By 2028, AI systems will reach 'independent researcher' level — driving autonomous scientific discoveries without human intervention.
Predictor: Sam Altman
Prediction text
By 2028, AI systems will reach 'independent researcher' level — driving autonomous scientific discoveries without human intervention. | Autonomous research agent benchmarks
Key catalyst: Autonomous research agent benchmarks
Watch events: Autonomous experiment-design systems; material discovery wins; AI-authored publications in top journals.
Resolution evidence
Anthropic/Amodei timeline (AGI 2026-2027) is even more aggressive; DeepMind AlphaEvolve and OpenAI internal research agents suggest directional validation.
Predictor: Sam Altman
Calibration plot (stated vs observed)
Evidence about this node from Sam Altman 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
- 2025-10-28hitSam Altman publicly commits OpenAI to fully autonomous AI researcher by 2028How: Altman states publicly that OpenAI roadmap targets a legitimate AI researcher by 2028 conducting independent scientific workSource: TechCrunch - Sam Altman says OpenAI will have a legitimate AI researcher by 2028conf 99%Notes: HIT - direct verbatim claim; this prediction source statement is now anchored to a public commitment.
- 2025-10-29hitOpenAI commits to 30 GW of compute infrastructureHow: OpenAI announces or contractually commits to 30 GW cumulative compute capacity to support automated researcher buildoutSource: TechCrunch / Technology.org reporting on Altman 30 GW commitmentconf 85%
- 2026-09-30pendingOpenAI delivers intern-level research assistant by Sept 2026How: OpenAI publicly releases an intern-level AI research assistant capable of literature review hypothesis generation basic experiment designSource: TechCrunch / eWeek - OpenAI plan: fully automated AI researchers by 2028conf 65%Notes: Critical interim milestone - slip past 2026-09-30 should reduce confidence in 2028 endgame.
- 2026-06-01 → 2027-12-31pendingAI system co-authors a peer-reviewed paper as primary authorHow: An AI system listed as first or sole author on a paper in a top-tier peer-reviewed venue Nature Science Cell NeurIPS ICMLSource: Nature / Science editorial policies on AI authorshipconf 45%Notes: Currently most journals prohibit AI as first author - change in policy would be a strong leading signal.
- 2027-06-01 → 2028-12-31pendingAI system independently produces novel scientific finding without human guidanceHow: Documented case where AI system independently formulates hypothesis designs experiment and produces validated novel result without per-step human directionSource: OpenAI / DeepMind / Anthropic research disclosuresconf 35%Notes: Cascade - direct event resolution. Altman 2028 target is aggressive; community estimates skew later.
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
Network propagation neighbors
Top incoming (parents)
Edges that influence THIS node's belief
| Kind | Node | Their prob | P(c|s=T) | P(c|s=F) | Δ implied |
|---|---|---|---|---|---|
| killer | TK06 China-Taiwan Military Conflict | 8.0% | 0.050 | 0.550 | +0.196 |
| killer | TK11 Autonomous Regulatory Block (Level 4 Halt) | 10.0% | 0.050 | 0.550 | +0.186 |
| killer | TK03 AI Regulatory Moratorium (EU/US Capability Freeze) | 10.0% | 0.050 | 0.550 | +0.186 |
| killer | TK01 AGI Capability Plateau (2026-27 Training Stall) | 15.0% | 0.050 | 0.550 | +0.161 |
| prereq | 241_006 Everyone is out of hardware and electricity due to AI boom — Eric Schmidt | 78.9% | 0.550 | 0.050 | +0.126 |
Top outgoing (children)
Predictions THIS node influences
| Kind | Node | Their prob | P(c|s=T) | P(c|s=F) | Δ implied |
|---|---|---|---|---|---|
| prereq | 235_030 Ray Kurzweil predicts Longevity Escape Velocity (LEV) by 203 — Ray Kurzweil | 39.2% | 0.750 | 0.050 | -0.127 |
| prereq | 239_009 People will be on Mars within 10 years — Elon Musk | 30.9% | 0.550 | 0.050 | -0.106 |
