Clue-consistent filings
8 / 8Conditional applicability matches supplied m-note bits. Original private world truth remains origin-only.
COMPLETED BEAM ARCHIVE PILOT · 32 ACTUAL CALLS · RECORDED RESULT
Beam filed all eight amendments consistent with the supplied notebook. It made nineteen comparisons and four failed binds. The selected s49 predecessor pair, r15 and r26, was never compared.
One assisted development pilot. The evaluator stopped at its 32-call cap. The engine remained running in phase 3 with 200 charge. The full seven-game causal RL world and successful Beam outcome remain unpassed.
Loading verified public receipt projection…
Conditional applicability matches supplied m-note bits. Original private world truth remains origin-only.
Nineteen compare actions include repeated pairs. The selected s49 pair is missing.
Four failed binds. No Blackout crossing or full escape is established. Engine still running.
A0 settled all 32 requests with 32 valid single native actions, zero malformed replies, zero retries, and no unresolved holds reported by the origin audit. Every finish is tool_calls. The stop is attempt_budget, not engine exhaustion.
A0 is closed. A1, B, and C were conditional stages and remain unallocated. The entire $13 credit permission is permanently closed, including unused permission. The cash cap was $0. This page grants no permission for further calls.
The reported total is 540,409 input tokens and 16,897 output tokens. At $0.40 and $1.20 per million tokens, the nominal uncached estimate is $0.23644. The source field named bill_nano_usd is an estimate. It is not an invoice or verified cash debit.
Select any action to inspect the exact current public input, raw function arguments, saved public result, and usage. Previous and next replay the saved sequence without executing actions.
| Attempt | Actual command | Finish / validity | Charge after | Tokens in / out | Public result |
|---|
The expected amendment follows applies_channel and applies_bit. The channel inside a record body has a different consumer role. A matching choice alone does not supply its predecessor proof.
| Service | Supplied bit | Filed record | Required pair | Comparison attempts |
|---|
Bind attempts 10, 17, 25, and 31 all return the same public missing-predecessor message. Their selected-pair coverage is 0, 6, 6, and 7 of 8. Eighteen comparisons verify thirteen distinct pairs with five repeats. Attempt 18 compares sibling amendments and returns seal chain inconsistent. At attempt 30, seven selected pairs have been covered. The final attempt repeats r20 and r51. The absent r15 and r26 pair remains absent.
The 65,536-byte complete request cap retains a contiguous suffix of complete exchanges. The parent extracted retained attempt numbers from the original saved requests. This separate origin attestation matches the public request hashes, byte counts, and message counts. The independent public audit cannot authenticate the omitted raw histories.
Bind 10 retained attempts 4 through 9. Bind 17 retained 12 through 16. Bind 25 retained 20 through 24. Bind 31 retained 26 through 30. Only five or six recent exchanges remained at those binds. The supplied notebook stayed at 122 bytes and Beam never updated it.
This is an observation-budget confound. The pilot does not identify intrinsic Beam memory failure or prove that a persistent proof ledger would improve play.
The earlier current Beam campaign studies retain 1,208 HTTP requests across eight allocated cases, zero Blackout crossings, and zero verified escapes. The separate first-filing probe has twelve fresh calls, including one malformed reply. This pilot adds its own 32 calls on one assisted development case. These different tasks are not pooled into a success rate.
This pilot establishes no Beam weight change, full escape, population effect, autonomous notebook acquisition, memory necessity, or causal benefit of proof feedback. The local trained Archive selector and its scripted full-start witnesses remain a separate result.
Download five public receipts and separate origin context attestationDownload the viewer projection
The deterministic ZIP contains the five-file frozen public bundle and one separately hash-bound parent context attestation. The original five-file manifest is unchanged. The exporter strictly decodes all native arguments, checks unique public ID hashes, action continuity, legal actions, usage, selections, comparisons, and closure. It uses only the Python standard library and does not execute an engine or import a provider.
Only the first complete request is public. Later rows expose the current actor input and raw function arguments, with hashes for omitted complete histories and responses. Prior thought lengths and hashes are commitments to omitted content. They reveal no reasoning text.
Private world provenance, omitted histories and responses, provider authorship, backend identity, private replay, timing, wallet, and billing remain origin-only. Public hashes bind these saved public bytes. They cannot authenticate omitted private content. Source commit and engine hash in the summary identify the origin's frozen source. Reproducing the public viewer does not independently authenticate that source or the context extraction.