AI Math Claims Remain Unverified as Apple Develops Agents

Several high-profile mathematical claims involving AI remain unverified. An unsigned four-page note on Levent Alpöge’s website says Claude helped prove the Riemann Hypothesis and its generalized form, describing a strategy involving theta lifts and saying Lean code is available. Separately, an Anthropic mathematician says the company’s non-public Claude helped produce a Lean proof exceeding 250,000 lines. The claims have not been independently checked; Terrance Tao says review could take months or longer. The unsigned note has not been published, and some of its steps are described as open questions or personal intuitions.

A separate, non-AI Riemann Hypothesis draft by Alessandro Masullo was released by American Mathematical Society President Ravi Vakil. Stanford mathematician Brian Conrad has reviewed it for months and said it was looking good, but has not issued a final verdict. Vakil released the draft earlier than planned and asked mathematicians to check and formalize it. Another disputed claim concerns the three-dimensional Navier–Stokes equations: Tristan Buckmaster and Levent Alpöge’s 79-page paper says smooth solutions can blow up without an external force. Its computer-assisted proof relies on a numerical certificate that has not been released, and the paper has not undergone peer review. Both mathematical claims still require scrutiny.

Apple is hiring a senior engineering leader to set direction for general-purpose AI agents that can plan and perform tasks across apps, browsers, command lines, code and APIs, coordinate specialized agents, and operate with privacy and safety in mind. The work involves Siri, UI frameworks and Apple’s AFM. A separate report says a Mac agent that can browse websites, use apps and write or debug code may launch by the end of 2027, though the project is early and its features and timing are unconfirmed. Bloomberg also reports macOS 27.3 may warn users before third-party agents receive broad device control.

Other coverage examines where AI is already being deployed and what its costs and effects may be. MIT FutureTech researchers Damian Saccocio and Neil Thompson argue adoption depends on whether AI’s full cost—including checking and correcting its output—is below the cost of human labor. They expect AI more often to assist workers than replace whole occupations at once, while noting automation can push wages in either direction. A study comparing 19 U.S. industries found job cuts and AI use rose together in several sectors, including aerospace and defense, whose combined score was 97.9 out of 100; transportation cuts rose 189.8%, from about 15,300 to 44,400. The study does not establish that AI caused layoffs, and construction saw adoption rise 85.5% as cuts fell 27.6%.

At the AASHTO 2026 Annual Meeting in Louisville, transportation officials described partnerships and pilots: Michigan is testing full-size automated buses with Michigan State University, Texas is testing automated freight vehicles and data-sharing, and Utah has connected its Transportation Management Center to traffic signals using AI. In financial advice, one firm linked meeting notes, client records and public information to identify a client’s retirement plans and assets; the resulting review moved the retirement target forward two years and brought $2.1 million in assets under management. The article urges firms to set 90-day goals for three revenue-focused AI uses, measure results and discontinue what does not work.

Governance and access also feature in the coverage. At a September 2026 Bonn Symposium session, Clara Iglesias Keller and Marie Kilg called for human-rights assessments, transparency and public oversight, arguing that people—not technology—must be accountable and that private firms wield too much power while laws lag. Kilg supports regional and multilateral agreements tailored to countries’ resources and economic conditions; the African Observatory on Responsible Artificial Intelligence is cited as an initial step, though enforcement institutions remain lacking in many places. A Deutsche Bank report by Adrian Cox, published Nov. 19, 2026, says India could benefit from AI without bearing the large upfront costs of building it, and argues its practical test will be in cities such as Mumbai. Separately, an image falsely presented as a college photo of Nvidia co-founder Jensen Huang and his wife Lori was generated by OpenAI’s DALL-E, according to OpenAI’s verification tool; it does not match their 1984 graduation photo and circulated after reports of Huang’s admitted affair with a Chinese executive accused of spying for Beijing.

