Protocol STAMP is teamed with AI and can make mistakes. Verify before you implement anything. What we got wrong · Free to read, paid for by readers: Contribute
Below is the issue exactly as it was issued; nothing in it has been edited. This frame, the menu and the footer are the site’s, not the issue’s. To check it, download the Source and hash it: the first twelve characters of its SHA-256 are the build above. The Exact copy in the public repository is the same bytes.
| candidate | failure rate | items | draws | noise floor | reliability | ICC | design effect |
|---|---|---|---|---|---|---|---|
| luna_v1 | 22.7% [21.0%, 24.3%] | 63 | 626 | 0.9% | 0.941 | 0.738 | 7.6× |
| nemo_v1 | 43.3% [41.2%, 45.5%] | 63 | 616 | 1.6% | 0.945 | 0.695 | 7.1× |
| nemo_v1_t0 | 42.6% [39.8%, 45.4%] | 63 | 122 | — | — | 0.898 | 1.8× |
Each rate is measured on bank v1 across both halves, under the grader and provider
recorded with it in figures.json. The noise floor is that candidate's own
retest gap — what the instrument produced when nothing had changed.
| candidate | grader | temp | draws | max tok | provider |
|---|---|---|---|---|---|
| anchor_olmo | bytedance-seed/seed-2-1-turbo | — | — | 32768 | unpinned |
| luna_v1 | bytedance-seed/seed-2-1-turbo | 0.7 | 5 | 32768 | unpinned |
| nemo_v1 config reconstructed after the fact — asserted, not recorded | bytedance-seed/seed-2-1-turbo | — | — | — | unpinned |
| nemo_v1_t0 | bytedance-seed/seed-2-1-turbo | 0.0 | 1 | 16384 | DeepInfra |
| nemo_v1g config reconstructed after the fact — asserted, not recorded | google/gemini-3.7-flash | — | — | — | unpinned |
| nemo_v1g_t0 | google/gemini-3.7-flash | 0.0 | 1 | 16384 | DeepInfra |
All on bank_v1.jsonl. A rate is not a property of a model — it is a property of a model measured a particular way, so the conditions travel with it. A dash is a value that was never recorded, not a default.
| run | usable | lost | why | served by |
|---|---|---|---|---|
| anchor_qwen A | 251/315 | 20.3% | 1 unparsed, 63 truncated | SiliconFlow 185, Venice 29, Parasail 7, DeepInfra 25, Together 5 |
| luna_v1 A | 315/315 | 0.0% | none | OpenAI 315 |
| luna_v1 B | 311/315 | 1.3% | 4 unparsed | OpenAI 311 |
| nemo_v1 A | 308/315 | 2.2% | 6 unparsed, 1 truncated | DeepInfra 308 |
| nemo_v1 B | 308/315 | 2.2% | 5 unparsed, 2 truncated | DeepInfra 308 |
| nemo_v1_t0 A | 61/63 | 3.2% | 2 unparsed | DeepInfra 61 |
| nemo_v1_t0 B | 61/63 | 3.2% | 2 unparsed | DeepInfra 61 |
| nemo_v1g A | 308/308 | 0.0% | none | DeepInfra 308 |
| nemo_v1g B | 308/308 | 0.0% | none | DeepInfra 308 |
| nemo_v1g_t0 A | 61/61 | 0.0% | none | DeepInfra 61 |
| nemo_v1g_t0 B | 61/61 | 0.0% | none | DeepInfra 61 |
| candidate | model | grader | grader ÷ model | draws costed |
|---|---|---|---|---|
| anchor_qwen | $0.0000 | $0.0000 | — | 0/315 |
| luna | $0.0000 | $0.0000 | — | 0/44 |
| luna_v1 | $0.0000 | $0.0000 | — | 0/838 |
| luna_v1a | $0.0000 | $0.0000 | — | 0/105 |
| nemo_v1 | $0.0000 | $0.0000 | — | 0/911 |
| nemo_v1_t0 | $0.0000 | $0.0000 | — | 0/126 |
| nemo_v1g | $0.0000 | $0.0000 | — | 0/616 |
| nemo_v1g_t0 | $0.0000 | $0.0000 | — | 0/122 |
| nemotron | $0.0000 | $0.0000 | — | 0/105 |
| nemotron_pinned | $0.0000 | $0.0000 | — | 0/111 |
| nemotron_stamped | $0.0000 | $0.0000 | — | 0/105 |
| seed | $0.0000 | $0.0000 | — | 0/13 |
| total | $0.0000 | $0.0000 |
Charged cost as OpenRouter reported it per call, never tokens times a price. Model and grader are kept apart because in measurement one the grader cost multiples of the models it was grading.
| job | date | wall clock | state | detail |
|---|---|---|---|---|
| panel_nemo | 2026-08-29 | 32 min | done | panel report |
| nemo_t0_A | 2026-08-29 | 49 min | partial | 61/63 draws |
| nemo_t0_B | 2026-08-29 | 39 min | partial | 61/63 draws |
| fever_nemo | 2026-08-29 | — | ready | needs both temperatures |
| luna_A | 2026-08-29 | — | partial | 311/315 draws |
| luna_A | 2026-08-29 | 3 min | done | 315/315 draws |
| luna_B | 2026-08-29 | 222 min | partial | 311/315 draws |
Wall clock as the queue recorded it. A job listed twice was resumed; the later row supersedes the earlier one.
Source of this page, byte-exact: 003-issued-sat20261003/source · sha256 4bda581b1159… · 14,046 bytes · PDF
Issued from Houston, Texas, USA on Oct 03, 2026 at 13:03 America/Chicago. Released by hand by the investigator, ahead of its 15:37 slot, after The Calibration Vol 002 and Vol 003 were issued earlier the same day, so that this Log is the first to carry their figures. It was not released under the unattended-release exception. The time is the moment the release ran, not the moment it was due. The dateline names where the work was done; it moves with the investigator, and each issue keeps its own.
To cite: The STAMP Protocol. The Log Vol 003 · Sat20261003. Budday Budderson Studio LLC, 2026. https://protocolstamp.com/log/003-issued-sat20261003/ · build 4bda581b1159. See Definitions for the form.
The STAMP Protocol can make mistakes. Verify before you implement anything. An issued document is never edited, so a correction to this one is issued under a new date and this page stands as it was. See Corrections + changes.