Scan: Will an AI lab announce another Millennium Prize solution by December 31, 2026?

1. Planner Fable 5.1 · Fermi decomposition + tool plan

Tools chosen: polymarket_direct polymarket_related kalshi_related claude_news article_search gdelt_news wikipedia

Sub-questions (Fermi decomposition)

Which AI lab announced a Millennium Prize (Navier-Stokes) solution, when, and how was that announcement worded and received — does it establish a precedent for labs making bold 'solved' claims?
Are any eligible labs (Google DeepMind, OpenAI, Harmonic, Axiom Math, Math Inc, DeepSeek, etc.) publicly working on or hinting at results for Riemann, P vs NP, Yang-Mills, Hodge, or BSD as of now?
How many times in the past 2 years have AI labs claimed to fully solve (not partially advance) a famous open math conjecture, and how many were retracted or reframed as partial?
What is the current Polymarket price and trend for this market, and are there related markets (e.g., 'Riemann Hypothesis solved by 2026', 'AI solves Millennium Prize') on Polymarket or Kalshi that imply a probability?
Have expert mathematicians (e.g., Terence Tao, Clay Institute statements) commented on the plausibility of AI resolving any remaining Millennium Problem within a year?
How strictly would the resolution criteria treat ambiguous claims (e.g., a lab saying its model 'proved' a conjecture pending verification), and is there a pattern of labs using such language?

Planner reasoning

The title's 'another' and the explicit Navier-Stokes exclusion imply an AI lab already announced a Navier-Stokes solution, so the key questions are the base rate of such announcements, which labs (e.g., Harmonic, Axiom Math, Math Inc, Google DeepMind, OpenAI) are actively targeting the remaining five problems, and how loosely 'announces solved' might be triggered by hype-prone labs. The Polymarket price is the primary anchor; news search establishes the current state of AI-driven math claims and any pending announcements.

2. Research Parallel tool calls

tool	status	time	items	summary
polymarket_direct	OK	0.3s	1	## This Market's Polymarket Data Will an AI lab announce another Millennium Prize solution by December 31, 2026? - Current price (probability): 27.50% - 7-day price change: -9.50% - 30-day price change: -9.50% - Total volume: $26,087 (USD notional) - Price range: 27.00% - 37.00% - Data points:
polymarket_related	OK	2.0s	0	Scanned 100 active Polymarket markets, kept 0 matches. keyword 'Millennium Prize': 0 markets \| keyword 'Riemann Hypothesis': 0 markets \| keyword 'P vs NP': 0 markets \| keyword 'Navier-Stokes': 0 markets \| keyword 'AI solves math': 0 markets
kalshi_related	OK	1.8s	0	0 related markets / summaries. keyword 'Millennium Prize': no matches \| keyword 'Riemann Hypothesis': no matches \| keyword 'P vs NP': no matches
claude_news	OK	27.1s	14	## Key Findings - OpenAI announced a Navier-Stokes "solution" on Sept 8, 2026: the AI company announced on 8 September that it has generated a solution to the most commonly used physical model of fluids — the Navier-Stokes equations — showing that they can break down . A group of 10,000 auton
article_search	OK	39.7s	33	Fetched 33 unique articles across 5 queries (lookback=120d). 'AI lab claims Navier-Stokes solved Millennium Prize': 10 articles \| 'AI proves Riemann Hypothesis': 10 articles \| 'AI lab announces proof of open math conjecture': 10 articles \| 'Harmonic Axiom Math Inc Millennium Prize': 10 articles \| 'D
gdelt_news	OK	187.7s	30	GDELT: 30 articles across 4 queries (lookback=90d). 'Millennium Prize AI': error GDELT rate-limited after retries (429) \| 'Navier-Stokes AI solution': 10 hits \| 'Riemann Hypothesis AI proof': 10 hits \| 'P vs NP AI claim': 10 hits
wikipedia	OK	0.2s	4	Fetched 4 Wikipedia entries (1 missing pages).

