0.020). Category E in E125 was 0.000. evidence
E126.b causal active selection · FAIL
25 active wins / 25 random wins / 30 ties. Mean MAE lower (0.1101 vs 0.1336) but preregistered paired test failed. Counterfactuals: 0.0768 vs passive 0.2812. negative result
E126.b transfer / collective / metacognition · PASS
Transfer +12.5 pts applied, 0/40 false accepts. Collective 0.666 vs 0.480 best single, 0/500 Byzantine flips. Metacognition Brier 0.0926 / ECE 0.0082. report
E125 compositional generalization · PASS
112/120 vs 98/120, p=0.00131, Holm 0.00262; category E 0.000; 81/81 verifier. report

RSI Lab — governed recursive self-improvement

E132 governed RSI loop · PARTIAL_RUN · 14/18 gates
Main 30/48 vs same-library matched-budget fixed 28/48; library 21→117; depth 1→8; tier-6 1.000 vs fixed 0.500; tiers 8/12/16 = 0.000; 1,740/1,740 offline checks. Verdict is bounded, not unrestricted RSI. bundle
E129 self-written library · POSITIVE_BOUNDED · 11/11
Sealed solves 8→28→35→35; library 21→96; depth 1→8; growth stops at generation 2. bundle
E130 scaled RSI · NEGATIVE_OR_PARTIAL · 9/14
Main 26/48 vs fixed 28/48; meta-improver hurt; generation 1 rolled back on dev-t4-0018. bundle
RSI ceiling · PARTIAL
Tiers 8, 12 and 16 remain 0/6. Constructed integer domain; no model trained; no external reviewer. This is not AGI, VAGI or unrestricted RSI.
E135 tier-8 RSI · POSITIVE_BOUNDED_RSI · 20/20
Main 35/48 sealed; tier-6 6/6; tier-8 5/6; flat 22; random 15; compute 6; frozen 6. Seed-131 replication repeats (35/48, tier-8 5/6). 1,551/1,551 offline checks. Tiers 12/16 remain 0/6; not unrestricted RSI. bundle · deep RSI lab page

Long-horizon infrastructure

24-hour soak · PASS · 24.003 h, 24 checkpoints, 0 divergences (2026-11-02).
48-hour soak · FAILED — NOT ESTABLISHED · Failed at checkpoint 0 after conflicting PBFT quorum certificates on cain-mr-02; cluster quarantined. No 48h pass exists.

Evidence and reproducibility

Every number above is drawn from a committed bundle with an offline verifier. AGI is not established. No external benchmark, certification or external human review is claimed. AGI+RSI evidence index · proof factory