1. xAI trained Grok on OpenAI's models, Musk testified. Then his lawyer fumbled the next witness's identity. Elon Musk testified Thursday in a California federal court that xAI used OpenAI's models to improve its own, The Verge reported.

2. Zig Bans AI-Generated PRs as Copilot and Claude Code Quietly Add New Restrictions For most of the past year, developers using AI coding tools worked under an unwritten contract: contributions were welcome, reviews came bundled in the subscription, and the assistant didn't care

3. Goodfire turned "open the model and twist a parameter" into a product The pitch from Goodfire's San Francisco office to model engineers is direct: stop guessing why your training run produced that output. Look at the parameters that fired. Change them.

In Brief

• Anthropic targets $900B valuation in two-week fundraise Anthropic asked investors to submit allocations within 48 hours for a round that could close at over $900 billion within two weeks, per sources. The pace would push Anthropic's private valuation past most public AI competitors.

• Stripe opens Link wallet to autonomous AI agents Stripe launched Link, a digital wallet that lets users connect cards, banks, and subscriptions, then authorize AI agents to spend through approval flows. The product gives agentic shopping a payments rail with explicit consent gates.

• Mistral ships Medium 3.5 with remote agents Mistral released Medium 3.5 alongside "vibe remote agents," extending the model into longer-running task execution. The release positions Mistral against Anthropic and OpenAI on agent infrastructure rather than raw benchmark scores.

• Google puts Gemini in millions of vehicles Google began deploying Gemini as the in-vehicle assistant across millions of cars, replacing earlier voice systems. The integration moves Gemini into a default-on surface drivers cannot easily swap out.

• SoftBank forms data-center robotics arm eyeing $100B IPO SoftBank is spinning up a robotics company focused on building data centers and is already targeting a $100 billion IPO. The unit feeds back into Masayoshi Son's AI infrastructure thesis.

• OpenAI restricts GPT-5.5 Cyber to vetted defenders OpenAI will release its cybersecurity testing tool GPT-5.5 Cyber only to "critical cyber defenders" at first. The gating arrives after OpenAI publicly criticized Anthropic for limiting access to its Mythos model.

• Legora hits $5.6B as Harvey fight escalates Legora reached a $5.6 billion valuation as it pushed deeper into Harvey's territory, with the two startups running dueling ad campaigns. Both have raised hundreds of millions and now compete directly for big-firm contracts.

• Apple says AI demand caught it short on Macs Apple told investors AI workloads drove unexpected demand for Mac mini, Studio, and the new Neo, and supply will stay constrained through next quarter. The Mac line is now selling as local-inference hardware.

• X rebuilds ad platform around AI X launched a rebuilt ads platform that uses AI for targeting and creative generation as it tries to grow revenue again. The relaunch arrives while advertisers continue to weigh brand-safety risk on the platform.

• Spotify adds "Verified" badge to flag human artists Spotify launched a verification program with a green checkmark for profiles confirmed to have a real person behind the music. AI personas will not qualify for the badge at launch.

Read the full edition →