Anthropic mathematician Levent Alpöge says Claude helped him develop a proof of the Riemann Hypothesis, the 167-year-old problem about the distribution of prime numbers proposed by Bernhard Riemann in 1859. Alpöge published the argument as Lean code and said the work took 41 days; he reported finding the solution on a Sunday night and contacting CEO Dario Amodei. The claim surfaced one day after Anthropic’s IPO, though Alpöge denied timing it to attract attention. The Clay Mathematics Institute, Anthropic and other outside groups have not confirmed the proof. MIT’s Andrew Sutherland and other researchers have compiled the code without errors, and the Lean kernel rechecked it using three standard axioms. Reviewers are still checking whether it proves the stated hypothesis without altered definitions or software exploits.
Anthropic raised about $81 billion in its IPO—reported as $81.3 billion at $83 a share—and closed its first trading day at $97.17, with a fully diluted valuation of about $1.8 trillion. Shares then fell: reports put the premarket decline at more than 10% ahead of the first full Nasdaq session and at 5.76% in broader Nov. 20 premarket trading. AI-related stocks including AMD, Intel and Nvidia also declined, alongside other chip and equipment companies, while Nasdaq 100 futures fell more than 1.4%. Asian markets were lower too. Federal Reserve officials warned that AI investment and data-center construction could add inflation and overheating risks; markets expected a quarter-point rate hike at the December meeting. One analysis says AI profits currently accrue mainly to providers of tools and infrastructure, though that distribution could change.
AI adoption is also creating operational questions across sectors. At HLTH 2026, healthcare providers and insurers said AI is raising costs even as they invest in it with hopes of improving efficiency and outcomes. Körber’s AI and Decision Complexity Report found 59% of companies piloting AI or using it across their businesses invest in AI-powered decision support. In the US, 77% said AI increasingly affects decisions and 78% said decision-making has become more complex. AI involvement was highest among supply-chain leaders, at 92%, and lowest in life sciences, at 60%. Körber says organizations need timely, trustworthy data, clear accountability, exception-handling processes and an appropriate balance of automation and human oversight.
Some deployments focus on practical assistance and research. Home Depot says its Magic Apron AI assistant handles millions of shopper questions each month, aiming to help customers get answers and avoid another store trip. Alabama A&M University, Mississippi State University Extension Service and Mississippi State’s Southern Rural Development Center are launching AI for Answers to train agricultural businesses and extension professionals in marketing, communications, program development and educational outreach. Participants will develop individual AI roadmaps, with an AI and marketing microcredential pathway also planned. In a separate investment-research project, Codex agents and LATO helped design a two-day study involving 12 experts, buyers and technical leaders across the US, Asia and Europe, with 14 hours of conversations; the resulting insights informed an investment decision.
Infrastructure and model research are advancing alongside these applications. Coontec said it will use Nvidia’s DGX B300 to strengthen threat detection for ASH, its AI-agent security platform. The system has eight Nvidia Blackwell Ultra GPUs, each with 288 GB of HBM3e memory, totaling more than 2 TB of GPU memory. Coontec plans to study models and attack scenarios, feed findings into ASH and its AI supply-chain checks, and offer pre-deployment risk assessments. Researchers introduced GLiDE, a family of decision models from 0.4B to 27B parameters that scores candidate answers in one pass without generating text or intermediate reasoning. Across five sizes, GLiDE outperformed tested models of equal or smaller size on the Hugging Face performance-and-parameter comparison. Its 27B model scored 71.52 with 2.0% expected calibration error and 92.9 ms median latency on one GPU; a further improvement raised the score to 73.37 with 2.1% error.
Key Takeaways
- Anthropic mathematician Levent Alpöge claims Claude helped him produce a Lean proof of the Riemann Hypothesis in 41 days, but the result remains unverified.
- Independent reviewers compiled Alpöge’s Lean code without errors; they are still checking that it proves the stated hypothesis and does not rely on altered definitions or software bugs.
- Anthropic raised about $81 billion in its IPO; its shares closed the first day at $97.17 after pricing at $83, then fell in premarket trading as AI stocks weakened.
- AMD, Intel and Nvidia were among AI-related companies whose shares fell amid broader market declines; Federal Reserve officials cited possible inflation and overheating risks from AI and data-center investment.
- Healthcare leaders at HLTH 2026 said AI is increasing costs even as they invest in it to pursue better efficiency and outcomes.
