AI will become smarter than any single human by end of 2025 or 2026, and smarter than all of humanity combined by 2030 or 2031.

AI will become smarter than any single human by end of 2025 or 2026, and smarter than all of humanity combined by 2030 or 2031. | Grok 5 or Gemini 4 capability evaluation

Key catalyst: Grok 5 or Gemini 4 capability evaluation

Watch events: Grok 5+ capability disclosures; frontier benchmark ensembles

xAI Grok 4.2, Colossus 200K-GPU training runs. Frontier models approach but do not surpass human-cognition benchmarks in generalist tasks as of Apr 2026.

No propagation data yet. Run inference/.venv/bin/python scripts/ops/run_loopy_belief_propagation.py on the droplet, or wait for the Sunday 02:00 UTC weekly cron.

{
  "nia": false,
  "qty": "smarter than humanity",
  "mode": "FORECAST",
  "role": "Cited-CEO",
  "context": "Musk aggressive-timing thesis. Pairs with INF_071 (Kurzweil 2029 AGI) as upper bound of elite-futurist consensus. Forcing function for trillion-dollar energy buildout.",
  "to_year": 2031,
  "conv_cues": "specific years; CEO FIRST_PERSON; superlative framing",
  "direction": "HAPPEN",
  "from_year": 2025,
  "timeframe": "2025-2031",
  "conv_level": "HIGH",
  "milestones": [
    {
      "kind": "llm_pre_event",
      "label": "OpenAI o3 surpasses 87.5% on ARC-AGI (above human baseline)",
      "notes": "First clear single-domain superhuman benchmark. Necessary but not sufficient for 'smarter than any single human'.",
      "source": "https://www.veracalloway.com/blog/ai-culture/agi-timeline/ — o3 hit 87.5% Dec 2024",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -9,
      "source_id": null,
      "confidence": 0.99,
      "source_url": "https://www.veracalloway.com/blog/ai-culture/agi-timeline/",
      "expected_date": "2024-12-20",
      "observed_date": "2024-12-20",
      "research_origin": "deep_research",
      "measurement_criterion": "OpenAI o3 publishes ARC-AGI score 85% or higher (human baseline) on official leaderboard"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q1 window check-in (25%)",
      "status": "overdue",
      "weight": 0.05,
      "ordinal": -8,
      "source_id": null,
      "expected_date": "2026-02-23",
      "observed_date": null,
      "miss_emitted_at": "2026-05-02T22:07:21.384228+00:00",
      "miss_emitted_by": "metadata_milestone_sweep"
    },
    {
      "kind": "llm_pre_event",
      "label": "GPT-5.5 (Spud) completes pretraining (March 2026)",
      "source": "https://www.veracalloway.com/blog/ai-culture/agi-timeline/",
      "status": "hit",
      "weight": 0.4,
      "ordinal": -7,
      "source_id": null,
      "confidence": 0.85,
      "source_url": "https://www.veracalloway.com/blog/ai-culture/agi-timeline/",
      "expected_date": "2026-03-24",
      "observed_date": "2026-03-24",
      "research_origin": "deep_research",
      "measurement_criterion": "OpenAI internal communications / leaked reporting confirms GPT-5.5 pretraining completion in March 2026"
    },
    {
      "kind": "llm_pre_event",
      "label": "Frontier model passes adversarial Turing test (high-quality, 2-hour panel)",
      "source": "Manifold market reference at https://aimultiple.com/artificial-general-intelligence-singularity-timing",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -6,
      "source_id": null,
      "confidence": 0.45,
      "source_url": "https://aimultiple.com/artificial-general-intelligence-singularity-timing",
      "expected_date": "2027-03-17",
      "research_origin": "deep_research",
      "expected_date_range": {
        "to": "2027-12-31",
        "from": "2026-06-01"
      },
      "measurement_criterion": "Independent academic study (Stanford/MIT/Oxford) confirms frontier model passes 2-hour adversarial Turing test with expert judges and 50% pass rate or higher"
    },
    {
      "kind": "llm_pre_event",
      "label": "AI matches or exceeds human expert performance across 5+ professional domains simultaneously",
      "source": "Stanford AI Index annual report",
      "status": "pending",
      "weight": 0.4,
      "ordinal": -5,
      "source_id": null,
      "confidence": 0.5,
      "expected_date": "2027-03-17",
      "research_origin": "training",
      "expected_date_range": {
        "to": "2027-12-31",
        "from": "2026-06-01"
      },
      "measurement_criterion": "Independent study (Apollo/METR/AI Index) shows frontier model matches expert humans in 5+ distinct professional benchmarks (medical diagnosis, legal reasoning, programming, mathematical proof, scientific writing)"
    },
    {
      "kind": "quartile_checkpoint",
      "label": "Q2 window check-in (50%)",
      "status": "pending",
      "weight"
... (truncated)