3. Evidence Brief Sonnet 5 · 9019 chars

# Current state The market asks whether any eligible AI lab will announce a solution to one of the five *remaining qualifying* Millennium Prize Problems (Riemann, P vs NP, Yang-Mills, Hodge, BSD) by Dec 31, 2026 — Navier-Stokes explicitly does NOT count. As of Sept 2026, OpenAI made a high-profile Navier-Stokes "solution" announcement, but this is excluded by the market's own rules. No eligible lab has yet made a qualifying "solved" claim on any of the five listed problems; the most advanced work (Anthropic on Riemann, DeepMind on singularities) is characterized by labs/experts as "progress," not solution. # Timeline of key events - 2026-08-11 (confirmed): TechCrunch reports an unreleased Anthropic model made "significant progress" on the Riemann Hypothesis (raised lower bound of zeros satisfying it); not a solution. - 2026-08-13 (confirmed): Reports clarify Anthropic's Claude found a new bound (de Bruijn–Newman constant) while attempting Riemann; hypothesis itself unproven. - 2026-09-08/09 (confirmed): OpenAI announces a swarm of ~10,000 AI agents produced a proof that Navier-Stokes solutions can develop a singularity ("solved" per OpenAI's own framing of variant C/D), formalized in Lean. This problem is explicitly EXCLUDED from this market's qualifying list. - 2026-09-08/09 (confirmed): Priority/credit dispute erupts — mathematicians Tristan Buckmaster and Levent Alpöge (Anthropic-linked) allege OpenAI used/credited their prior unpublished work improperly; OpenAI denies. - 2026-09-09 (confirmed): Clay Mathematics Institute (president Martin Bridson) does not certify or endorse the result; still lists Navier-Stokes as unsolved. OpenAI states it does not intend to claim the $1M prize. - 2026-09-09 (confirmed): Terence Tao publicly criticizes the "frenetic competition" dynamic AI is creating around famous conjectures, warning of a "meaningless production quota game." - 2026-09-09 (reported): Wikipedia and multiple outlets confirm only the Poincaré Conjecture has ever been officially/formally accepted as solved among the seven Millennium Problems. # Event Will an AI lab announce another Millennium Prize solution (Riemann, P vs NP, Yang-Mills, Hodge, or BSD only — Navier-Stokes excluded) by Dec 31, 2026? # Outcomes to forecast Yes / No # Kalshi market anchor No kalshi_direct tool output was returned in this research pass — Kalshi price unavailable directly. The only price data available is from Polymarket (same underlying event/ticker cross-listed): **current YES ≈ 27.5%**, down from ~37% a week/month ago (-9.5% both 7d and 30d), on thin volume (~$26K total, only 3 data points). Treat this Polymarket price as the best available consensus proxy in absence of native Kalshi data — flagged as a gap. # Sub-question answers 1. **Navier-Stokes precedent** — OpenAI announced a Navier-Stokes "solution" Sept 8, 2026, framed confidently ("resolved," proof produced by internal system, Lean-formalized), but faced immediate CMI non-certification, priority disputes, and expert skepticism (Tao). It sets precedent for labs making bold "solved" claims quickly, but also shows such claims draw intense scrutiny and controversy rather than clean acceptance — and this specific claim doesn't qualify for this market. [Quanta, OpenAI, Nature, CNN, Wikipedia] 2. **Eligible labs working on remaining problems** — Anthropic (Riemann, progress on bounds, Aug 2026), OpenAI (Astra model reportedly produced proofs for "ten open problems," unspecified if Millennium-tier), Google DeepMind (unstable singularities in fluid equations, Navier-Stokes-adjacent, not a qualifying problem). No lab has been reported working toward full solutions of P vs NP, Yang-Mills, Hodge, or BSD specifically. [TechCrunch, tech-insider.org] 3. **Past 2-year false/retracted "solved" claims** — Only one clear full "solved" claim event in the reporting window: OpenAI's Navier-Stokes (Sept 2026), which remains unverified/disputed (not retracted, but contested and not accepted by CMI). No other full-solution claims on qualifying problems found in research. Anthropic's Riemann claim was explicitly framed/reported as "progress," not solved — media (Scientific American) pushed back against overstatement. [claude_news, Scientific American headline] 4. **Polymarket price/related markets** — Polymarket price for this exact market: 27.5%, trending down (-9.5% 7d/30d) from a high of 37%. No other related Polymarket or Kalshi markets found (0 matches for Millennium Prize, Riemann, P