| prereq | 232_047 Mass drivers on the moon will shoot AI satellites into deep — Elon Musk | 30.0% | 0.550 | 0.050 | -0.096 |
| prereq | 239_008 Moon base will exist in 10 years — Elon Musk | 28.8% | 0.550 | 0.050 | -0.084 |
| prereq | SEM_034 True artificial general intelligence will be achieved betwee — Demis Hassabis | 28.7% | 0.550 | 0.050 | -0.083 |
Ticker exposure
Beneficiaries (13)
Adverse (4)
Prerequisites (130)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| prereq | 248_040 | Pausing AI will fail and only accelerate race dynamics. | AI | — |
| prereq | 241_052 | Standard AI data centers being built at 400 MW scale, half-mile long | Energy | — |
| prereq | 232_014 | Recursive self-improvement is already here, not 12 months away. | AI | — |
| prereq | 229_013 | The next 12-18 months will be the largest AI transformation ever seen. | AI | — |
| prereq | 238_009 | Recursive self-improvement is already happening now (no longer three years out) | AI | — |
| prereq | 233_020 | Texas Sports Academy will operate at $300/month per parent using Texas billion-dollar voucher program. | Education | — |
| prereq | 241_004 | AI progress is not stoppable or controllable by any government, individual, or corporation | AI | — |
| prereq | 241_006 | Everyone is out of hardware and electricity due to AI boom | AI | — |
| prereq | 229_047 | Figure's 3,000 B200 GPU cluster is coming online for Helix pre-training with much larger GPUs planned. | AI | — |
| prereq | 235_038 | David Sinclair begins partial epigenetic reprogramming trials with Life Biosciences in March 2026. | Biotech/Longevity | — |
| prereq | 238_020 | Math field is 'cooked' — AI will solve research-level mathematics (first open hard math problem imminently) | AI | — |
| prereq | CMQ_001 | By 2026, AI will reach 'intern-level' capability — millions of virtual interns performing supervised, economically useful tasks. | AI | — |
| prereq | 237_025 | We are currently in the midst of the singularity, not waiting for it in the 2040s as originally expected. | AI | — |
| prereq | 241_026 | Space data center technology is understood and largely figured out | Space | — |
| prereq | 233_016 | Alpha will keep opening schools coast-to-coast; 13 new schools this year with more to come. | Education | — |
| prereq | 241_024 | Heat dissipation and radiation issues in space data centers will be addressed | Space | — |
| prereq | 241_060 | Data centers will be built in space / rockets will carry data centers | Space | — |
| prereq | 247_041 | AI-powered drugs have 85% phase 1 success vs 52% traditional | Biotech/Longevity | — |
| prereq | 235_022 | US will add a record 86 gigawatts of utility-scale capacity this coming year. | Energy | — |
| prereq | 235_033 | Chinese eVTOL taxi service moving toward commercial operations in 2027. | Auto/Transport | — |
| prereq | 237_002 | We will see a lot of evolution and many OpenClaw variants emerging very quickly as an early domain being developed. | AI | — |
| prereq | 241_011 | By end of 2026, no one will write code manually - it'll be a quaint skill | Labor/Jobs | — |
| prereq | 232_010 | Voice becomes the new interface in the AI era, replacing typing. | AI | — |
| prereq | 241_054 | Space wins by far on energy argument for data centers | Space | — |
| prereq | 238_004 | Future Vision X-Prize will produce at least one (possibly two) feature films depicting the future | Media/Ads | — |
| prereq | 243_044 | Tesla with FSD is 10 times safer than human driving (host claim) | Auto/Transport | — |
| prereq | 231_005 | Next year frontier reasoning costs will be pennies (down from ~$7 now). | AI | — |
| prereq | 231_012 | Within 6 months to a year physics problems will be massively addressed by AI. | AI | — |
| prereq | 240_056 | Anthropic enterprise market share to continue growing (went from 40% to 73% in 3 months) | AI | — |
| prereq | 242_057 | Artemis 2 launches April 1, 2026 - circumlunar mission | Space | — |