- Körber reported that 78% of US respondents say decision-making has become more complex, while AI involvement reaches 92% among supply-chain leaders and 60% in life sciences.
- Home Depot says its Magic Apron assistant handles millions of shopper questions monthly; Alabama A&M and Mississippi State partners are launching AI training and microcredentials for agricultural businesses and extension professionals.
- Codex agents and LATO helped build an investment-research study that included 12 experts and 14 hours of conversations over two days.
- Coontec plans to use Nvidia’s DGX B300, equipped with eight Blackwell Ultra GPUs and more than 2 TB of GPU memory, to strengthen AI security testing and ASH’s threat detection.
- GLiDE decision models range from 0.4B to 27B parameters; the 27B model recorded 92.9 ms median latency on one GPU and scored up to 73.37 in reported testing.
Claude Helped Levent Alpöge Develop a Claimed Riemann Hypothesis Proof in 41 Days
A timeline in a note accompanying Levent Alpöge’s claimed proof says the work took 41 days, after the Riemann Hypothesis had remained unsolved for 167 years. The timeline describes how Claude helped with the work. Alpöge released the proof as Lean code, but it has not been independently verified. Lean experts are working to check it.
Anthropic Mathematician Says Claude Helped Find a Riemann Hypothesis Proof
An Anthropic mathematician says Claude helped him discover a proof of the Riemann Hypothesis, a famous problem that has resisted solutions for 167 years. The claim came one day after the company’s IPO, according to the article. The proof has not yet been independently verified.
Independent Mathematicians Check Claude Assisted Riemann Hypothesis Proof
Levent Alpöge says he used Anthropic’s Claude to help produce a proof of the Riemann Hypothesis, then wrote it in Lean. Andrew Sutherland of MIT and other independent researchers have compiled the Lean code without errors. Sutherland said the Lean kernel rechecked the proof using only three standard axioms. Researchers are still checking that the code proves the stated hypothesis and does not change mathematical definitions or exploit a software bug.
Anthropic Researcher Claims Claude Helped Solve the Riemann Hypothesis
Anthropic mathematician Levent Alpöge claimed on November 20, 2026, that he had solved the Riemann Hypothesis with help from Claude. First proposed by Bernhard Riemann in 1859, the problem concerns the distribution of prime numbers and is one of the Clay Mathematics Institute’s seven Millennium Prize Problems, each worth $1 million for a solution that meets the institute’s requirements. Alpöge published his proof in Lean, but the Clay Mathematics Institute, Anthropic and other outside groups had not confirmed it. He said he found the solution Sunday night and contacted Anthropic CEO Dario Amodei, while denying that the timing was intended to draw attention to the company’s IPO.
Anthropic Mathematician Claims Claude Helped Solve 167 Year Old Math Problem
An Anthropic mathematician claims he used Claude to discover a proof of the Riemann Hypothesis, a famous problem in number theory that has remained unsolved for 167 years. Bernhard Riemann proposed the hypothesis in 1859. The article says the claimed proof remains unconfirmed.
Healthcare leaders say AI raises costs as they pursue better results
At HLTH 2026, healthcare providers and insurers said AI is increasing healthcare spending. They are still investing in AI because they expect it to improve efficiency and outcomes. The article highlights four takeaways about AI’s next phase in healthcare.
Anthropic’s $81 billion IPO signals a new phase for AI
Anthropic raised $81 billion in its initial public offering, marking a new phase in the AI boom. The article says AI profits now mostly go to companies that supply the tools and infrastructure, but that may change. The shift could have major effects on investors and businesses.
Anthropic shares drop ahead of first full Nasdaq trading day
Anthropic shares fell more than 10% to $97.17 in premarket trading ahead of their first full day on the Nasdaq. The decline followed a 17% rise on Thursday after the company priced its upsized IPO. The Nasdaq was also down, with technology stocks, especially AI-related ones, facing pressure.
LATO and Codex agents help speed up expert research for an investment
For an investment opportunity, a researcher needed to quickly understand how a company’s product stood out from well-funded competitors. Over two days, 12 category experts, buyers, and technical leaders across the US, Asia, and Europe took part in 14 hours of conversations. The researcher said Codex agents and LATO designed the study together, using earlier work to improve questions and identify experts. The resulting insights helped support a more informed investment decision.