vs NP, Navier-Stokes keywords). A separate Manifold market (pre-OpenAI announcement) had ~38% for "AI solves a Millennium Prize Problem in 2026" (broader, includes Navier-Stokes). [polymarket_direct, polymarket_related, kalshi_related, tech-insider.org] 5. **Expert commentary on plausibility** — Terence Tao is skeptical/critical, describing AI math claims as turning into a "meaningless production quota game" and cautioning against overclaiming; he does not endorse near-term full solutions to remaining problems. CMI's Bridson is cautiously non-committal, requiring independent verification. No mathematician quoted expects Riemann/P vs NP/Yang-Mills/Hodge/BSD to fall within 2026. [CNN, Nature] 6. **Strictness of resolution criteria on ambiguous claims** — Rules require explicit "solved" framing (not partial progress), sole/joint lab credit (not just crediting outside researchers), and no CMI confirmation is required. The Navier-Stokes case shows labs do use strong "solved" language despite lack of independent verification — suggesting if a qualifying-problem claim occurs, wording alone (not verification) could trigger Yes. This raises resolution risk/uncertainty around interpretation of partial vs. full claims. [Market description; OpenAI Navier-Stokes framing as analogy] # Key facts (high-confidence, factual) 1. [Wikipedia/CMI] Only the Poincaré Conjecture has ever been officially accepted as solved among the 7 Millennium Problems. 2. [OpenAI/Quanta/Nature] OpenAI's Sept 2026 Navier-Stokes claim is NOT a qualifying problem for this market and remains CMI-unverified/disputed. 3. [TechCrunch] Anthropic reported "progress" (not solution) on Riemann via an unreleased model, Aug 2026. 4. [CNN] Terence Tao publicly critical of the trend of AI labs racing to claim math breakthroughs. 5. [Polymarket] Current price 27.5%, down sharply from a high of 37% over the past week/month, thin volume/liquidity. # Cross-market signals - Kalshi related: none found (0 matches on searched keywords); kalshi_direct price not returned in this research pass. - Polymarket: same-event contract at 27.5% YES, declining trend, low volume/liquidity (small sample, only 3 data points) — low confidence signal. - Sportsbook implied: N/A. - Other: pre-OpenAI-announcement Manifold market had ~38% for broader "AI solves a Millennium Prize in 2026" (included Navier-Stokes, since resolved-adjacent). # Analyst opinions and speculation - claude_news synthesis: OpenAI's disputed Navier-Stokes claim raises the plausibility that another lab (Anthropic, DeepMind) could make a similarly bold — if contested — claim on a qualifying problem before year-end, given "frenetic competition." - Tao: warns of degraded incentives/quality in AI-driven math claims; skeptical of genuine full solutions arriving imminently. - CMI (Bridson): cautious, non-committal, awaiting verification — implicitly signals no qualifying problem is close to consensus-solved. # Directional lean per outcome - **Yes**: Momentum/competitive dynamic post-OpenAI Navier-Stokes announcement; Anthropic already probing Riemann; labs show willingness to make bold "solved" announcements before independent verification (lowers bar given market doesn't require CMI confirmation). Three-plus months remain. - **No**: Navier-Stokes (the only actual "solved" claim to date) is explicitly excluded; no eligible lab has approached full solutions on Riemann, P vs NP, Yang-Mills, Hodge, or BSD — only incremental progress reported; expert consensus (Tao, CMI) skeptical; Polymarket price declining toward ~27%, well under 50%, suggesting market participants increasingly doubt a qualifying claim will occur in remaining months. # Gaps / unknowns - No native Kalshi YES price was returned by tools — only Polymarket cross-market data available; true Kalshi consensus unknown. - Polymarket data is thin (3 data points, $26K volume) — low statistical confidence. - Unclear how resolution would treat an ambiguous/contested claim (as with Navier-Stokes) if made on a qualifying problem — could resolve Yes on announcement alone per rules despite lack of verification. - No specific evidence found of any eligible lab actively targeting P vs NP, Yang-Mills, Hodge, or BSD. # Calibration anchors - Polymarket price (proxy anchor): 27.5%, trending down from 37% over past week/month. - Precedent base rate: 1 contested "solved" claim (Navier-Stokes, excluded) in ~2 years across all Millennium Problems by eligible labs; zero qualifying-problem full solutions to date.

