Are OpenAI and Anthropic Moving Away From Reasoning Tech?

Elon Musk rocked the business world again by announcing Tuesday that his rocket-satellite-social media firm SpaceX has signed a deal agreeing to buy AI coding app Cursor for $60 billion at an undisclosed later date or else SpaceX would pay a $10 billion breakup fee.

It’s one of the biggest acquisitions of a venture backed startup ever, though it’s not much money in the context of an initial public offering that could value SpaceX at $1.5 trillion this summer.  

Cursor would give xAI a substantial AI application business—recently on pace to generate billions of dollars in revenue annually—in a market that xAI has struggled to crack with its Grok product.

The question is whether xAI and Cursor will be able to develop good-enough coding models, and a sticky enough product, to avoid being lapped by rivals Claude Code and Codex.

On to the column…

Giving models more computational power to think through tough problems, known as reasoning, has been one of the biggest AI research breakthroughs of the last two years.

Researchers also refer to it as test-time compute.