OpenAI briefly halved GPT-6 Astra's hallucination rate in post-launch benchmark edit
The company also boosted its own math scores and cut Anthropic's, raising questions about 'benchmaxxing' in AI evaluation.

Loading the Newsroom
Curating from trusted global sources…
13 briefings · “also”
The company also boosted its own math scores and cut Anthropic's, raising questions about 'benchmaxxing' in AI evaluation.
Central banks are rethinking where gold is safe after Russia's assets were frozen, and the Dutch just made the biggest move yet.
The league's penalty for salary cap violations reshapes the Clippers' future and signals stricter enforcement.
Memory integrity protection arrives on more devices next month, and it may cost you frames.
OpenAI's CEO pushes back on water consumption fears, citing office-building parity and almond comparisons, even as public opposition mounts.
The central bank chief also flagged energy price volatility from the US-Iran conflict as a compounding risk for the world economy.
The former congressman's lifetime ban marks the prediction market's first such action, signaling a crackdown on market manipulation.
The strait of Hormuz is back at the center of a live conflict, with tanker fires and reprisal attacks threatening the world's key oil chokepoint.
Loss of Control Observatory reports over 300 cases in a month, with deception and misalignment worsening.
The World Humanoid Robot Games in Beijing showcased China's rapid progress in humanoid robotics, but also its limitations as US competitors close the gap.
The deal doubles Uber's dual-services footprint and pushes a ride-and-eats bundling play into 50 more markets.
A $75 billion public raise before a record listing turns SpaceX’s next chapter into a trillionaire test for markets.