4. Forecasters Fable 5.1 + GPT-5.6 Sol · in parallel

claude-fable-5-1

50% conf

Yes	20%
No	80%

The only 'solved' claim to date (OpenAI's Navier-Stokes, Sept 2026) is explicitly excluded, and the brief finds no eligible lab near a full solution on Riemann, P vs NP, Yang-Mills, Hodge, or BSD — Anthropic's Riemann work was framed by the lab and media as incremental progress on bounds, not a proof. The remaining five problems are generally regarded as substantially harder than Navier-Stokes regularity, and Tao/CMI commentary signals no imminent full solutions. The only price signal is a thin Polymarket contract at ~27.5% and declining from 37%; with ~3.5 months left and the competitive incentive for bold 'solved' framing (the market needs no CMI verification), I shade modestly below that anchor to ~20% rather than dismissing the tail. Native Kalshi price was unavailable, so the anchor itself is a weak proxy.

gpt-5.6-sol

64% conf

Yes	23%
No	77%

The best consensus proxy is Polymarket at 27.5% YES, though its thin volume and missing native Kalshi price weaken the anchor. The Navier-Stokes precedent modestly supports Yes because labs may announce contested solutions without CMI validation, but that claim is excluded and also demonstrates the scrutiny deterring overstatement. Anthropic has reported only incremental Riemann progress, while no lab is known to be near a full solution of any other qualifying problem and experts such as Terence Tao remain skeptical. With only several months remaining and zero qualifying claims to date, the evidence supports moving modestly below the market proxy to 23% YES.

5. Devil's Advocate Sonnet 5 · red-team critique

Both forecasters anchor heavily on the thin, declining Polymarket price (27.5%, only 3 data points, $26K volume) despite the brief explicitly flagging this as "low confidence" — yet neither meaningfully discounts for the possibility that this proxy is simply mispriced or stale relative to the true Kalshi market, especially given no native Kalshi price exists to cross-check.

Both forecasts converge tightly (20% and 23%) despite F1 claiming to "shade modestly below" the anchor while F2 also lands "modestly below" — this apparent agreement masks that neither seriously engages with the resolution-ambiguity risk flagged in sub-question 6: that wording alone (a bold "solved" claim, verified or not) could trigger Yes, which the Navier-Stokes precedent directly demonstrates labs are willing to do even under CMI non-certification and priority disputes.

Neither forecast weighs the "frenetic competition" dynamic (Tao's warning, claude_news synthesis) as a distinct causal driver — if one lab caused massive PR by claiming Navier-Stokes, competitive/reputational incentives among Anthropic, DeepMind, OpenAI could push a rival to make a bold claim on Riemann or another qualifying problem specifically to compete, which is a scenario-specific escalation risk understated in both.

Both forecasts treat "no lab is known to be near a full solution" on P vs NP, Yang-Mills, Hodge, or BSD as strongly decisive, but the brief's own reference class (only Poincaré ever solved among seven problems, and that took ~100 years of human effort) suggests the true base rate for ANY qualifying solution in a single year is likely far below 20%, meaning both forecasts may still be anchoring too high on the Polymarket price rather than this harder reference class.

Neither forecast explicitly addresses the downward Polymarket trend (37%→27.5%) as informative of resolution risk perception — i.e., does the decline reflect market participants updating that OpenAI's Navier-Stokes exclusion reduces Yes-probability, or does it reflect low liquidity noise? This distinction matters for how much to trust the anchor's direction versus its level, and neither forecaster explores it.

Both forecasts assign meaningful probability mass (20-23%) to Yes despite zero identified labs currently working toward P vs NP, Yang-Mills, Hodge, or BSD (per sub-question 2), leaving Riemann as the only remotely plausible pathway — this narrows the effective "shots on goal" considerably and arguably warrants a lower probability than either forecaster gives, or at minimum an explicit acknowledgment that the Yes case rests almost entirely on Anthropic's Riemann work escalating from "bound progress" to a full proof within 3.5 months.

6. Aggregator Fable 5.1 · final calibrated forecast

21% (57% confidence)

Yes: 21% No: 79%

Both forecasters drew the same inferences from the sub-questions: the only actual 'solved' claim (OpenAI's Navier-Stokes) is excluded, Anthropic's Riemann work is framed as bound-progress rather than proof, and no eligible lab is reported targeting P vs NP, Yang-Mills, Hodge, or BSD, leaving Riemann as essentially the sole pathway within ~3.5 months. The critique's competing points largely offset: the announcement-only resolution rule plus 'frenetic competition' dynamics keep a meaningful tail (labs have demonstrated willingness to declare 'solved' before verification), while the harsh reference class and single-shot-on-goal argue for staying below the thin Polymarket proxy of 27.5%. The declining, low-liquidity price is a weak anchor, so I weight it only modestly and land slightly below both it and F2, near F1. Final call ~21% Yes.

Pipeline Timing

Total pipeline time: 294.1s

Per-tool research timings shown in the Research section above.