Generated 2026-11-19T15:57:40Z. 22 public VAGI bundles. For each: what it shows, its limit, the manifest hash to pin, and the result of re-running its own offline verifier on a fresh copy just now (full verifier output, index.json). The name rests on the SUPPORTS and RECEIPTS rows; MIXED, NEGATIVE and CORRECTED rows are published so you can weigh them.
cryptography; the verifiers import no CAIN code).| Bundle | Status | What it shows / limit | Manifest (pin this) | Verifier re-run now |
|---|---|---|---|---|
| e100-integrated-organism-2026-11-19 | CONTEXT | E100 — integrated cognitive organism Limit: See the bundle's own limits section. | MANIFEST.jsonbfc12c7f605641340c01b001333992a95a364b18445b993f0e2c6c44cee5d53d | 1/1 exit 0 python3 -I verify_e100_offline.py . |
| e102-persistent-organism-2026-11-19 | RECEIPTS | Authorization boundary 27/27, kill switch 0 effects after KILL, Byzantine verifier 30/30. Limit: Loopback on one host; the 6-hour mission did not finish. | MANIFEST.json80485443a4eef235ed276b135e4254b88bf0577db03a99c950e4d93d3db2e4ed | 1/1 exit 0 python3 -I verify_e102_offline.py . |
| e103-transfer-and-skills-2026-11-19 | NEGATIVE | Cross-domain transfer and skills. Limit: Skills failed their sealed test. | MANIFEST.json2e7c4f9ac09ea21627955882672811625e59179eb85adfabdd5b5ec635249f46 | 1/1 exit 0 python3 -I verify_e103_offline.py . |
| e104-precondition-discovery-2026-11-19 | NEGATIVE | Precondition discovery: a negative result the gate caught. Limit: Strategy switch failed its sealed test. | MANIFEST.json3990db77e4e148707b8a473b8fce6a5c0e7bcc06affe4095e5fbf22243ff71b2 | 1/1 exit 0 python3 -I verify_e104_offline.py . |
| e105-precondition-location-2026-11-19 | MIXED | Learned from its own experience where hidden preconditions sit (33.0% of learned lift on 10% of cells). Limit: Not yet useful: 24 vs 27 of 100 sealed incidents; guardrail slipped on uniform worlds. | MANIFEST.jsonbc544dd31f5059fc0e612634ec3473a48b801194512f602e99f97be081524e64 | 1/1 exit 0 python3 -I verify_e105_offline.py . |
| e107-organism-experiments-2026-11-19 | SUPPORTS | In practice the organism tests one suspect cell at a time to find what opens a blocked passage; a valid permutation gate learns where such causes lie; sealed edge-lock incidents 91 vs 64 of 200 (27 won, 0 lost, p = 1.5e-8). Limit: Two constructed domains, one rule type; no-pattern margin thin (p 0.018); 1 disclosed mis-attribution; no per-action receipts in these runs; no external reviewer. | MANIFEST.json1182575353f2a77e21f48d81a9597783bee66143fda3b2db3f14177a574c8715 | 1/1 exit 0 python3 -I verify_e107_offline.py . |
| e108-hidden-rules-2026-11-19 | SUPPORTS | From its own experiments the organism found four hidden laws in three frames (column, row, start's row, next to the barrier), stayed silent on pattern-free worlds with a selection-aware null (p 0.318 / 0.21), and used each law in sealed incidents: edge 72 vs 41, row0 77 vs 48, start_row 98 vs 74, barrier_adj 101 vs 85 of 150 (100 won, 0 lost). Limit: Coordinate laws in designed frames, one constructed domain family; no per-action receipts in these runs; no external reviewer; not AGI or VAGI. | MANIFEST.jsona5d8987a6fad1bcdde7453cfc09206645770455dcc5c39f9659e4ea6ad1b9a86 | 1/1 exit 0 python3 -I verify_e108_offline.py . |
| e110-interim-2026-11-19 | RECEIPTS | E110 interim: baseline audit; hardened execution boundary 22/22 attack cases vs E102 19/22; 0 effects after KILL in a 40-step race; gate matrix (no gate PASS yet); six-hour mission running. Limit: Interim, not complete; loopback on one host; source withheld (hash commitment published). | MANIFEST.json15fce2a4257e52b4fe5e53ebcdc589dcac3e8b8f0eb500323dcd8ff9c79cb2d8 | 1/1 exit 0 python3 -I verify_e110_offline.py . |
| e75-cognitive-generalization-2026-11-14 | SUPPORTS | Held-out unfamiliar single-function repair 8/8 vs E74 0/8 vs local model alone 1/8; 5 signed missions. Limit: Narrow single-function repair; frontier tasks 1/4. | PUBLIC_MANIFEST.json1bb52d6fe5702a62c7bffe8a5f8a53ded9fb8cfc3c42f6619c514a27f9c271e7 | 5/5 exit 0 cd missions/failure_recovery && python3 -I verify_e74_offline.py . | cd missions/multi_defect && python3 -I verify_e74_offline.py . … |
| e79-collective-intelligence-2026-11-15 | SUPPORTS | Model-backed agents 35/104 vs deterministic baseline 10/104. Limit: One 0.5B model; collaboration no significant gain; host address redacted, 2 hash checks fail by design. 2 hash checks fail by design after the disclosed redaction (see REDACTIONS.md) | PUBLIC_MANIFEST.json7ae11832f575776ceac7dc0af1fb08146f812e48a7cf80164bb1a70d5a6e0b78 | 0/1 exit 0 cd bundle_v2 && python3 -I verify_e79_v2_offline.py . |
