CAIN-42 CAIN StudioCAIN-42 evidence library
CAIN-42 Evolution 30 -- Governed Machine Autonomy Fabric
Evidence bundle for evolution E30: 6 claims, 166 of 166 invariants held.
Last reviewed 2026-10-01
Browse the raw bundle · How to reproduce it · SHA-256 manifest
Claims (6)
| Claim | Level |
|---|---|
| C42-E30-KERNEL: one action path (kernel -> E28 -> E25 -> E8) that checks autonomy state, level, budget, containment, environment, perception, memory and intent before authorizing, and a signed universal action receipt binding a 20-field action digest for every action, allowed or refused | TESTED |
| C42-E30-E2E: a 28-step run from agent identity through perception, memory, reasoning, a governed transaction, E8, self-test, a firewalled self-improvement with human review and canary, promotion, rejection and continuous reauthorization | TESTED |
| C42-E30-BENCH: 889 of 889 adversarial and integration scenarios held; 166 of 166 invariants; mutation 11 of 11 | TESTED |
| C42-E30-COVERAGE: a governance coverage map that reports UNCONTROLLED and UNKNOWN paths and never counts monitored as enforced | IMPLEMENTED |
| C42-E30-TS-SDK: a TypeScript verifier SDK re-checks the receipts without CAIN code and detects a tampered receipt | TESTED |
| C42-E30-SCALE: synthetic in-process runs at 10/100/1,000/10,000 real agents and a 100,000-agent virtual model | SIMULATED |
Invariants: 166 of 166 held
Rules the code must never break, each checked across many scenarios. See all 166 on one page.
| Niche | Held |
|---|---|
| Transactions, markets & economics | 30 of 30 |
| Policy, law & governance | 21 of 21 |
| Autonomy, control loops & recovery | 21 of 21 |
| Tools, MCP, protocols & adapters | 21 of 21 |
| Benchmarks, coverage & performance | 18 of 18 |
| Evidence, receipts & proofs | 14 of 14 |
| Trust & reputation | 12 of 12 |
| Prediction, world models & simulation | 10 of 10 |
| Memory, data & privacy | 9 of 9 |
| Core guarantees | 7 of 7 |
| Identity, authority & delegation | 2 of 2 |
| Attacks, threats & containment | 1 of 1 |
Verify it in your browser
Your browser downloads the bundle's SHA256SUMS manifest and the file(s) behind this page, hashes them with SHA-256 locally (WebCrypto), and compares. A match shows the record you are reading is the published one; it does not by itself prove who published it (see the signed claims registry and the bundle verifier for that).
Known limitations
- An integration library in one process; not hosted; not wired into the gateway, MCPGate or the clusters.
- Governance coverage outside the E25/E8 boundary is UNCONTROLLED or UNKNOWN (see COVERAGE_MAP.json); robots, vehicles, laboratories and physical actuators are UNENFORCED -- CAIN refuses to act there, it does not govern them.
- Perception VERIFIED means two independent registered sensor keys agreed; it says nothing about the physical truth of the reading.
- Injection detection in memory, perception and intent is a marker list; recall against real attacks is UNKNOWN.
- The self-improvement firewall governs the process; it cannot tell whether a promoted change is actually better or safe beyond the recorded stage evidence.
- Grid scenarios are distinct inputs over shared defences; the mutation self-test shows which defences they exercise.
- The TypeScript SDK is a verifier; agent-side TypeScript integration is not built.
- The research engine and frontier lab are registers only; no automated research feed runs.
- Scale rows are synthetic and in-process; the 100,000 row is a virtual model without signatures or E8.
- Hardware attestation UNKNOWN; zero-knowledge proofs NOT IMPLEMENTED; no external review.
The bundle's own README
CAIN-42 Evolution 30 -- Governed Machine Autonomy Fabric#
An integration kernel over E8 and E12-E29: every consequential action is normalized into a 20-field Universal Machine Action, checked against autonomy state, level, budget, containment, environment, perception, memory and intent, executed only through E28 -> E25 -> E8, and recorded with a signed universal action receipt. Verify with verify_e30.py.txt or the TypeScript SDK cain-verify.ts.txt (see REPRODUCTION.md). Status: TESTED, pre-production.
Try CAIN-42 on your own agents
Create a free account and every new account starts with a 7-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