AI will be 'smarter than the smartest human' by 2026, driven by fully automated recursive self-improvement loops compounding cognitive gains without human intervention.
Predictor: Elon Musk
Prediction text
AI will be 'smarter than the smartest human' by 2026, driven by fully automated recursive self-improvement loops compounding cognitive gains without human intervention. | Frontier-model generalist benchmark sweep
Key catalyst: Frontier-model generalist benchmark sweep
Watch events: Grok 5 / GPT-6 benchmark disclosures; recursive-self-improvement demos
Resolution evidence
xAI Grok 4.2, OpenAI o-series, Claude Opus 4.7 all match or exceed expert humans on narrow benchmarks (IMO, USAMO, FrontierMath, GDPval); generalist-single-human-parity more contested.
Predictor: Elon Musk
Calibration plot (stated vs observed)
Evidence about this node from Elon Musk is multiplied by κ in /api/intake. Lower κ = less weight; floors at 0.10 (effectively silenced) and caps at 1.00 (full weight).
Reference class
This node isn't linked to a reference class. The Bayesian update applies without outside-view blending.
Probability over time
Milestone chain
- 2026-02-22overdueQ1 window check-in (25%)
- 2026-03-05hitMajor model release exceeds human OSWorld baseline by clear marginHow: Frontier AI model exceeds 72.4% human-expert OSWorld baseline by >=2pp, signaling superhuman computer-use capabilitySource: https://nerdleveltech.com/gpt-5-4-beats-humans-computer-use-ai-agents — GPT-5.4 hits 75.0%conf 95%
- 2026-04-16overdueQ2 window check-in (50%)
- 2026-04-29hitHolo3-35B-A3B leads OSWorld with 82.6%How: OSWorld-Verified leaderboard shows top model at >=82% accuracy, exceeding human baseline by >=10ppSource: https://benchlm.ai/benchmarks/osWorldVerified — Holo3-35B-A3B 82.6%conf 92%
- 2026-06-07pendingQ3 window check-in (75%)
- 2026-06-30pendingAnthropic CEO publicly maintains 2026 AI > most humans timelineHow: Dario Amodei reaffirms in published interview/essay that AI surpasses human intelligence in most domains by end-2026 or early-2027Source: https://www.bloomberg.com/news/newsletters/2024-10-18/anthropic-ceo-thinks-ai-may-outsmart-most-humans-as-soon-as-2026 — Amodei 2026 timelineconf 80%
- 2026-05-01 → 2026-09-30pendingOpenAI articulates 'hundreds of thousands of automated research interns' planHow: OpenAI executive publicly outlines path to 100K+ automated research agents within 9 months, indicating material progress on RSI loopSource: https://openai.com/index/next-phase-of-enterprise-ai/ — OpenAI roadmapconf 70%
- 2026-09-01 → 2026-11-30pendingFrontier benchmark composite (MMLU/SWE/MATH/OSWorld) exceeds expert humans on 4 of 4How: At least one frontier model exceeds expert-human baseline on 4 of 4 standard capability benchmarks (MMLU, SWE-Bench Verified, MATH, OSWorld)Source: https://hai.stanford.edu/ai-index/2026-ai-index-report/technical-performance — Stanford AI Index 2026conf 60%
- 2026-10-01 → 2027-06-30pendingCascade: AGI-popular-narrative drives capex acceleration past $700B/yrHow: Hyperscaler combined annual AI capex exceeds $700B run-rate as direct response to claimed AGI/ASI capability milestonesSource: https://aijourn.com/700-billion-ai-capex-in-2026-following-the-capital-flows-from-hyperscalers-to-chipmakers/ — $700B 2026 capexconf 65%
What if this resolves?
Click a button to clamp this prediction and run a Gibbs sample. Returns the predictions whose marginals shift most. ~30s per run; ideal for stress-testing "if X resolves, what else moves?"
Evidence chain
Raw metadata
{
"trf": 0.6338685427014515,
"kappa": 0.6429,
"base_rate": null,
"predictor": "Elon Musk",
"total_llr": -0.8109302162163288,
"grace_days": 7,
"bayesian_v2": true,
"prior_logit": -1.0986122886681098,
"bayes_factor": "1.7:1 against",
"blend_reason": "no reference_class linked",
"inside_prior": 0.25,
"kappa_source": "predictor_table",
"n_milestones": 2,
"blend_applied": false,
"contributions": [
{
"llr": -0.4054651081081644,
"kind": "quartile_checkpoint",
"kappa": 0.6429,
"label": "Q1 window check-in (25%)",
"weight": 0.05,
"strength": "weak",
"confidence": null,
"source_url": null,
"adjusted_llr": -0.2606735180027389,
"expected_date": "2026-02-22",
"measurement_criterion": null
},
{
"llr": -0.4054651081081644,
"kind": "quartile_checkpoint",
"kappa": 0.6429,
"label": "Q2 window check-in (50%)",
"weight": 0.05,
"strength": "weak",
"confidence": null,
"source_url": null,
"adjusted_llr": -0.2606735180027389,
"expected_date": "2026-04-16",
"measurement_criterion": null
}
],
"evidence_kind": "metadata_milestone_miss_sweep",
"inside_source": "prior_prob",
"inside_weight": 0.5562920201089838,
"outside_weight": 0.4437079798910162,
"posterior_prob": 0.1652104798916139,
"posterior_logit": -1.6199593246735877,
"predictor_brier": 0.01,
"inside_posterior": 0.1652104798916139,
"blended_posterior": 0.1652104798916139,
"reference_class_id": null,
"total_adjusted_llr": -0.5213470360054778,
"predictor_n_resolved": 2
}Network propagation neighbors
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.
