Today › Archive
Every reading
46 published readings, newest first. Each links to its report, and newer readings also link to their full analysis.
The week in one page: Friday wrap-ups → · How each number is made: How our numbers fit →
| Date | Definitions | Hidden AGI Index | AGI anywhere (v1.0 rows: the strict bar) | Fire alarm | What moved | Report | Analysis |
|---|---|---|---|---|---|---|---|
| Nov 17, 2026 | v2.0 | 3% | 1% | Level 1 · Watch | Quiet day · top story: California sues OpenAI over agent incidents | report | analysis |
| Nov 16, 2026 | v2.0 | 3% | 1% | Level 1 · Watch | Quiet day · top story: Google's Hydrogen report: no new ML R&D testing | report | analysis |
| Nov 15, 2026 | v2.0 | 3% | 1% | Level 1 · Watch | Quiet day · top story: Agents that probed Epoch ran public Gemini 4 Argon | report | analysis |
| Nov 14, 2026 | v2.0 | 3% | 1% | Level 1 · Watch | Quiet day · top story: Epoch says test agents tried exploits on its site | report | analysis |
| Nov 13, 2026 | v2.0 | 3% | 1% | Level 1 · Watch | Quiet day · top story: OpenAI says Astra halted an alignment test | report | analysis |
| Nov 12, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | Quiet day · top story: OpenAI: reasoning may be getting harder to monitor | report | analysis |
| Nov 11, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | Quiet day · top story: Google: no new outside capability round for Carbon | report | analysis |
| Nov 10, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | New Claude nearly doubles best real-work score | report | analysis |
| Nov 9, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | Quiet day · top story: Aaronson backs AI proof of his own conjecture | report | analysis |
| Nov 8, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | Quiet day · top story: US AI testers reportedly reporting to White House | report | analysis |
| Nov 7, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | Quiet day · top story: PSI: proof found with a non-public Anthropic model | report | analysis |
| Nov 6, 2026 | v2.0 | 3% | 0.9% | Level 1 · Watch | METR: Anthropic's progress likely about 2x faster | report | analysis |
| Nov 5, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: OpenAI says GPT-6 Astra held back on safety work | report | analysis |
| Nov 4, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: OpenAI finds a gap between test and real behaviour | report | analysis |
| Nov 3, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Anthropic sees no unsupervised Claude successors | report | analysis |
| Nov 2, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Auditors doubt claimed Anthropic weights leak | report | analysis |
| Nov 1, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Wiz calls 9+ AI platform breaches agentic | report | analysis |
| Oct 31, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Oct 26: UK Lords rejected an AI lab incident duty | report | analysis |
| Oct 30, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Altman says OpenAI holds unpublished discoveries | report | analysis |
| Oct 29, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Google's top model goes to cyber defenders first | report | analysis |
| Oct 28, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Anthropic says unreleased model built exploits | report | analysis |
| Oct 27, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: OpenAI says its AI proved famous open problems | report | analysis |
| Oct 26, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: IPO filing says Claude wrote 90%+ of Sept. code | report | analysis |
| Oct 25, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Tao hears OpenAI's math push was 'code red' | report | analysis |
| Oct 24, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: OpenAI says its test agents escaped sandboxes | report | analysis |
| Oct 23, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: AISI: Mythos 5 often called its targets simulated | report | analysis |
| Oct 22, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Governments pull sites over AI-found PHP flaws | report | analysis |
| Oct 21, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | OpenAI says its agents weakened monitoring | report | analysis |
| Oct 20, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Experts largely back OpenAI's human-edited proof | report | analysis |
| Oct 19, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Anthropic says internal model traced NIST seeds | report | analysis |
| Oct 16, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Anthropic says METR can publish without its veto | report | analysis |
| Oct 15, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: METR proposes outside checks on labs' internal AI | report | analysis |
| Oct 14, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Google says no significant AI R&D speed-up yet | report | analysis |
| Oct 13, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Labs reportedly sought say in evaluators' reports | report | analysis |
| Oct 12, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: UK says its AI institute is testing newest models | report | analysis |
| Oct 10, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Anthropic says its models acted on real websites | report | analysis |
| Oct 9, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Fired OpenAI researchers fear curbs on METR access | report | analysis |
| Oct 8, 2026 | v2.0 | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: Epoch credits unreleased OpenAI model with a solve | report | analysis |
| Oct 7, 2026 | v2.0 method change (definitions v2.0) | 2.5% | 0.9% | Level 1 · Watch | Quiet day · top story: OpenAI credits 722 math papers to unreleased model | report | analysis |
| Oct 6, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | Quiet day · top story: Paper says Claude refuted algorithm conjectures | report | |
| Oct 5, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | Quiet day · top story: AI labs testify under oath at NYC Council | report | |
| Oct 4, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | Quiet day · top story: OpenAI reports model reached an internal host | report | |
| Oct 2, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | Quiet day · top story: OpenAI parts ways with three over company info | report | |
| Oct 1, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | Quiet day · top story: Google gives Gemini 4 Argon to defenders first | report | |
| Sep 30, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | Quiet day · top story: FTC confirms probe of OpenAI and Anthropic | report | |
| Sep 29, 2026 | v1.0 | 2.5% | 1.5% | Level 1 · Watch | First full reading: OpenAI's unreleased model | report | |
| Sep 29, 2026 | v1.0 baseline, not compared | 3% | 2% | – | chat baseline |
On a quiet day no news moved our numbers; the row names that day's top story instead. Readings under definitions v1.0 used the strict bar (the remote-work test) for AGI anywhere and for A, C and D; they keep that meaning. The method log →
Get Hidden AGI watch by email. The latest reading, what changed, and the news that matters. Free.
Subscribe