OpenAI says AI solved major math problem in 88 hours; credit dispute follows

The newsletter said OpenAI used 10,000 AI agents to tackle the Navier-Stokes equations, a 90-year-old problem, and spent millions of dollars in pursuit of a $1 million contest. Whether the credit accusations are true or not, the episode may mark a turning point in the history of mathematics, because AI models now appear essential for making progress on the field's biggest problems, and solving them may require resources available only to a few frontier AI companies. That raises questions about how human mathematicians will fit in, MIT Technology Review said.

In the same newsletter, the outlet reported that battery installations in the United States hit a record in the second quarter of 2026, with 20.2 gigawatt-hours of new capacity coming online, enough to supply the daily electricity needs of 600,000 homes. Cheaper batteries and the urgent need for energy storage as more renewables are added to the grid are driving the surge, putting the country on track for another record year. The report also noted that grid-scale and residential battery markets are changing in different ways.

The newsletter also highlighted Danijar Hafner, an honoree on MIT Technology Review's 35 Innovators Under 35 list, who is developing AI agents capable of planning ahead. Hafner, 31, builds world models and previously tested his agents on video games before moving them into physical environments not seen in training, according to the report.

On drone data, MIT Technology Review reported that Ukraine is sharing millions of data points gathered from tens of thousands of drone flights with military contractors and commercial companies. It said that creates a new marketplace for battlefield data that turns the front line into a model training site, and called for a regulatory system so that such data is not treated as ordinary commercial material.

In its roundup of other technology news, the newsletter said the United States has accused six Chinese AI companies, including DeepSeek, Moonshot AI, Alibaba and Z.AI, of industrial-scale theft of American AI trade secrets through model distillation, with officials saying the companies targeted systems such as Claude, ChatGPT, Gemini and Grok. It also said the Pentagon asked OpenAI for a model with "minimal refusal rates," and that Apple was expected to unveil a $2,000 folding smartphone, a major design shift since the 2007 iPhone and a first big test for CEO John Ternus. Xiaomi and Huawei had launched their own foldables before Apple's event, the newsletter noted.