Körber report finds AI makes decisions more complex for many leaders
Körber’s AI and Decision Complexity Report found that 59% of companies piloting AI and those using it across their businesses invest in AI-powered decision support. In the US, 77% of respondents said AI is increasingly involved in decisions, while 78% said decision-making has become more complex. Supply chain leaders reported the highest AI involvement at 92%, while life sciences leaders reported 60%. Körber says organizations need timely, trustworthy data, clear accountability, defined ways to handle exceptions, and a suitable balance of automation and human oversight to turn AI investment into results.
Home Depot’s Magic Apron AI fields millions of shopper questions
Home Depot says its Magic Apron AI assistant now handles millions of questions each month. The tool is designed to help shoppers get the information they need and avoid making a second trip to the store.
Alabama A&M initiative will teach AI skills to agricultural businesses
Alabama A&M University is joining Mississippi State University Extension Service and Mississippi State’s Southern Rural Development Center to lead AI for Answers. The initiative will train agricultural businesses and extension professionals to use AI for marketing, communications, program development, and educational outreach. Participants will create an individual AI roadmap for their operations. The program will also offer an AI and marketing microcredential pathway to help participants build skills and earn credentials.
Coontec adds Nvidia DGX B300 to strengthen AI security testing
Coontec said on Nov. 19 it will use Nvidia’s DGX B300 system to improve threat detection for ASH, its AI agent security platform. The system has eight Nvidia Blackwell Ultra GPUs, each with 288 GB of HBM3e memory, for more than 2 TB of total GPU memory. Coontec plans to use it to study large AI models and attack scenarios internally, then add findings to ASH and its AI supply chain checks. The company also plans to offer services that assess AI model risks before use.
GLiDE models deliver fast decisions with confidence scores
Researchers introduced GLiDE, a family of decision models ranging from 0.4B to 27B parameters. The models score candidate answers in one pass without generating text or intermediate reasoning, while using training methods that transfer reasoning skills from several teacher models. Across five sizes, GLiDE outperformed tested models of equal or smaller size on the Hugging Face performance and parameter comparison. The 27B model scored 71.52 with a 2.0% expected calibration error and a 92.9 ms median latency on one GPU; further model improvements raised its score to 73.37 with a 2.1% error.
AI stocks fall as Anthropic drops 5.76% after market debut
AI-related stocks fell in U.S. premarket trading on Nov. 20, one day after Anthropic went public, and Anthropic shares dropped 5.76%. SanDisk, Coherent, Lumentum, AMD, Intel, TSMC, and Nvidia also fell, while Nasdaq 100 futures declined more than 1.4%. Anthropic raised $81.3 billion at $83 a share and closed its first trading day at $97.17, giving it a fully diluted market value of about $1.8 trillion. Asian markets also fell, while Federal Reserve officials warned that AI investment and data center construction could add to inflation and overheating risks. Markets expected a quarter-point rate hike at the December meeting.
Sources
- Riemann Hypothesis In 41 Days: Levent Alpöge’s Timeline Shows How Fast Claude Helped Crack The 167-Year-Old Problem
- Anthropic Mathematician Says AI Has Solved the Most Famous Problem in Math
- The AI Riemann hypothesis proof is almost certainly true, independent mathematicians attest
- Anthropic Researcher Claims AI Has Solved Math’s White Whale
- Anthropic mathematician claims to have solved 167-year-old math problem with Claude — Riemann hypothesis proof allegedly discovered but remains unconfirmed
- HLTH 2026: 4 takeaways on AI’s next phase
- Anthropic Shares Are a Rational Bet on the Next Phase of AI
- Anthropic Stock Falls Sharply After Record IPO
- LATO Labs seed investment
- AI investment is widespread, but 3 in 4 leaders say decision-making is now more complex
- Home Depot's AI bot helps shoppers avoid the dreaded second trip
- New Alabama A&M University Initiative to Provide AI Training to Agricultural Businesses | The Journal of Blacks in Higher Education
- Coontec Adopts Nvidia DGX B300 for In-House AI Security Validation
- GLiDE: A Fast, Calibrated Decision Model Distilled from Reasoning Teachers
- Anthropic Plunges Nearly 6% in Premarket a Day After Debut as AI Stocks Retreat; Fed Officials Warn of Overheating Risks
Comments
Please log in to post a comment.