OpenAI's Model Solves an 80-Year-Old Math Problem — Refuting a Central Conjecture in Discrete Geometry

OpenAI's reasoning model has produced a proof that refutes a central conjecture related to the "unit distance problem" in discrete geometry, which had remained unsolved since 1946. According to OpenAI's official announcement, this achievement marks a milestone in AI-driven mathematical discovery.

According to TechCrunch, the last time OpenAI claimed a mathematical breakthrough, it ended in an embarrassing retreat after mathematicians sharply rebutted it. This time is markedly different: the very mathematicians who pointed out problems with that earlier work have provided the verification. It is a significant step, suggesting that the model's reasoning capabilities are now making substantive contributions to mathematical research.

NVIDIA's Revenue Forecast Lands at the Low End of Analyst Expectations Amid Intensifying Chip Competition

NVIDIA, the world's most valuable company by market capitalization, has released its quarterly revenue forecast. The revenue outlook for the July quarter came in at around $91 billion. While that beat the analyst consensus of $87 billion, the top of the estimate range stood at $96 billion, and the reaction from investors was largely lukewarm.

According to Bloomberg, data center revenue continues to surge, but intensifying chip competition is being recognized as a source of concern. While demand for AI semiconductors itself remains solid, the rise of competitors and customers moving to develop chips in-house are increasingly seen as risks of slower future growth.

Google Publishes Exploit Code for an Unpatched Chromium Vulnerability — Left Unfixed for 29 Months

Google has published proof-of-concept code for an unpatched vulnerability affecting Chromium-based browsers (including Chrome and Edge). According to Ars Technica, the vulnerability exploits the Browser Fetch API and has remained unfixed in the 29 months since it was reported.

An attacker exploiting this vulnerability can monitor a user's browsing activity or use their browser as a proxy to relay DDoS attacks. Connections can persist even after the browser or device is restarted. In effect, there is a risk that a device becomes part of a limited botnet without the user's knowledge.

Google's decision to release the exploit code before a patch is available appears to be an effort to pressure for a higher fix priority. With hundreds of millions of users on Chromium-based browsers such as Chrome and Edge, the scale of the potential impact is a serious concern.

Socket Raises $60 Million at a $1 Billion Valuation — Security Demand in the Era of AI Coding

Socket, a startup that protects the security of open source code, has closed a $60 million funding round led by Thrive Capital, reaching a $1 billion valuation. According to Bloomberg, Andreessen Horowitz and Abstract Ventures participated as existing investors, with Capital One Ventures joining as a new investor.

The company's service has taken on new importance with the spread of AI coding tools offered by Anthropic, OpenAI, Cursor, and others. Because these AI systems train on open source code repositories, the impact of vulnerabilities lurking in that code is magnified. In fact, Anthropic and Cursor are Socket customers. It is fair to say that investors have recognized the importance of supply chain security as AI development accelerates.

Ramp Speeds Up Code Reviews with Codex × GPT-5.5

Ramp's engineering team has shared a code review approach that combines OpenAI's Codex and GPT-5.5. According to OpenAI's official case study, the process of obtaining review feedback, which previously took hours, was cut down to minutes.

AI-powered code review is entering practical use, contributing not only to developer productivity but also to maintaining consistency in code quality. Notably, Codex's autonomous code improvement suggestions directly reduce the review burden on engineers.

---