| prereq | 247_046 | AI investment will reach $3 billion per day, up from $2B in 2025 | AI | — |
| prereq | 244_032 | Uber will cash flow $10B this year (2026) | Markets/Stocks | — |
| prereq | 232_018 | We are living through a moment that will be in history for thousands of years: nonhuman intelligence arrived. | AI | — |
| prereq | 241_014 | The world a year from today will be nothing like the world today | AI | — |
| prereq | 242_012 | Flying cars (eVTOLs) will be operating in US within 18 months | Auto/Transport | — |
| prereq | 237_001 | Increased hacker profile and AI-driven social engineering attacks will become more and more prevalent. | AI | — |
| prereq | 231_013 | Math is cooked (will be solved), physics cooked, biology char broiled. | AI | — |
| prereq | 241_036 | No frontier AI company will emerge in Russia due to war | AI | — |
| prereq | 241_062 | Best AI labs will need to be controlled by humans (at the moment) | AI | — |
| prereq | 243_041 | Delivery/retail (non-food) business growing much faster than expected | Consumer | — |
| prereq | 243_043 | Tesla Cybercab priced around $30K vs Waymo cars around $150K | Auto/Transport | — |
| prereq | 244_016 | AV will grow very quickly but remains very small industry today | Auto/Transport | — |
| prereq | 244_035 | Uber will only enter adjacencies that 'rhyme' with core capabilities | Other | — |
| prereq | 247_006 | Anthropic ARR will reach $100 billion by end of 2026 | AI | — |
| prereq | 246_026 | In next year, significant cyber threats to mitigate from AI models (cited Altman). | AI | — |
| prereq | 246_027 | AI models will be capable enough for terrorist groups to create novel pathogens (near future). | AI | — |
| prereq | 246_035 | Terafab pilot phase is $25B, yielding ~$4B/year revenue for Intel. | AI | — |
| prereq | CMQ_004 | AGI-like models matching or outperforming human experts across most professional domains could arrive as early as 2026-2027. | AI | — |
| prereq | 235_014 | Sam Altman predicted 40x year-over-year hyperdeflation of AI costs at constant capability. | AI | — |
| prereq | 236_029 | Regulatory regime will be very pro-business despite 80% wanting AI more regulated | AI | — |
| prereq | 229_007 | Figure's entire 2026 roadmap will be built around Helix 2, extending from logistics to full-body long-horizon control. | Robotics | — |
| prereq | 237_024 | Skippy will listen to this YouTube video and self-improve based on the transcript (Peter's expectation). | AI | — |
| prereq | 238_017 | Uber will automate approximately 30% of its employment this year (2026) | Labor/Jobs | — |
| prereq | 240_005 | The organizational singularity will impact every company, nonprofit, and government | AI | — |
| prereq | 240_014 | Cost of reasoning models has dropped 1,000x in 16 months | AI | — |
| prereq | 241_016 | 92 GW power shortage in America between now and 2030 | Energy | — |
| prereq | 241_017 | Equivalent of ~60 nuclear plants needed by 2030 but essentially zero being built | Energy | — |
| prereq | 241_019 | AI scaling laws have not hit an asymptote yet | AI | — |
| prereq | 241_031 | Scientists don't agree yet on approach for recursive self-improvement | AI | — |
| prereq | 241_055 | Next frontier of AI infrastructure is space | Space | — |
| prereq | 241_037 | Chinese AI strategy will stay open source / open weights | AI | — |
| prereq | 241_047 | Universities should create prompt engineering course starting September for all freshmen | Education | — |
| prereq | 241_049 | Underage kids / AI vulnerability problem must be solved - line we can't cross | AI | — |
| prereq | 237_003 | 12 months in AI will be like 12 years; the next year will be wild. | AI | — |
| prereq | 241_040 | A Chernobyl-like AI tragedy is likely to happen, will awaken world to act | AI | — |
| prereq | 241_041 | Biological and nuclear attacks spawned by AI are real dangers | AI | — |
| prereq | 241_046 | Gemini 3 is probably the broadest of the non-Chinese AI systems | AI | — |
