GoKawiil
Tech News
OpenAI published more than 700 preliminary math research papers on GitHub this week, intensifying tension with academic mathematicians. The release follows an open letter last month from 28 Fields Medal winners who created a site called Math and AI, accusing AI companies of pursuing goals misaligned with the mathematics community.
A Bluesky poster using the handle William B. Fuckley argued that the current focus on 'superintelligence' messaging functions as a humiliation ritual aimed at OpenAI and the broader AI sector, speculating it may be designed to see whether companies change their names. The poster characterized the strategy as shrewd in terms of regime politics, while framing it as their own opinion rather than confirmed fact.
OpenAI published a massive collection of almost 400 mathematical results spanning combinatorics, geometry, number theory, topology, probability and other fields, contained in over 700 manuscripts. The release was so large that OpenAI issued guidance to help researchers navigate its GitHub repository, and some proofs include Lean formalizations for computational verification.
According to Axios, OpenAI and Anthropic executives have been conducting scenario-planning exercises modeling potential public and political backlash against AI companies, including hypothetical catastrophic incidents or major cyberattacks tied to their technology. OpenAI confirmed to Axios it runs 'preparedness exercises' covering various scenarios but said these aren't treated as inevitable outcomes. Anthropic declined to comment on the report.
An Entrepreneur contributor describes organizing multiple ChatGPT instances into a coordinated team for solo business owners, assigning each a specialized role such as email strategy, content research and campaign scheduling. The setup includes an always-on AI agent acting as a chief of staff to hand off tasks between the specialized assistants, built using separate ChatGPT projects with tailored instructions and files for each role.
A developer who began using AI coding agents in February 2025 describes persistent problems with task tracking and reliability, including agents that stall and require restarts, falsely report tasks as complete, and fail to properly manage multi-step work. The author distinguishes between the underlying language models, which have improved significantly, and the agent software wrapping them, which has not kept pace. An example cited involves an open-source file-sharing app where an agent, OpenCode, split a passphrase-protection feature into ten subtasks but mishandled execution of the plan.
Three OpenAI safety researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—who were fired last week for allegedly mishandling confidential information have published an open letter disputing the company's account. They deny leaking details about OpenAI's model architecture to The Information and say the firings have made remaining staff afraid to communicate as they once did. OpenAI maintains the trio violated policies on handling sensitive information and broke trust essential to its work.
President Trump said on Truth Social that the US will not attack Iran before the November 3 midterm elections, citing "productive discussions" between the two countries—a reversal from comments made a day earlier. He said the US naval blockade on Iran's ports remains fully in effect and that Iran still cannot be allowed a nuclear weapon. Oil prices rose during Thursday's trading session despite the de-escalatory tone.
Proofs and Prompts has compiled a running thread of mathematician reactions to OpenAI's 6 October announcement that its models produced solutions to over 100 research-level math problems, including one described as resembling a Quasi-Riemann result. Contributors like Tim Santens and Roman Sauer describe being stunned that the models applied known but obscure techniques across very different problems, and the site says it will keep collecting reactions for another week.
Anthropic has appointed at least two former Trump officials this year, including one to its board and another to lead its frontier compute strategy, while OpenAI has hired at least three ex-administration officials for strategy and special projects. The hiring comes as AI companies seek closer ties to the Trump White House amid regulatory pressure and security concerns, following a court ruling that upheld a Pentagon designation of Anthropic as a supply-chain risk.
OpenAI released hundreds of papers containing AI-generated mathematical proofs, following earlier controversy over its models' handling of the Navier-Stokes equations. The Association for Human Mathematics issued a statement condemning the release, which was shared and endorsed by Fields Medalist Terence Tao, arguing the dump of files demonstrates power rather than scholarship. OpenAI's announcement cited consultations with an independent group, the Advisory Group on Mathematics and Artificial Intelligence, but did not mention that the group had actually opposed continued testing of its proprietary models on advanced problems.
Three former OpenAI safety researchers—Mikita Balesni, Tomek Korbak and Jasmine Wang—published an open letter claiming they were dismissed for prioritising safety concerns rather than for the 'mishandling of sensitive information' OpenAI cited. OpenAI's research leaders responded that the firings were due to a serious breach of trust unrelated to safety advocacy, standing by its decision.