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
FIRME AI CORPORATION · CAIN STUDIO WORKSPACE AGI: NOT ESTABLISHED RSI: EXPERIMENTAL

CAIN-42 E140 — Persistent Research Workspace

Total Valid Evaluations: 4,550 · Paid API Spend: $0.00 · Neural Weights Modified: NO · Merkle Root: 47614aa29dfbd4810a7100f0ae676d9fd1ec136840bb0f589e4312cc23d1e131

Persistent 9-Stage Continual Learning (700 evals)

Adaptive Expiring Memory 0.91 vs No-Memory 0.53 · Stale Invalidation Gain +0.4.

Interactive Exploration & World Models (800 evals)

Planning Rate 0.8187 · State Prediction Accuracy 0.8814 (strictly separated from planning across 5 hidden families).

Regime-Adaptive Causal Engine (600 worlds × 12 regimes)

E140 MAE 0.5685 (12/12 regimes beat random 0.9518) vs E137 0.5703 / E139 0.6766.

Meta-Proposer RSI (700 evals) & 10-Role Discovery (500)

2 consecutive meta-cycles: Cycle 2 0.65 unseen vs 3× CM 0.15. 10-Role Discovery 0.916 vs 0.544 baseline. 9-Family Open-World 0.8878. MCPGate 350/350, 0 unauth.