| prereq | 229_015 | The current Figure facility will top out around 50,000 units per year at full capacity. | Robotics | — |
| prereq | 232_034 | Apple owning an open-weight LLM hosting strategy would be the single best product strategy for their hardware. | AI | — |
| prereq | 232_013 | Recursive self-improvement loops likely to go live in the next 12 months; 2026 will be the most consequential year for humanity. | AI | — |
| prereq | 241_032 | World can accommodate ~10 frontier AI companies at scale | AI | — |
| prereq | 241_033 | Few frontier AI companies will be in China | AI | — |
| prereq | 241_034 | One or two frontier AI companies in Europe | AI | — |
| prereq | 232_009 | Startups will continue to deploy dangerous AI tech even after hyperscalers voluntarily pause it. | AI | — |
| prereq | 246_025 | Sam Altman predicts world-shaking cyber attack this year (2026). | AI | — |
| prereq | 241_051 | AI systems eventually won't have the ability to do completely new things | AI | — |
| prereq | 241_039 | In a year we'll know better which AI labs pull ahead | AI | — |
| prereq | 248_008 | Jack Dorsey aims to reduce Block's org depth from 5 to 2-3 layers this year. | AI | — |
| prereq | 240_034 | Morgan Stanley forecasts 13-40GW data center power shortfall through 2028 | Energy | — |
| prereq | 240_006 | OpenClaw on AWS will be the biggest enterprise unlock in decades | AI | — |
| prereq | 232_008 | Real-time AI video generation will break any video/voice-based security mechanisms. | AI | — |
| prereq | 248_024 | AI will drive an explosion of new religions and AI cults. | Other | — |
| prereq | 238_025 | AI computer-use benchmarks (OSWorld, Tbench) have broken through human level | AI | — |
| prereq | 229_006 | Figure will graduate from room-scale autonomy to full-house autonomy next. | Robotics | — |
| prereq | 237_012 | Over the next 12 months OpenClaw technology will be digested into corporations, causing destruction (layoffs) but also growth. | Labor/Jobs | — |
| prereq | 245_035 | Gene drive tech is safer than prior dispersals and can be rolled back | Biotech/Longevity | — |
| prereq | 229_004 | By summer 2026, Figure will have almost none of its supply chain in China. | Geopolitics | — |
| prereq | 229_043 | Figure expects to hire a small handful of lead customers rather than broad distribution over next 6-12 months. | Robotics | — |
| prereq | 230_001 | A billion-dollar revenue company is already being run by an AI CEO (with human as legal puppet). | AI | — |
| prereq | 230_042 | A pure AI organization will emerge soon — won't look efficient, will look 'literally alien'. | AI | — |
| prereq | 231_004 | The solution wavefront will spread from math/coding to physics, chemistry, and other fields. | AI | — |
| prereq | 234_026 | AI rent-a-human service feature of scoring humor/visuals will be gone in months | AI | — |
| prereq | 234_027 | Element Biosciences will launch $600K desktop $100-per-genome machine in second half of 2026 | Biotech/Longevity | — |
| prereq | 234_050 | Open-source maintainers will be overwhelmed by AI-discovered software vulnerabilities | AI | — |
| prereq | 237_016 | A specific-sliver OpenClaw-powered vertical SaaS could become a $5 million company overnight for ~$200 subscription cost. | Markets/Stocks | — |
| prereq | 237_018 | We are seeing a Cambrian explosion of OpenClaw variants (PicoClaw, IronClaw, NanoClaw, Nanobot, etc.) with many more to come. | AI | — |
| prereq | 238_005 | By September 2026 there will be 1000+ ultra-high-inspirational-quality videos of the future generated nearly free | AI | — |
| prereq | 238_039 | Apple's underused unified memory overhang will likely collapse within the next year | AI | — |
| prereq | 238_047 | US will add 86 GW of new grid capacity in 2026, 51% from solar | Energy | — |
| prereq | 240_015 | Post-transformer architectures will make a 1000x cost reduction look like child's play | AI | — |
