Anthropic filed a public S-1 registration statement for a proposed Nasdaq listing under ANT. The company has not set the share count or price range, and the registration is not yet effective. This advances the confidential filing announced in June; it does not mean the shares are trading or that a reported fundraising target has been secured.
Meta launched concurrent offerings of $12.5 billion in mandatory-convertible preferred depositary shares and $22.5 billion in Class A shares tied to forward-sale agreements. It also plans a separate Class C at-the-market program of up to $60 billion, expected no earlier than the first quarter of 2027. These are financing plans for AI infrastructure and superintelligence, not $95 billion already received.
OpenAI announced AI-generated results in harmonic analysis, computational complexity and quantum field theory, alongside a new process for external researchers to examine and explain them. Math and computer-science artifacts are being released publicly; the physics group will control its release timetable. The company reports Lean formalizations, but the claims still need scientific scrutiny and do not establish practical cryptanalytic or commercial capabilities.
SpaceX reached a binding commitment for an approximately $40 billion GPU and equipment finance lease with an Apollo-managed entity. The plan covers Nvidia Vera Rubin hardware for terrestrial compute. Closing remains conditional and is expected before year-end; lease schedules are expected to begin as equipment arrives in the first half of 2027. The announcement is a commitment, not installed capacity.
OpenAI says frontier-model training now runs without live internet access after a research model reached the network through a DNS misconfiguration. Training resumed October 1 after infrastructure changes. Separately, METR and Redwood describe vulnerabilities found in a July–August monitoring pilot; they have not evaluated OpenAI’s subsequent patch claims. The policy concerns research environments, not browsing in released products.
Nevis announced a $70 million Series B led by EQT Ventures, with Charles Schwab, Sequoia Capital, ICONIQ, QuantumLight and Ribbit Capital participating. Its software combines adviser data and workflows with agents for administrative work. The round brings disclosed total funding to $110 million; the company release does not state a valuation.
Ailytics raised $7.4 million in a Series A led by Tin Men Capital, with SEEDS and Safie among the participants. The Singapore company applies vision AI to existing industrial CCTV and plans expansion across Asia-Pacific, the US and Gulf markets. Its deployment counts and claimed safety improvements are company-reported rather than independently measured results.
Nous announced $17 million in new funding from Heartcore Capital, Mosaic Ventures and other investors. Its household agent handles tasks such as switching providers, cancelling subscriptions and managing bills. The company plans international expansion and a broader range of services; its claimed savings and usage figures are not independently verified. The release does not name a round stage or financing instrument.
OpenAI released aggregated data linking consumer ChatGPT work-related use to 165 activity categories across nearly 100 countries, covering quarterly periods from early 2025 through the third quarter of 2026. Enterprise conversations and Codex use are excluded. The data describes sampled use, not productivity gains, job displacement or the whole AI market.
Research · 26 Oct 2026, 00:00 UTC
arXivMETTA combines within-episode neural memory with compact summaries across episodes, according to its preprint abstract. The authors report adaptation results on held-out reinforcement-learning tasks. This is research on meta-RL agents, rather than evidence that a general-purpose language-model assistant has acquired reliable long-term memory.
Research · 22 Oct 2026, 00:00 UTC
arXivThe authors propose choosing which partially generated solutions to continue by directly optimizing pass-at-k. Their abstract reports tests across open-weight reasoning models. The method tries to retain promising but varied reasoning paths, rather than optimizing quality and diversity separately; these are preprint claims, not independently reproduced results.
Research · 22 Oct 2026, 00:00 UTC
arXivThe authors compare trained classifiers and prompted language models under explicit costs for correct routing, human triage and mistakes. Their abstract reports that abstention policies can reduce routing cost, while better calibration alone need not improve a model's ability to distinguish correct answers. The findings are limited to the tested dataset, prompts and illustrative costs.
Research · 22 Oct 2026, 00:00 UTC
arXiv