CAIN-42 CAIN Studio
FIRME AI CORPORATION · CAIN-42 INVESTOR EVIDENCE PORTAL

CAIN-42 Proto-VAGI: 42 claims. Every one links to its proof, and you can check the proof yourself.

CAIN-42 PROTO-VAGI · INFRASTRUCTURE GOVERNOR OF AGI · NOT YET AGI

CAIN-42 Proto-VAGI is the infrastructure governor of AGI and a VAGI research platform. It is not yet AGI; its intelligence results are early and bounded. The infrastructure evidence is real; the intelligence claim is deliberately limited.

AGI dossier → · Proof factory → · E100 organism · E99 recurrent · VAGI suite · E95 harness

CAIN-42 PROTO-VAGI CAIN-42 Proto-VAGI is the infrastructure governor of AGI. CAIN-42 Proto-VAGI: early, measured signs of intelligence, and every action it takes comes with a signed receipt. Auditor index: every VAGI bundle, verifier re-runs, manifest hashes →
Early, measured signs of intelligence
  • E95: on a frozen held-out set, composing procedures it had not seen went from 0/90 to 79/90 after one harness improvement (two candidates rejected). evidence (Constructed tasks; a configuration change, not a weight update.)
  • E93: open procedure synthesis solved 369/510 held-out tasks versus 200 for selecting from a supplied library. evidence (At equal compute the two were about equal (208 vs 200).)
  • E105: it learned from its own experience where hidden preconditions sit (33.0% of its learned lift on the column holding 10% of cells), with no code that knows the answer. evidence (Learned, not yet useful: steering with it did not raise success (24 vs 27 of 100).)
  • E75: on held-out unfamiliar code repairs it solved 8/8, where the previous version solved 0/8. evidence (Narrow single-function repair; frontier tasks 1/4; the local model alone 1/8.)
  • E79: model-backed agents answered 35/104 tasks vs 10/104 for the deterministic baseline (about 3.5x). evidence (One 0.5B model; collaboration gave no significant gain; host address redacted (disclosed).)
Every action comes with a signed receipt
  • 12 CAIN-42 VAGI Agents live: 35/35 published signed records verify offline with a verifier that imports no CAIN code. evidence (Deterministic rule engines; nothing executed on real systems.)
  • Authorization boundary: 27/27 adversarial cases pass, 0 unauthorized effects; kill switch: 0 effects after the KILL order. evidence (Loopback on one host, not the production MCPGate.)
  • E110 governed organism: every action of the E108 organism in 120 sealed incidents went through the hardened boundary: 9,240 signed, hash-chained receipts; outcomes identical to the direct runs (120/120); 0 effects after an independent KILL. evidence (Loopback on one host, not the production MCPGate.)
  • E110 (in progress): hardened execution boundary passes 22/22 attack cases (the E102 boundary: 19); independent emergency KILL/REVOKE path; 0 effects after KILL in a 40-step race. Six-hour governed mission running. evidence (Interim; loopback, one host; no gate claimed PASS yet.)

Proto means early. CAIN-42 is governed AGI infrastructure; it is not yet AGI. The intelligence results are on constructed tasks, measured by the same project and not independently reproduced. Also on record: negative results (E103, E104, E106) in the proof factory.

New: CAIN-42 VAGI Agents — governed vertical agents (12 of 12 live)

The commercial wedge: industries where the buyer must keep records of what an agent did (insurers of AI agents, carriers and MGAs under the NAIC AI bulletin, investment advisers under SEC record-keeping). Competitors such as SUPERAGENT (insurance, marketed as "Vertical AGI") gate actions with an approval click; CAIN-42 VAGI Agents return a signed record a third party can re-check offline.

Verify: evidence bundle · offline verifier · signing key. Tests 99 passed / 0 failed.

Not claimed: customers or revenue for these agents, third-party audit, certification, or AGI. CAIN-42 is governed AGI infrastructure; it is not yet AGI. Prices are proposals.

AGI INFRASTRUCTURE & RESEARCH LAB · EVIDENCE DOSSIER · 2026-11-19

CAIN-42 is AGI governed infrastructure and an AGI research lab. It is not yet AGI.