Ticker exposure
Beneficiaries (1)
Prerequisites (1)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| correlate | S_AI_PAUSE_2026 | Major-country AI pause beginning 2026 | ai_regulatory_pause | — |
Dependents (0)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| No dependents | ||||
Linked documents (10)
Raw metadata
{
"nia": false,
"qty": "smartest-human-parity",
"mode": "FORECAST",
"role": "Cited-CEO",
"context": "Distinct from INF_073 (AI smarter than all humanity combined by 2030/2031) — this is the lower-bound 2026 prediction that any-individual-human-parity is reached first. Musk has repeatedly revised this forward from 2028 (2022) to 2026 (2024).",
"to_year": 2026,
"conv_cues": "CEO FIRST_PERSON; specific year; superlative",
"direction": "HAPPEN",
"from_year": 2026,
"timeframe": "2026",
"conv_level": "HIGH",
"milestones": [
{
"kind": "quartile_checkpoint",
"label": "Q1 window check-in (25%)",
"status": "overdue",
"weight": 0.05,
"ordinal": -7,
"source_id": null,
"expected_date": "2026-02-22",
"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": "Major model release exceeds human OSWorld baseline by clear margin",
"source": "https://nerdleveltech.com/gpt-5-4-beats-humans-computer-use-ai-agents — GPT-5.4 hits 75.0%",
"status": "hit",
"weight": 0.4,
"ordinal": -6,
"source_id": null,
"confidence": 0.95,
"source_url": "https://nerdleveltech.com/gpt-5-4-beats-humans-computer-use-ai-agents",
"expected_date": "2026-03-05",
"observed_date": "2026-03-05",
"research_origin": "deep_research",
"measurement_criterion": "Frontier AI model exceeds 72.4% human-expert OSWorld baseline by >=2pp, signaling superhuman computer-use capability"
},
{
"kind": "quartile_checkpoint",
"label": "Q2 window check-in (50%)",
"status": "overdue",
"weight": 0.05,
"ordinal": -5,
"source_id": null,
"expected_date": "2026-04-16",
"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": "Holo3-35B-A3B leads OSWorld with 82.6%",
"source": "https://benchlm.ai/benchmarks/osWorldVerified — Holo3-35B-A3B 82.6%",
"status": "hit",
"weight": 0.4,
"ordinal": -4,
"source_id": null,
"confidence": 0.92,
"source_url": "https://benchlm.ai/benchmarks/osWorldVerified",
"expected_date": "2026-04-29",
"observed_date": "2026-04-29",
"research_origin": "deep_research",
"measurement_criterion": "OSWorld-Verified leaderboard shows top model at >=82% accuracy, exceeding human baseline by >=10pp"
},
{
"kind": "quartile_checkpoint",
"label": "Q3 window check-in (75%)",
"status": "pending",
"weight": 0.05,
"ordinal": -3,
"source_id": null,
"expected_date": "2026-06-07",
"observed_date": null
},
{
"kind": "llm_pre_event",
"label": "Anthropic CEO publicly maintains 2026 AI > most humans timeline",
"source": "https://www.bloomberg.com/news/newsletters/2024-10-18/anthropic-ceo-thinks-ai-may-outsmart-most-humans-as-soon-as-2026 — Amodei 2026 timeline",
"status": "pending",
"weight": 0.4,
"ordinal": -2,
"source_id": null,
"confidence": 0.8,
"source_url": "https://www.bloomberg.com/news/newsletters/2024-10-18/anthropic-ceo-thinks-ai-may-outsmart-most-humans-as-soon-as-2026",
"expected_date": "2026-06-30",
"research_origin": "deep_research",
"measurement_criterion": "Dario Amodei reaffirms in published interview/essay that AI surpasses human intelligence in most domains by end-2026 or early-2027"
},
{
"kind": "llm_pre_event",
"label": "OpenAI articulates 'hundreds of thousands of automated research interns' plan",
"source": "https://openai.com/index/next-phase-of-enterprise-ai/ — OpenAI roadmap",
"status": "pending",
"weight": 0.4,
"ordinal": -1,
"source_id": null,
"confidence": 0.
... (truncated)