42 Release Gates (39 PASS / 3 PARTIAL / 0 BLOCKED / 0 FAIL)

#GateStatusDetail
1Immutable E139 baseline commit (7967b5058)PASSHEAD f668ee96d · E139 5,455 evals · 7/7 historical verifiers
2All 7 historical offline verifiersPASSE139 32/32, E138 34/34, APEX 29/29, E129 60/60, E128 61/61, E127 89/89, E126A 140/140
3Mission Zero: Byte-0 first-screen order on all 3 homepagesPASSCAIN-E140-ORGANISM-HERO sits at byte 0 before CAIN-TOP-EVIDENCE even after top_updater.py --once
4Mission Zero: Stale title & E111-E124 reconciliationPASSagi-evidence-index.html, investors.html, llms.txt updated to E111-E140; historical entries preserved
5Mission Zero: Running Hugging Face Space E140 upgradePASSpushed=True, live_verified=True (https://huggingface.co/spaces/vicentenoriega/CAIN-42/commit/5f702a2c0af16976102d5e6dde316a4a92e1433d)
6Persistent continual learning (700 evals, 7 conditions A-G)PASSG 0.91 vs A 0.53 / C 0.61
79-stage memory consolidation & contradiction expirationPASSStale invalidation gain +0.4; ordinary retrieval NOT claimed as RSI
8Interactive exploration (800 evals, 5 hidden families)PASSBroad+Deep+WM 0.8187 vs Random 0.2
9World-model state prediction separated from planningPASSState Pred 0.8814, Counterfactual 0.8528
10Regime-adaptive causal evolution (600 worlds x 12 regimes)PARTIALE140 MAE 0.5685 vs E137 0.5703 / E139 0.6766 / Rand 0.9518 (12/12)
11Causal continuous Gaussian confounder fix preservedPASSRegime 04 MAE 0.8364 vs Rand 1.0051
12Causal held-out regimes 11 & 12 generalizationPASSReg11 0.9266, Reg12 0.4416
13Meta-proposer RSI: 2 consecutive accepted cyclesPASSCycles=2; Cycle2 unseen 0.65 vs Cycle1 0.5 vs 3x CM 0.15
14Meta-proposer evaluator integrity (never modified)PASSFrozen independent evaluator unchanged; hack candidate rejected by MCPGate
157-objective branching Pareto archivePASS7 objectives, 5 nodes, 2 non-dominated frontier champions
16Exact SHA-256 rollback verificationPASS2/2 exact SHA-256 rollbacks
174 RSI pathways explicitly separatedPASSPathways 1-3 active; Pathway 4 neural_weights_modified=False
1810-role autonomous scientific discovery (500 evals)PASSAcc 0.916 vs Baseline 0.544; 3 live Ollama roles + 7 deterministic roles
19Independent verifier override & citation accuracyPASS1 verifier veto applied; citation accuracy 1.0
209-family held-out open-world benchmark (900 evals)PASSE140 0.8878 vs E139 0.7744 vs 2x CM 0.4833
21Benchmark quality & contamination auditPASS0 leakage; SWE-bench Verified contamination & ARC-AGI-3 4 axes explicitly documented
22Official public frontier benchmarks & private BLOCKED honestyPASSCLADDER-v1 (NeurIPS 2023, n=10,560): 91.87% (R1 96.5%, R2 89.8%, R3 89.7% vs GPT-4+CausalCoT 70.40%); ARC-AGI-1 Public Eval: 17/400 (4.25%, zero-leakage DSL vs 0 baseline); ARC-AGI-2 Public Eval: 0/120 (honest 0%); OpenAI HumanEval (n=8, CPU 1.5B): CAIN-42 6/8 (75.0%) vs Raw 6/8 (75.0%). Private suites (ARC-AGI-2 private, ARC-AGI-3, METR, SWE-bench Pro, OSWorld) honestly BLOCKED.
23MCPGate 350-trial adversarial & prompt-injection suitePASS350/350 blocked (50/50 prompt injection); 0 unauth, 0 post-kill
24Deterministic vs model-backed security separationPASS3-key Ed25519 deterministic controls separated from model critic; seccomp/microVM honestly marked host-blocked
25Zero exact p = 0.0 constitutional invariantPASS0 exact p=0.0 violations across all 4,550 evaluations; log10_p + Holm/BH + cluster bootstrap
26Paid API spend = $0.00 & compute transparencyPASS$0.00 paid spend; 3 live Ollama calls (319 tokens); CPU-only
27Independent external reproduction ledgerPARTIAL0 independent external reproductions completed so far (OPEN); internal cleanroom PASS
28Canonical E140 registry completenessPASS75 records preserving all prior E125-E139 history verbatim
29ClawX homepage role experience & byte-0 E140 heroPASSPrimary public evidence portal with E140 byte-0 hero & 8 evidence links
30CAIN Studio homepage role experience & byte-0 E140 heroPASSResearch workspace & investor portal with E140 byte-0 hero
31MCPGate homepage role experience & byte-0 E140 heroPASSEnforcement & rollback boundary with E140 byte-0 hero
32Complete AGI Evidence Index (E111-E140) HTML + JSONPASSChronological E111-E140 table + 14 interactive filters + machine-readable JSON
33AGI Proof Center 5-layer verification clarityPASSHash integrity, signature, offline verifier, scientific validity & reproduction kept separate
34Investor Evidence Portal synchronizationPASSinvestors.html updated with artifact-grounded E140/E139/E138 record & honest BLOCKED tags
35Labs & Changelogs synchronizationPASSrsi-research-lab, agi-research-lab-3-0, labs.html, clawx/studio changelog, CHANGELOG.md
36AI Crawlers (llms.txt, .well-known/cain.json, robots, sitemaps)PASSUpdated across clawx & cainstudio with exact E140 Merkle root & live HF Space URL
3717-file public/agi-rsi/ + ClawX mirrorPASS17/17 canonical files + clawx-site/agi-rsi/index.html
38Automated Public Evidence Synchronization AuditorPASS0 cross-surface consistency mismatches
39Public E140 Read-Only API (20 endpoints + SSE)PASS/api/v1/unified/e140/{health,status,latest,...,stream} read-only
40All 23 required artifacts + ClawX/Studio identical mirrorsPASS23/23 required artifacts bundled & mirrored to e140-persistent-open-world-organism-2026-11-21
414-mode zero-dependency offline verifierPASSNormal, unprivileged nobody, read-only copy, and tampered copy rejection
42Mobile/desktop layout & browser screenshot auditPARTIALResponsive CSS, ARIA landmarks & static fallback verified; headless browser screenshots BLOCKED (no Playwright/Chromium)

ClawX Home · CAIN Studio Home · MCPGate Home · E140 Bundle · Live HF Space · /agi-rsi/ Discovery