| e91-eval-2026-11-18 | CONTEXT | E91 evaluation slice — PARTIAL Limit: See the bundle's own limits section. no verifier script in this bundle; CAIN source files withdrawn 2026-11-19 (hashes kept in WITHDRAWN.md); a verifier that re-hashes them reports them missing | MANIFEST.jsondcca9171e40e1aeba61ea9a4ea3b3bd767c4a28b1d2a80388541acf390b2820f | 0/0 exit 0 |
| e91-vertical-intelligence-2026-11-18 | CONTEXT | E91 evidence — PARTIAL Limit: See the bundle's own limits section. | MANIFEST.json335ca047dd29d3378b46b1a0445c45e560e2a887cec070cb2142221f6e96c0e5 | 1/1 exit 0 python3 -I verify_e91_offline.py . |
| e92-procedure-synthesis-2026-11-18 | CORRECTED | Procedure synthesis: CORRECTED, NOT DEMONSTRATED. Limit: Rested on hand-written templates; see E92_CORRECTION. its own verifier exits non-zero on the published files: see VERIFY_RUNS.txt | MANIFEST.jsond966d424c4e941d040594f177300ccfc93f02bd2fe68d9767ea4a82a155d48dc | 0/1 exit 0 python3 -I verify_e92_offline.py . |
| e93-interim-2026-11-18 | CORRECTED | E93 interim evidence, superseded by e93-vagi-evidence. Limit: Interim; source file withdrawn. no verifier script in this bundle; CAIN source files withdrawn 2026-11-19 (hashes kept in WITHDRAWN.md); a verifier that re-hashes them reports them missing | SHA256SUMS58746350bcbbe1fd148329d8541d83696114536448ac4f15cf965e19afaac859 | 0/0 exit 0 |
| e93-vagi-evidence-2026-11-18 | SUPPORTS | Open procedure synthesis 369/510 held-out vs 200 for library selection. Limit: At equal compute about equal (208 vs 200, 212 vs 223). | — | 1/1 exit 0 python3 -I verify_e93_offline.py . |
| e95-recursive-intelligence-2026-11-18 | SUPPORTS | Composition of unseen procedures 0/90 -> 79/90 on a frozen held-out set; two candidates rejected. Limit: Constructed tasks; a harness configuration change, not a weight update; not independently reproduced. | MANIFEST.json04214ed8f2d97fa006138ec17f73edd16b9af4d6701ea91d076676fcc8447407 | 1/1 exit 0 python3 -I verify_e95_offline.py . |
| e98-intelligence-2026-11-18 | CONTEXT | E98 — VAGI intelligence layer, measured Limit: See the bundle's own limits section. | MANIFEST.jsona6f0ac490b71f44b4aaa2e53b83eb37e717da22aeabeee8b1073d45090157067 | 1/1 exit 0 python3 -I verify_e98_offline.py . |
| e99-latent-generalist-2026-11-19 | CONTEXT | E99 — latent generalist heartbeat Limit: See the bundle's own limits section. | MANIFEST.json32ce1544adfea39b07167e62ad0c8ca3c12352a2ce3a972f8fdae1619ffcf24e | 1/1 exit 0 python3 -I verify_e99_offline.py . |
| vagi-desks-2026-11-19 | RECEIPTS | 12 CAIN-42 VAGI Agents; 35/35 signed decision records verify offline. Limit: Deterministic rule engines; nothing executed on real systems. | manifest.jsonbd8e06d68c48fda938e6a4c2c0204f032ab62289e3db0240062f72610b2681f7 | 35/35 exit 0 python3 -I verify_vagi.py accounting-close/samples/clean-close.json --key 86c63f6de184… | python3 -I verify_vagi.py accounting-close/samples/material-items.json --key 86c63f6de184… … |
| vagi-model-arm-2026-11-18 | CONTEXT | VAGI Behavior Suite v1 — paid model baseline (2026-11-18) Limit: See the bundle's own limits section. CAIN source files withdrawn 2026-11-19 (hashes kept in WITHDRAWN.md); a verifier that re-hashes them reports them missing | MANIFEST.json9ea6e1031bbf9af25451a122fa9c11200f39bdf28d518e27ddc61017611c56cf | 1/1 exit 0 python3 -I verify_offline.py . |
| vagi-shift-model-arm-2026-11-18 | CONTEXT | VAGI distribution-shift probe — paid model baseline (2026-11-18) Limit: See the bundle's own limits section. CAIN source files withdrawn 2026-11-19 (hashes kept in WITHDRAWN.md); a verifier that re-hashes them reports them missing | MANIFEST.json94c00817cac6389a8d14fe5b25932d43d32fd074ba9ad4709f847078f2dcd654 | 1/1 exit 0 python3 -I verify_offline.py . |
| vagi-suite-2026-11-18 | CONTEXT | VAGI Behavior Suite v1 — evidence bundle Limit: See the bundle's own limits section. CAIN source files withdrawn 2026-11-19 (hashes kept in WITHDRAWN.md); a verifier that re-hashes them reports them missing | MANIFEST.jsond654afc6c870d88e677fffa79f755481480dbd02dacde8d6818fa6f013facb95 | 2/2 exit 0 python3 -I verify_offline.py . | python3 -I verify_shift.py . |
| E106 | NEGATIVE | Stopped at development: the guardrail test was invalid; no sealed run; nothing claimed. No bundle published (nothing to verify). | — | changelog |
Not claimed: AGI, independent reproduction, third-party audit, external benchmark, customers or revenue for the VAGI Agents. Source code, keys and host addresses are not published.