Key Takeaways

  • Anthropic’s non-public Claude is said to have helped produce a Lean proof of the Riemann Hypothesis exceeding 250,000 lines; experts say verification may take months or longer.
  • An unsigned note on Levent Alpöge’s website also claims a Claude-assisted proof of the Riemann Hypothesis and its generalized form, but it is unpublished and unchecked.
  • Alessandro Masullo’s separate, non-AI Riemann Hypothesis draft has been reviewed by Brian Conrad, who has not given a final verdict; Ravi Vakil released it early for further checking and formalization.
  • Tristan Buckmaster and Levent Alpöge’s 79-page Navier–Stokes paper claims smooth solutions can blow up, but its numerical certificate is unavailable and the work has not been peer-reviewed.
  • Apple is developing general-purpose AI agents across apps, browsers, command lines, code and APIs; a Mac agent is reportedly targeted for late 2027, but remains early-stage and unconfirmed.
  • MIT researchers say AI adoption depends on total cost and reliability, including the human work needed to check and correct results; AI may assist workers rather than replace whole occupations at once.
  • A 19-industry study found AI use and job cuts rising together in some sectors, including aerospace and defense, but it does not establish causation; construction saw AI adoption rise while cuts fell.
  • Transportation agencies reported AI and automated-vehicle pilots in Michigan, Texas and Utah, including automated buses, freight tests and traffic-signal network connections.
  • A financial advisory firm’s AI-supported client analysis preceded a plan update, a retirement target two years earlier and $2.1 million in assets under management; the article recommends measurable 90-day goals.
  • Experts called for human-rights assessments and public oversight of AI; a Deutsche Bank report says India could gain from AI without the high upfront costs, while OpenAI’s tool identified a viral image of Nvidia’s Jensen Huang and his wife as DALL-E-generated.

Mathematicians Claim Navier Stokes Breakthrough but Key Proof Check Is Missing

Tristan Buckmaster and Levent Alpöge say their 79 page paper proves that smooth solutions to the three dimensional Navier Stokes equations can blow up without an external force. The paper, dated November 20, uses a computer assisted proof involving a self similar fluid pattern and a numerical certificate. The certificate has not been released, and the interval arithmetic check described in the README is not included in the repository or Lean formalization. The work has not received peer review, so independent experts still need to examine it before the claim can be settled.

Unsigned Note Claims Claude Helped Prove the Riemann Hypothesis

An unsigned four page note on Levent Alpöge’s website claims a proof of the Riemann Hypothesis and its generalized form, developed with Claude. It presents a strategy based on theta lifts and ideas from modern number theory, with a timeline running from early October to November 16, 2026. The note says a Lean code version is available, but no paper has been published and the claim has not been checked. Alpöge has not shared the note himself, and it describes some steps as open questions or personal intuitions.

Ravi Vakil Releases Alessandro Masullo’s Human Riemann Hypothesis Proof Draft

Hours after a claim of an AI assisted proof appeared, American Mathematical Society President Ravi Vakil announced a separate proof of the Riemann Hypothesis by Alessandro Masullo, stressing that it used no AI. Vakil said Stanford number theorist Brian Conrad had reviewed the draft for months and that it was looking good, but had not given a final verdict. The plan had been to bring Masullo to Stanford for a year to improve the paper and seek more expert review, but Vakil released it early and asked mathematicians to check and formalize it. The draft has not yet been formalized, and both proof claims still need further checking.

Apple Job Listing Reveals Work on Computer Using AI Agents

A job listing for a senior engineering leader says Apple is building general purpose AI agents that can work across apps, browsers, command lines, code and APIs. The agents would plan and carry out tasks, choose tools and coordinate specialized sub agents, while working privately and safely. The leader would set the technical direction for the agent platform and work with teams including Siri, UI frameworks and AFM. The listing suggests Apple is developing a product in a field where competing AI agents are already available to consumers.

Apple Reportedly Plans a Mac AI Agent for Late 2027

Apple is reportedly developing an AI agent for Mac that could browse websites, use apps, run commands and write and debug code for users. A recent job listing describes agents that can handle multi step tasks across apps, browsers, command lines and APIs. The project is still in its early stages, and Apple may aim to launch it by the end of 2027, though no exact release date or final features have been announced. Bloomberg also reported that macOS 27.3 may warn users before they give third party agents broad control of their devices, following added security protections announced to developers in October 2026.

AI image falsely shown as Jensen Huang and his wife in college

An image shared online as a college photo of Nvidia co-founder Jensen Huang and his wife, Lori, is not real. OpenAI’s verification tool found that its DALL-E model created the image. It does not match a photo of the couple from Huang’s 1984 graduation. The fake image circulated after news of Huang’s admitted affair with a Chinese executive accused of spying for Beijing.

