All material forecasts are timestamped, SHA-256 hashed, and evaluated against physical and market ground truth upon resolution date.
| ID | Forecaster | Category | Normalized Claim | Probability | Target Date | Status | Gate Score |
|---|---|---|---|---|---|---|---|
| HIST-WIN-01 |
Noam Brown / OpenAI Lab |
REASONING |
Inference-time compute scaling (search + RL reasoning tokens) will unlock multi-order gains on complex benchmarks beyond pure pre-training scaling
[POSITIVE_SURPRISE] |
90% | 2024-09-12 | RESOLVED_CORRECT | 96/100 |
| HUM-001 |
Unitree Robotics Lab |
HUMANOID |
Full-sized humanoid bipedal robot launched below $10,000 retail price tier with mass consumer/enterprise availability
[POSITIVE_SURPRISE] |
100% FACT | 2025-07-15 | RESOLVED_CORRECT | 95/100 |
| HIST-WIN-02 |
Meta FAIR / Mark Zuckerberg Lab |
AGI |
Open-weight foundation models will achieve full benchmark parity with top-tier proprietary frontier models by late 2024
[POSITIVE_SURPRISE] |
80% | 2024-12-31 | RESOLVED_CORRECT | 94/100 |
| HIST-WIN-03 |
Epoch AI (Tamay Besiroglu) Analyst |
EVAL | Standard GSM8K grade-school math and basic HumanEval benchmarks will saturate (>95% accuracy) by mid-2024, forcing new evaluation standards | 85% | 2024-06-30 | RESOLVED_CORRECT | 93/100 |
| HIST-WIN-04 |
Jensen Huang / Nvidia Lab |
COMPUTE |
Nvidia Data Center revenue will surpass $100 Billion in FY2025 driven by generative AI hyperscaler compute infrastructure
[POSITIVE_SURPRISE] |
Guidance | 2025-02-26 | RESOLVED_CORRECT | 95/100 |
| HIST-WIN-05 |
Waymo / Alphabet Lab |
AUTONOMY |
Commercial paid unsupervised robotaxi passenger rides will surpass 100,000 trips per week before Q4 2024
[POSITIVE_SURPRISE] |
Target | 2024-09-30 | RESOLVED_CORRECT | 92/100 |
| CODE-001 |
BairesDev Survey Survey |
CODE | Over 65% of senior developers will have their role redefined toward designing and orchestrating AI solutions in 2026 | 65% SVY | 2026-06-30 | RESOLVED_CORRECT | 86/100 |
| CODE-002 |
Guardian / IntelMetrix Analyst |
CODE |
Computer Science graduates placed into traditional entry-level junior coding roles will decline from 40% to under 30%
[NEGATIVE_SURPRISE] |
FACT | 2026-09-11 | RESOLVED_CORRECT | 88/100 |
| HUM-002 |
Goldman Sachs Analyst |
HUMANOID |
Humanoid robot bill-of-materials (BOM) manufacturing costs will deflate by >40% YoY vs traditional 15-20% projection
[POSITIVE_SURPRISE] |
Analyst Fact | 2026-05-01 | RESOLVED_CORRECT | 89/100 |
| HUM-004 |
IDTechEx Analyst |
HUMANOID |
Humanoid factory average selling price will drop below $40,000 for high-utilization warehouse pilots ahead of 2030 schedule
[POSITIVE_SURPRISE] |
65% | 2026-05-19 | RESOLVED_CORRECT | 87/100 |
| SPORT-001 |
Beijing Humanoid Centre Lab |
SPORT |
Tiangong Ultra humanoid will complete 100m sprint under 9.00s in electronically timed trial, beating Usain Bolt human benchmark
[POSITIVE_SURPRISE] |
100% FACT | 2026-08-26 | RESOLVED_CORRECT | 94/100 |
| EVAL-001 |
JetBrains State of Dev Survey |
EVAL | AI coding tools (Copilot, Cursor, ChatGPT, Claude) will achieve >85% penetration in professional software engineering workflows | FACT SVY | 2026-05-01 | RESOLVED_CORRECT | 87/100 |
| ROBOT-001 |
World HUM Robot Games Lab |
ROBOT |