Its intelligence results are early and bounded. The infrastructure claim is backed by authorization, enforcement, BFT quorum, kill-switch, persistence and signed evidence. The research claim is backed by executed, verifier-checked experiments — including their failures.

Proof factory → · Machine-readable dossier →

1. Infrastructure evidence

ControlStatusEvidenceLink
Agent identity and scoped authorityIMPLEMENTEDEd25519 grants, capability attenuation, revocation; E96 governance suite 11/11.evidence
Default-deny enforcementIMPLEMENTEDIn-process CAIN/MCPGate bridge; every real E98/E99/E100 action authorized; 0 unauthorized effects.evidence
Byzantine quorumIMPLEMENTEDN=4, f=1, Q=3; 3-of-4 BFT receipt on every governed action.evidence
Kill switchTESTEDMonotonic fence; 0 effects after the fence; post-fence attempt recorded.evidence
Persistent mission and resumeTESTEDE100 3/3 organism missions; resume 2/2; hash-chained memory.evidence
Real OS workersTESTED10 processes; 0 duplicate seeds; 28/40 success.evidence
48h multi-region soakRUNNING — NOT PASSED22.9/48h at last heartbeat; 0 divergences, 0 anomalies; not complete.evidence

2. Research lab evidence

RunResultLabelVerifierLimitRaw
E95 — recursive harness improvement0/90 → 79/90 held-out compositionTEST56/56Constructed numeric substrate; no external reviewer.raw
E98 — learned-control pulse35/50 model-based vs 4/50 randomTEST38/383 gain families; skill weak; failure F1 0.raw
E99 — latent recurrent generalist93/84/88 vs MLP 68/63/70 vs random 20/24/23TEST40/406 constructed families; no external reviewer.raw
E100 — integrated cognitive organismdelayed 52/60 vs MLP 43/60; hysteresis 58/60 vs 56/60TEST32/32zero-shot 53/60; few-shot 19/60 rejected; skill 45/50; failure F1 0.6667.raw
VAGI Behavior Suite v1library 500/500 in-distribution; shift 5/100TEST49/49 + 12/12Hand-written procedure library; constructed tasks; shift gap is the finding.raw
Paid model baselinesHaiku 481/500 in-distribution; shift Haiku 74/100, Sonnet 81/100TEST49/49 + 87/87Constructed suite; policy card supplied; no external reviewer.raw

3. What is not proven

4. Verify it yourself

Self-reported by FIRME AI CORPORATION; no independent third-party evaluation, no AGI certification. The 48-hour soak is still running and is not counted as passed. Historical evidence keeps its dated verdicts; current claim: not yet AGI.

VAGI PROOF FACTORY · INVESTOR VIEW · 2026-11-19

A proof factory, not a claims page.

Each lane has a metric, a verifier and a stated failure boundary. CAIN-42 Proto-VAGI is the infrastructure governor of AGI. It is not yet AGI; its intelligence results are early and bounded.

