Today › Daily report
Daily reading · 12 November 2026
Nov 12: OpenAI reports chain-of-thought controllability rose in its Astra-class models during normal training and calls its results 'sobering'. C rises to 0.4% from time passing alone.
1 · Is AGI here?
Not in public: 0 of 8 parts met (2 partial, 6 far).
AGI anywhere, public or hidden: 0.9% today
- Breadth, status Far, step 1 of 4. On real freelance projects the best system delivers 41.30% (Fable 5.2, Nov 9, pending our weekly check), against 80%; results by occupation for the other half are unpublished.
- Quality, status Far, step 1 of 4. Wins or ties professionals in 84.9% of comparisons on well-specified tasks (overall only); on real projects 41.30% (pending our weekly check), against 50%.
- Reliability, status Far, step 1 of 4. On a hard agent test the best model gets 35.1% of tasks right on all four tries, against 80%.
- Reasoning, status Partial, step 2 of 4. Expert exams are passed (GPQA Diamond 96%); new research results by AI alone: 4 of the 10 we ask for.
- Long projects, status Far, step 1 of 4. The best measured 80% time horizon is about 3.1 hours, some 54 times short of a work-month.
- Autonomy, status Far, step 1 of 4. Hands-off on few-minute tasks (86%); on 1.6-hour workflows 44.3% strict, against 80%.
- Learning, status Far, step 1 of 4. Captures about a quarter of the available gain from experience (25.4%, against 50%); keeping it is unmeasured.
- Generalization, status Partial, step 2 of 4. At first contact with a new kind of game the best score was 1.5%, against 80%; 62.7% came only after 5 months.
3 new results pending our weekly check: reliability, autonomy, long projects.
2 · Could it be hidden?
3% Hidden AGI Index, no change
Above AGI anywhere because 2.15 points are hidden self-improvement (B), which needs no AGI.
3 · Should I worry today? Separate: set by published rules, never by the odds aboveRules, not odds
Fire alarm: Level 1 · Watch since Sep 29 · W2, W3, W4, W5 met
Quiet day
OpenAI: reasoning may be getting harder to monitor
The rise in chain-of-thought controllability that OpenAI reports persisted when it trained its latest Astra-class model with only length limits and final-answer grading. UK AISI's Sept 3 tests put Astra's no-CoT time horizon on competition math (50% reliability) at 30.9 minutes, against 3.6 for GPT 5.6 Sol (UK AISI, via OpenAI). It is a warning sign already in our numbers, so C's share holds. verified fact (part our inference)OpenAI's alignment blog (Nov 10)
Probabilities
| Hypothesis | Today | By end-2030 | By end-2035 | Change today | Confidence |
|---|---|---|---|---|---|
| AGI anywhere | 0.9% | 36% | 59% | no change | low–medium |
| A · Hidden AGI | 0.6% | 20% | 32% | no change | low–medium |
| B · Hidden self-improvement | 2.5% | 25% | 34% | no change | low |
| C · Covert AGI actor | 0.4% | 8.5% | 14% | ▲ +0.1 pts | low |
| D · Covert government influence | 0.2% | 3% | 7% | no change | low |
| D-open · Open government influence | 0.2% | 24% | 46% | no change | low |
How the "today" numbers are derived: analysis, Step 5 →
Signals
4 confirmed (count toward Watch) · 6 open · 1 quiet. No signal changed today. All signals →
Escape watch: 0 of 8 confirmed · 8 open · no status changed; new evidence on 8 (weights, selfrep, compute, money, identity, supplychain, coordination, leakage). Escape watch →
Capability gap ~3 months · AI-led R&D 26% · Disclosure lag ~112 days (111–120) floor · Big Tech capex $615B / yr · Delegation rung 3 of 5. No gauge moved.
Is AGI here, today?
No public system meets every part's bar at once; today's leads are an expert-built coding benchmark, AWS's agent-written emulator and Anthropic's claimed three-day AMD port. Full reasoning →
Top stories
- Nov 10 OpenAI says controllability, which can make chains of thought harder to monitor, rose in Astra-class models during normal training. verified factOpenAI · more →
- Nov 11 – Nov 12 The Epoch Capabilities Index adds Gemini 4 Carbon at 169.76, 3.68 below a re-scored Claude Fable 5.2. verified fact (part our inference)Epoch; Google · more →
- Nov 10 Anthropic's founders' letter, filed with the SEC, says safety may require slowing capability growth, 'especially when this is driven by AI systems building themselves'. verified fact (part our inference)Anthropic (SEC FWP) · more →
- Nov 10 – Nov 12 Hijacked GitHub accounts planted a CI secret-stealer in thousands of repositories, per Aikido and SafeDep; neither blames an AI agent. verified fact (part our inference)Aikido; SafeDep · more →
- Nov 11 Nvidia's board accepted Jensen Huang's resignation as CEO on Nov 11; CFO Colette Kress is interim CEO. The 8-K gives no reason. verified fact (part our inference)Nvidia (SEC 8-K) · more →
Dates to watch
- Nov 15Our forecast: Anthropic's registration statement on EDGAR (filed Oct 26; we said 65%)
- Nov 15Our forecast: Big Four July–September capex (filed Oct 22–30; we said 55%)
- Dec 13W6 is met unless OpenAI's post-Astra model is released or its capabilities published
Reasoning that gets harder to read makes outside testing matter more, and outside testers' access to new models remains restricted on our record. What would change our mind: an outside evaluator documenting a frontier model evading its developer's monitors in use, or OpenAI's post-Astra model passing Dec 13 unreleased and unevaluated.