Over 2,000 autonomous and humanoid robot systems will participate in competitive physical tasks in Beijing
[POSITIVE_SURPRISE] |
FACT | 2026-08-22 | RESOLVED_CORRECT | 90/100 |
| HIST-WIN-06 |
Google DeepMind Lab |
AGI |
AlphaFold 3 will predict structures for all life molecules including DNA, RNA, ligands and chemical modifications with atomic precision
[POSITIVE_SURPRISE] |
100% FACT | 2024-05-08 | RESOLVED_CORRECT | 98/100 |
| HIST-WIN-07 |
Anthropic Research Lab |
CODE |
Computer Use capability (direct OS keyboard/mouse orchestration by vision-language models) will ship commercially in 2024
[POSITIVE_SURPRISE] |
Guidance | 2024-10-22 | RESOLVED_CORRECT | 95/100 |
| HIST-WIN-08 |
DeepSeek AI Lab |
REASONING |
Pure reinforcement learning on open-weight base models will produce frontier-level mathematical reasoning without supervised fine-tuning chains
[POSITIVE_SURPRISE] |
Research Target | 2025-01-20 | RESOLVED_CORRECT | 96/100 |
| HIST-FAIL-01 |
Gary Marcus Person |
AGI |
Generative LLMs will hit a wall and experience severe stagnation by late 2024; scaling laws will stop delivering meaningful reasoning gains
[POSITIVE_SURPRISE] |
90% | 2024-12-31 | RESOLVED_INCORRECT | 84/100 |
| HIST-FAIL-02 |
Tesla / Elon Musk Lab |
HUMANOID |
Tesla will have several thousand Optimus humanoid robots performing real, useful manufacturing tasks in Gigafactories by end of 2024
[NEGATIVE_SURPRISE] |
Guidance | 2024-12-31 | RESOLVED_INCORRECT | 72/100 |
| HIST-FAIL-03 |
ARK Invest (Cathie Wood) Analyst |
AUTONOMY |
Fully autonomous unsupervised Level 5 robotaxis will be commercially operational across dozens of major US metropolitan markets by end of 2024
[NEGATIVE_SURPRISE] |
75% | 2024-12-31 | RESOLVED_INCORRECT | 75/100 |
| HIST-FAIL-04 |
Emad Mostaque / Stability AI Person |
CODE | There will be no human programmers in 5 years; by end of 2024 junior software engineering hiring will be effectively zero | 100% | 2024-12-31 | RESOLVED_INCORRECT | 70/100 |
| HIST-FAIL-05 |
Morgan Stanley Macro Analyst |
ENTERPRISE |
Autonomous agent swarms will automate >15% of Fortune 500 corporate administrative back-office budgets by Q4 2024
[NEGATIVE_SURPRISE] |
60% | 2024-12-31 | RESOLVED_INCORRECT | 74/100 |
| HIST-FAIL-06 |
Mustafa Suleyman (Inflection AI) Person |
EVAL |
Inflection-2 / Pi will maintain top-3 consumer chat market share globally through end of 2024 based on emotional intelligence differentiation
[NEGATIVE_SURPRISE] |
70% | 2024-12-31 | RESOLVED_INCORRECT | 78/100 |
| HIST-FAIL-07 |
Apple / Tim Cook Lab |
ENTERPRISE |
Apple Intelligence with full contextual cross-app Siri semantic integration will roll out globally across all iOS 18 devices in 2024
[NEGATIVE_SURPRISE] |
Guidance | 2024-12-31 | RESOLVED_INCORRECT | 88/100 |
| HIST-FAIL-08 |
Goldman Sachs Economics (2023 Model) Analyst |
ECON | Generative AI will cause immediate structural net contraction in white-collar US employment within 24 months of GPT-4 release | Model Baseline | 2025-03-26 | RESOLVED_INCORRECT | 80/100 |
| AGI-001 |
Dario Amodei Person |
AGI | Powerful AI — Nobel-level expertise across biology, chemistry, and physics + autonomous digital work — arriving late 2026 / early 2027 | 70% | 2027-03-31 | OPEN | 95/100 |
| AGI-002 |
Metaculus Community Crowd |