VAGI ORGANISM · ALL EVIDENCECAIN-42 Proto-VAGI: early, measured signs of intelligence, and every action it takes comes with a signed receipt.From a single pulse to an organism that learns hidden rules from its own experiments — every step, including the negatives and corrections.
E98 · PulseGAIN
A learned world model improves closed-loop decisions.
35/50 model-based vs 4/50 random (+62.0 pp)
E99 · Rapid heartbeatCORRECTED
Recurrent advantage did not survive E102's re-score; the world model did.
GRU 93/84/88 of 100 vs MLP 68/63/70 vs random 20/24/23
E100–E101 · Integrated organismCORRECTED
Measured on already-solved environments; re-scored by E102.
Delayed 52/60 vs MLP 43/60; hysteresis 58/60 vs 56/60
E102 · Persistent governed organismGAIN
Learns the world while it acts; beats frozen and reactive baselines.
252/300 vs frozen 198/300 vs reactive 154/300 (McNemar p=8.1e-09); 40 poisoned memories: undefended -44.5 pp, causal -1.0 pp
E103 · Transfer and skillsMIXED
Cross-domain transfer positive; skills failed held-out.
service -> grid: faster on 56 vs 14 tasks (p=4.3e-07); grid -> service: 78 vs 52 (p=0.03)
E104 · Strategy switchNEGATIVE
Rejected by the preregistered gate, which prevented a loss.
sealed grid 256 vs 257 of 300; key-and-door 35 vs 35; service lock 18 vs 18; plain service 54 vs 65 (a loss the gate prevented)
E105 · Where preconditions areNEGATIVE
Learned the drain column, but it did not raise success.
edge-lock 24 vs 27 of 100 with vs without the prior; uniform-key worlds 25 vs 26 of 60 (no harm)
E106 · Subgoal + permutation gateSTOPPED
Stopped before a sealed run: the gate was inverted. Nothing claimed.
development only; no sealed run
E107 · Runs its own experimentsGAIN
Interventional attribution + valid gate: more incidents solved, none lost.
edge-lock 91 vs 64 of 200 (27 won, 0 lost, p = 1.5e-08); uniform-key worlds unchanged 57 vs 57 and 74 vs 74
E108 · Learns many kinds of hidden rulesGAIN
Selection-aware gate; four rules in three frames found and used; none lost.
edge 72 vs 41/150 · row0 77 vs 48/150 · start_row 98 vs 74/150 · barrier_adj 101 vs 85/150; uniform controls unchanged (63 vs 63, 57 vs 57)
E110 · Every action receiptedRECEIPTS
Organism acts only through the hardened boundary; a signed receipt per action; KILL fences it.
9240 receipts for 9120 actions + 120 opens; parity with direct runs 120/120; permit effects after KILL 0; pre-issued permit after KILL refused

Every step links to its evidence bundle and offline verifier (hover for limits). TEST results on constructed tasks; no external reviewer. Receipts: the governed boundary issues an Ed25519-signed receipt for every action it admits (E102, loopback); the E107/E108 sealed runs called the world directly and carry no per-action receipts. CAIN-42 Proto-VAGI is the infrastructure governor of AGI; it is not yet AGI. Source code is not published. Auditor index →

AGI infrastructure — governance holds at the actionTEST
27/27 adversarial cases pass; 0 unauthorized effects; 30 false success claims, all rejected
CAIN decides, the resource enforces, and every effect is recorded in hash chains; consensus proofs are published every 30 minutes and anchored daily.
Limit: Loopback boundary, not production MCPGate; vendor-operated validators; PBFT incidents 2026-11-15 and 2026-11-18 disclosed; no external audit.
E110 every organism action carries a signed receiptTEST
9240 receipts for 9120 actions + 120 opens; parity with direct runs 120/120; permit effects after KILL 0; pre-issued permit after KILL refused
The E108 organism ran 120 sealed incidents with every action routed through the hardened resource boundary: signed permit, resource-side check, Ed25519-signed hash-chained receipt. A KILL sent to the resource on an independent key stopped every later effect while the decision service still answered ALLOW.
Limit: Loopback, OS processes on one host (not separate failure domains, not the production MCPGate); a subset of E108 sealed incidents; no external reviewer. Not AGI.
E108 the organism learns many kinds of hidden rulesTEST
edge 72 vs 41/150 · row0 77 vs 48/150 · start_row 98 vs 74/150 · barrier_adj 101 vs 85/150; uniform controls unchanged (63 vs 63, 57 vs 57)
From its own experiments the organism found four different hidden laws (a column, a row, its start's row, next to the barrier) in three frames, stayed silent on pattern-free worlds with a selection-aware null, and used each law: 100 incidents won, 0 lost.
Limit: Coordinate laws in designed frames, one constructed domain family; uniform gates off at p 0.318 / 0.21; no per-action receipts in these runs; no external reviewer. Not AGI or VAGI.

Open the full proof factory → · E98 raw rows →

Self-reported by FIRME AI CORPORATION; no independent third-party evaluation. Not an offer to sell, or a solicitation of an offer to buy, any security.

E98 · VAGI INTELLIGENCE PULSE · 2026-11-19 · NOT A GENERAL VAGI CLAIM

A learned neural world model, closed-loop, under the governor.

Sealed closed-loop success 35/50 (70%) for the model-based planner vs 4/50 random; +62.0 pp. Model: 6902 transitions, state MAE 0.017954, ECE 0.022124.