MIT researchers say AI’s impact on jobs depends on its cost

MIT FutureTech researchers Damian Saccocio and Neil Thompson argue that companies will adopt AI when it costs less than human labor, not simply when it can perform a task. AI’s full cost includes the work people must do to check and correct its results, and moving from usually correct to reliably correct can be expensive. The authors say AI is more likely to assist people than replace entire occupations at once. Automation can also raise or lower wages, depending on which parts of a job it takes over. They urge business leaders and policymakers to track AI’s cost and reliability, while workers and employers consider how automation changes the remaining work.

State transportation agencies prepare for AI and automated vehicles

At the American Association of State Highway and Transportation Officials 2026 Annual Meeting in Louisville, transportation officials discussed how AI and automated vehicles are changing their work. AECOM’s Suzanne Murtha moderated a panel with leaders from transportation departments in Utah, California, Texas, Michigan and Georgia. Michigan’s Kimberly Zimmer said agencies need partnerships with universities and private companies to test connected and automated vehicles, including full size automated buses at Michigan State University. Texas is testing automated freight vehicles and working on ways to share data with them. Utah’s Lisa Wilson said AI has connected the state’s Transportation Management Center to its traffic signal network.

Financial firms urged to measure AI’s effect on growth

The article argues that financial advisory firms should measure whether AI brings in revenue instead of focusing only on tools such as meeting summaries and email drafts. It describes an advisory firm that used AI to connect meeting notes, client records and public information, uncovering a client’s retirement plans and assets. The resulting conversation led to a financial plan update, moved the client’s retirement target up by two years and brought $2.1 million in assets under the firm’s management. The author recommends choosing three AI uses tied to revenue, setting goals for 90 days and measuring the results. Firms should expand what works and stop what does not.

Anthropic mathematician claims AI helped prove Riemann hypothesis

A mathematician at Anthropic says the company’s non-public version of Claude helped prove the Riemann hypothesis, a famous problem about prime numbers that has remained unsolved for nearly 170 years. The proposed proof, written in the proof-checking language Lean, runs to more than 250,000 lines. Mathematician Terrance Tao said reviewing it will take time, and experts may need months or longer to assess the claim. The announcement follows OpenAI’s earlier claim that its AI tools helped solve the Navier-Stokes equations, a result that has not been confirmed. Scholars have also raised concerns about AI’s effects on intellectual work.

India may gain from AI without paying the high upfront costs

A Deutsche Bank report by analyst Adrian Cox names India as a country that could benefit greatly from the artificial intelligence revolution. The report appeared Thursday, Nov. 19, 2026, alongside Anthropic’s high-profile Nasdaq listing. It argues that India could gain from AI without making the large upfront investments required to build the technology. Cox says AI’s real test will be in Mumbai and other cities, not only in San Francisco’s frontier labs.

Experts call for human rights rules to govern artificial intelligence

At a Bonn Symposium session in September 2026, Brazilian legal scholar Clara Iglesias Keller and German AI expert Marie Kilg argued that people, not technology, should be held accountable for AI’s effects. They warned that a few private companies have too much power and that laws are lagging behind, leaving AI governance largely to voluntary rules. Keller and Kilg called for human rights assessments, transparency and public oversight, and said product regulation could help shape AI rules. Kilg supports regional and multilateral agreements that account for differences in countries’ public resources and economic conditions. The article also notes the African Observatory on Responsible Artificial Intelligence as a positive first step for tracking practices, while saying many places still lack an institution to enforce existing policies.

AI spending grows as job cuts rise in several US industries

A study by HR.software compared announced job cuts in 19 US industries from January to September 2025 and 2026, along with changes in AI adoption and spending. Aerospace and defense had the highest combined AI Adoption & Job Cut Score, at 97.9 out of 100, as job cuts rose 132.1 percent and AI adoption rose 66.7 percent. Transportation job cuts increased 189.8 percent, from about 15,300 to 44,400, while AI adoption also grew. Chemical, food manufacturing and industrial goods sectors also recorded rising cuts alongside greater AI use or spending. The study shows a correlation, not that AI caused the layoffs, and construction is a counterexample: AI adoption grew 85.5 percent while job cuts fell 27.6 percent.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI-assisted mathematical proofs Riemann Hypothesis Navier–Stokes equations Lean theorem proving Claude Anthropic AI agents Apple Intelligence Siri AI and employment AI adoption costs AI in transportation automated vehicles AI in financial advice AI governance human rights AI oversight AI in India DALL-E AI-generated misinformation

Comments

Loading...