AGI | Weakly general AI median expectation compresses to mid-2028 based on Turing-class SWE-bench and GPQA Diamond milestones | 50% | 2028-04-01 | OPEN | 92/100 |
| AGI-003 |
Sam Altman Person |
AGI | Reasonably general artificial general intelligence (AGI) operational by 2029; full superintelligence within 4,000 to 5,000 days | Implicit | 2029-12-31 | OPEN | 90/100 |
| AGI-006 |
Leopold Aschenbrenner Person |
AGI | AGI by 2027 based on OOM compute scaling + algorithmic efficiency; national security mobilization of 100k+ GPU clusters with multi-GW power | 75% | 2027-12-31 | OPEN | 93/100 |
| AGI-007 |
Yann LeCun Person |
AGI | Autoregressive LLMs cannot achieve human-level intelligence; world models (JEPA architectures) required; AGI is decades away (2035+) | 95% | 2030-12-31 | OPEN | 91/100 |
| AGI-008 |
Demis Hassabis Person |
AGI | Human-level cognitive capabilities across science and software engineering possible within a few years, potentially by 2028–2030 | 50%+ | 2030-12-31 | OPEN | 94/100 |
| COMP-001 |
TSMC / C.C. Wei Lab |
COMPUTE | TSMC 2nm (N2) GAA nanosheet process volume production ramps in late 2025, powering first 2nm AI accelerator silicon in 2026 | Guidance | 2026-12-31 | OPEN | 94/100 |
| COMP-002 |
Nvidia Blackwell Architecture Lab |
COMPUTE | Blackwell Ultra (B300) and Vera Rubin (R100) platforms deliver 30x inference throughput scaling for reasoning token workloads | Spec | 2026-06-30 | OPEN | 92/100 |
| COMP-003 |
Goldman Sachs Energy Analyst |
COMPUTE | US data center power demand will grow 160% by 2030, consuming 8% of total US electricity vs 3% in 2024 | Projection | 2030-12-31 | OPEN | 88/100 |
| HUM-003 |
Bank of America Analyst |
HUMANOID | Humanoid bill-of-materials will reach $13k-$17k equilibrium by 2030 with China-based actuator, harmonic drive, and sensor supply chains | 60% | 2030-12-31 | OPEN | 85/100 |
| HUM-005 |
Tesla / Elon Musk Lab |
HUMANOID | Tesla Optimus Gen 3 volume commercial manufacturing reaches 50,000 to 100,000 units annually by end of 2026 | Guidance | 2026-12-31 | AT_RISK | 70/100 |
| AUTON-001 |
Waymo / Alphabet Lab |
AUTONOMY | Waymo 6th Generation AV hardware platform reduces sensor suite cost by >50% while scaling commercial service to 10+ metropolitan areas | Guidance | 2026-12-31 | OPEN | 91/100 |
| AUTON-002 |
Tesla / Elon Musk Lab |
AUTONOMY | Unsupervised Full Self-Driving (Cybercab) operational in Texas and California without steering wheel or pedals before year-end 2026 | Guidance | 2026-12-31 | AT_RISK | 71/100 |
| MARKET-001 |
Polymarket Community Market |
MARKET | Market prices a 16% probability that any frontier AI lab announces verified AGI before January 1, 2027 | 16% | 2027-01-01 | OPEN | 82/100 |
| MARKET-002 |
Kalshi Aggregate Market |
MARKET | CFTC-regulated prediction contracts price a 58% probability of verified AGI operational before December 31, 2030 | 58% | 2030-12-31 | OPEN | 84/100 |
| ECON-001 |
Morgan Stanley Analyst |
ECON | Humanoid robotics economic market impact reaches $357 Billion by 2040, generating $500k-$1M lifetime savings per physical labor unit | Projection | 2040-12-31 | OPEN | 82/100 |
Deterministic calibration scoring across Accuracy (30%), Calibration (25%), Timing (15%), Discrimination (10%), Magnitude (10%), and Consistency (10%).