| prereq | 240_046 | Public interest in any alien disclosure will fade within one or two days | Media/Ads | — |
| prereq | 241_035 | Maybe one frontier AI company in India | AI | — |
| prereq | 246_023 | Grok 5 release ~20% probability for Q2 2026 per PolyMarket. | AI | — |
| prereq | 247_048 | Battery energy densities improving 8% year-over-year per constant dollar | Energy | — |
| prereq | 248_028 | Personalized AI-designed medicine stories will repeat until it's n=10 billion. | Biotech/Longevity | — |
| prereq | 248_032 | First-generation neural uploads will be destructive; 2nd-4th generation will be non-destructive. | Biotech/Longevity | — |
| prereq | 248_036 | AI will enable one-person conglomerates and one-person religions. | AI | — |
| prereq | 248_049 | Humans (weaker intelligences) can successfully align/contain super-intelligences via weak-to-strong supervision. | AI | — |
| prereq | 240_045 | White House to disclose something about aliens/UAPs in the next few months (rumors of June/July/summer) | Geopolitics | — |
| prereq | 240_004 | A new repo from 2027 will go from zero to a billion stars in 5 minutes | AI | — |
| prereq | 248_025 | Religion will consolidate globally around fewer major religions while proliferating micro cults/sects. | Other | — |
| prereq | 248_031 | Enhanced Games world records will accelerate on a scaling-law-style trajectory year over year. | Biotech/Longevity | — |
| prereq | 248_044 | Alex: likely user ~1 million of a BCI implant (not 101) once upgradeable. | Biotech/Longevity | — |
| prereq | 233_009 | A motivational learning bundle with video game designers will be released in 2026. | Education | — |
| prereq | 246_030 | DeepSeek V4 will NOT shock markets like prior DeepSeek releases. | Markets/Stocks | — |
| prereq | 246_048 | Anthropic Mythos is ~5x more expensive than Opus, likely won't be widely run. | AI | — |
| prereq | 247_012 | The bar for AI startups will rise to require being recursively self-improving | AI | — |
| prereq | 245_017 | Disease/drought-resistant plants and animals market is currently hundreds of billions of dollars | Biotech/Longevity | — |
| prereq | 248_007 | We will see an explosion of AI-driven pop-up shops, retail venues, and malls orchestrated by AI. | AI | — |
| prereq | 240_031 | Samsung will be optimized by orders of magnitude by Elon | AI | — |
| prereq | 248_014 | If China leapfrogs the West algorithmically, the open/closed publication equilibrium will flip. | AI | — |
| prereq | 235_035 | Apple finally launching AI with Gemini integration at WWDC in June 2026. | AI | — |
| prereq | 248_029 | Regime change at the FDA is not beyond the realm of reason. | Biotech/Longevity | — |
| prereq | 233_004 | AI will be able to bootstrap four-year-olds into reading within roughly 18 months. | AI | — |
| correlate | S_AGI_FAST_2027 | AGI fast: drop-in remote worker by 2027-09 | agi_general_capability | — |
| correlate | S_AGI_SLOW_2031 | AGI slow: Schmidt/Hassabis 5-10 year path | agi_general_capability | — |
| killer | TK01 | AGI Capability Plateau (2026-27 Training Stall) | — | — |
| killer | TK11 | Autonomous Regulatory Block (Level 4 Halt) | — | — |
| killer | TK03 | AI Regulatory Moratorium (EU/US Capability Freeze) | — | — |
| killer | TK06 | China-Taiwan Military Conflict | — | — |
Dependents (5)
| Type | Pred | Title | Domain | Lag |
|---|---|---|---|---|
| prereq | 235_030 | Ray Kurzweil predicts Longevity Escape Velocity (LEV) by 2033. | Biotech/Longevity | — |
| prereq | SEM_034 | True artificial general intelligence will be achieved between 2032 and 2042 — 'first we solve AI, then use AI to solve everything else'. | AI/AGI | — |
| prereq | 239_008 | Moon base will exist in 10 years | Space | — |
| prereq | 239_009 | People will be on Mars within 10 years | Space | — |
| prereq | 232_047 | Mass drivers on the moon will shoot AI satellites into deep space; self-sustaining lunar city will follow. | Space | — |