What the evidence supports
A measured learned-control pulse: prediction, closed-loop planning, real OS workers, and CAIN/MCPGate/BFT authorization with a working kill switch.
What it does not support
General VAGI intelligence; broad open-world generalization; strong skill synthesis (7/24); reliable natural-failure localization (F1 0.0); external verification.
What would change the rating
A sealed 80%+ result on a new rule family, skill synthesis near 80%, natural-failure F1 ≥0.8, and an external reviewer.

Raw rows, model, failures and verifier → · Mirror →

Self-reported by FIRME AI CORPORATION; no independent third-party evaluation. Not an offer to sell, or a solicitation of an offer to buy, any security.

E95 · RECURSIVE INTELLIGENCE IMPROVEMENT · 2026-11-18 · NOT A VAGI CLAIM

A measured harness weakness, five candidates, two rejections, one held-out gain.

Atom baseline 0/90 on frozen held-out composition; selected adaptive harness 79/90; familiar non-regression 60/60 (baseline 60/60). Backward revision kept 9 of 9 macros.

What the evidence supports
A controlled improvement loop: observe a measured weakness, generate competing configurations, reject two, promote one, verify on a frozen split, and preserve the history.
What it does not support
AGI/VAGI; weight-level self-improvement; general professional work; multi-agent collaboration gain; independent reproduction; production readiness.
What would change the rating
An external reviewer, a new frozen atom pool, the two unsolved nested families, and a real-domain task where the harness gain survives.

Raw rows, candidates, failures and verifier → · Mirror →

Self-reported by FIRME AI CORPORATION; no independent third-party evaluation. This page is not an offer to sell, or a solicitation of an offer to buy, any security.

VAGI PROGRAM · EVIDENCE OF RECORD · 2026-11-18 · NOT A VAGI CLAIM

10 domains. 500 held-out tasks. A published failure boundary.

In-distribution procedure library 500/500; no-procedure floor 72/500; one fixed procedure 112/500. On 100 unfamiliar variants the same library scores 5/100, while raw models score far higher: that is a library-coverage map, not a general-intelligence result.

Frontier baseline: Haiku 5.5 481/500 (96.2%) and Fable 5.2 96/100, with the same policy card, no tools, one attempt. This is why the in-distribution 500/500 is a ceiling, not a moat. Model baseline + verifier →

Shift probe: library 5/100 vs Haiku 5.5 74/100 and Sonnet 5.5 81/100. The library gap is real, but it is a coverage gap, not a general-intelligence gap. Shift baseline + verifier →

What the evidence supports
A repeatable, deterministic method for measuring capability and publishing its failure boundary; a governed procedure runtime; an offline verifier that imports no CAIN code.
What it does not support
VAGI/AGI attained; generalization to unfamiliar variants; external validation; a frontier-model benchmark; revenue, valuation or product-market fit.
What would change the rating
An external verifier, an independent witness, shift performance near the in-distribution target, and a paying customer using it in production.
Not checked yet. Pinned sha256 56876ef17f245a9d… Evidence bundle + verifier → Per-domain table →

Self-reported by FIRME AI CORPORATION; independent third-party evaluation has not yet been obtained. VAGI infrastructure is a 2027 goal, not a current capability. This page is not an offer to sell, or a solicitation of an offer to buy, any security.

CAIN-42 gives AI agents memory and enforceable limits. On an independent benchmark it doubled a frontier model's score over no memory, at about one-seventh the cost of the brute-force approach, and every number is published with a verifier you can run yourself.

Each claim shows its statement, its stated limits and its evidence files. Check downloads a proof file in your browser, computes its SHA-256 and compares it with the signed digest. The full register, including failed and unverified items, is here. Status words: VERIFIED and REPRODUCED = re-checked against raw evidence; MEASURED = an experiment result; TESTED = code with passing tests, not a deployed service.

Self-reported by FIRME AI CORPORATION; independent third-party evaluation has not yet been obtained. VAGI infrastructure is a 2027 goal, not a current capability. This page is not an offer to sell, or a solicitation of an offer to buy, any security; any offering is made only through definitive documents to eligible investors. Questions and data-room requests: firmeai2026@gmail.com.