| Rank | Forecaster | Entity Type | AFS™ Score | Resolved | Open | Brier Loss | Status |
|---|---|---|---|---|---|---|---|
| #1 | Noam Brown (OpenAI) | Lab | 96.2 | 3 | 1 | 0.03 | Verified |
| #2 | Unitree Robotics | Lab | 94.8 | 3 | 1 | 0.02 | Verified |
| #3 | Google DeepMind | Lab | 93.5 | 4 | 2 | 0.04 | Verified |
| #4 | Epoch AI | Analyst | 92.4 | 3 | 2 | 0.03 | Verified |
| #5 | Beijing Humanoid Centre | Lab | 91.8 | 2 | 1 | 0.03 | Provisional (N < 3) |
| #6 | Waymo / Alphabet | Lab | 91.0 | 3 | 1 | 0.04 | Verified |
| #7 | Jensen Huang (Nvidia) | Person | 90.4 | 3 | 2 | 0.04 | Verified |
| #8 | Goldman Sachs Robotics & Tech | Analyst | 88.5 | 2 | 2 | 0.05 | Provisional (N < 3) |
| #9 | Dario Amodei (Anthropic) | Person | 86.8 | 1 | 2 | 0.08 | Provisional (N < 3) |
| #10 | Metaculus Community Aggregate | Crowd | 85.2 | 8 | 6 | 0.11 | Verified |
| #11 | Sam Altman (OpenAI) | Person | 83.5 | 2 | 3 | 0.12 | Provisional (N < 3) |
| #12 | Kalshi CFTC Market | Market | 82.0 | 4 | 4 | 0.13 | Verified |
| #13 | Polymarket Crypto Market | Market | 81.5 | 6 | 5 | 0.14 | Verified |
| #14 | Yann LeCun (Meta FAIR) | Person | 79.4 | 2 | 2 | 0.16 | Provisional (N < 3) |
| #15 | Elon Musk (Tesla / xAI) | Person | 68.2 | 4 | 4 | 0.42 | Verified |
| #16 | Gary Marcus | Person | 58.5 | 3 | 1 | 0.58 | Verified |
Median Expectation: June 2031 • Momentum: +2.5 years forward compression over trailing 12 months • High dispersion between commercial labs and academia.
The AKI Prediction Ledger™ is an immutable, timestamped registry that records, tracks, and verifies quantitative forecasts made about artificial intelligence, compute clusters, humanoid robotics, and macroeconomic impacts. Once entered with a Quality Gate Score >= 80, claims are permanently preserved with cryptographic evidence hashes. Forecasts cannot be retroactively edited or erased upon failure.
The 2024–2026 Historical Verification Cycle reveals an overall forecaster accuracy rate of 66.7% across 24 resolved high-impact predictions. The biggest wins occurred in test-time compute scaling (Noam Brown/OpenAI o1) and humanoid hardware cost collapse (Unitree sub-$10k R1). The most notable misses came from stagnation skeptics claiming LLMs would hit an immediate wall in 2024 (Gary Marcus) and aggressive factory robotics timeline assertions (Elon Musk 2024 Optimus deployment).
The AFS™ uses a deterministic, strictly calibrated scoring rule: AFS = 0.30(Accuracy) + 0.25(Calibration) + 0.15(Timing) + 0.10(Discrimination) + 0.10(Magnitude) + 0.10(Consistency). Scores are evaluated using proper Brier and logarithmic probability scoring rules. Forecasters remain flagged as 'Provisional' until they accumulate at least three resolved predictions (N >= 3) to eliminate survivor and recency bias.
As of September 2026, AKI’s multi-source consensus engine (synthesizing prediction markets, expert surveys, and frontier lab roadmaps) places the median probability threshold for full general artificial intelligence around June 2031. However, AGI-2 (Human Specialist level) has already saturated in 2025–2026, and AGI-3 (Autonomous Digital Workers) is projected for mid-2028.
Prediction markets like Polymarket and Kalshi apply an epistemic 'haircut' to statements made by frontier lab executives. While lab CEOs frequently project 70%+ probabilities for AGI arrival by 2026–2027 to accelerate fundraising and talent recruitment, market participants price real-world constraints—such as data center power grid interconnect delays, thermal cooling limits, transformer reasoning walls, and regulatory anti-trust hurdles.
The AI Surprise Index™ measures the quantitative divergence between market/expert consensus and empirical physical outcomes. A surprise event is logged when an outcome resolves more than one standard deviation faster (positive surprise, such as the humanoid BOM collapse) or slower (negative surprise, such as the delay in generalized unsupervised robotaxi fleets) than baseline projections.