Linked documents (2)
| Sim | Source | Title | Market prob | Polarity | Reviewed | Published |
|---|---|---|---|---|---|---|
| 0.691 | arxiv | When AI Meets Science: Research Diversity, Interdisciplinarity, Visibility, and Retractions across Disciplines in a Global Surge | — | mentions | pending | 2026-05-07 |
| 0.664 | arxiv | SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones? | — | mentions | pending | 2026-05-28 |
Raw metadata
{
"nia": false,
"qty": "researcher-level",
"mode": "FORECAST",
"role": "Cited-CEO",
"caveats": "Assumes continued compute + algorithmic scaling; requires unhobbling gains (memory, tool use, RL).",
"context": "Second rung of Altman's capability roadmap; step up from supervised intern work to autonomous discovery. Requires long-horizon agent loops, tool use, and self-verification.",
"to_year": 2028,
"conv_cues": "will achieve; specific year; CEO",
"direction": "HAPPEN",
"from_year": 2028,
"timeframe": "by 2028",
"conv_level": "HIGH",
"milestones": [
{
"kind": "llm_pre_event",
"label": "Sam Altman publicly commits OpenAI to fully autonomous AI researcher by 2028",
"notes": "HIT - direct verbatim claim; this prediction source statement is now anchored to a public commitment.",
"source": "TechCrunch - Sam Altman says OpenAI will have a legitimate AI researcher by 2028",
"status": "hit",
"weight": 0.4,
"ordinal": -12,
"source_id": null,
"confidence": 0.99,
"source_url": "https://techcrunch.com/2025/10/28/sam-altman-says-openai-will-have-a-legitimate-ai-researcher-by-2028/",
"expected_date": "2025-10-28",
"observed_date": "2025-10-28",
"research_origin": "deep_research",
"measurement_criterion": "Altman states publicly that OpenAI roadmap targets a legitimate AI researcher by 2028 conducting independent scientific work"
},
{
"kind": "llm_pre_event",
"label": "OpenAI commits to 30 GW of compute infrastructure",
"source": "TechCrunch / Technology.org reporting on Altman 30 GW commitment",
"status": "hit",
"weight": 0.4,
"ordinal": -11,
"source_id": null,
"confidence": 0.85,
"source_url": "https://www.technology.org/2025/10/29/openais-ceo-sets-2028-deadline-for-openai-to-have-a-fully-automated-ai-researcher/",
"expected_date": "2026-02-14",
"observed_date": "2025-10-29",
"research_origin": "deep_research",
"expected_date_range": {
"to": "2026-06-30",
"from": "2025-10-01"
},
"measurement_criterion": "OpenAI announces or contractually commits to 30 GW cumulative compute capacity to support automated researcher buildout"
},
{
"kind": "prereq",
"label": "The next 12-18 months will be the largest AI transformation ever seen.",
"status": "hit",
"weight": 0.5,
"ordinal": -10,
"source_id": "229_013",
"expected_date": "2026-04-29",
"observed_date": "2026-04-29"
},
{
"kind": "prereq",
"label": "Figure will graduate from room-scale autonomy to full-house autonomy next.",
"status": "pending",
"weight": 0.5,
"ordinal": -9,
"source_id": "229_006",
"expected_date": "2026-06-17",
"observed_date": null
},
{
"kind": "prereq",
"label": "The current Figure facility will top out around 50,000 units per year at full capacity.",
"status": "pending",
"weight": 0.5,
"ordinal": -8,
"source_id": "229_015",
"expected_date": "2026-06-18",
"observed_date": null
},
{
"kind": "prereq",
"label": "Figure's entire 2026 roadmap will be built around Helix 2, extending from logistics to full-body long-horizon control.",
"status": "pending",
"weight": 0.5,
"ordinal": -7,
"source_id": "229_007",
"expected_date": "2026-06-28",
"observed_date": null
},
{
"kind": "prereq",
"label": "By summer 2026, Figure will have almost none of its supply chain in China.",
"status": "pending",
"weight": 0.5,
"ordinal": -6,
"source_id": "229_004",
"expected_date": "2026-07-18",
"observed_date": null
},
{
"kind": "llm_pre_event",
"label": "OpenAI delivers intern-level research assistant by Sept 2026",
"notes": "Critical interim milestone - slip past 2026-09-30 should redu
... (truncated)