Anthropic and OpenAI Court Wall Street as Retracted Study Undercuts AI's School Push
1. Anthropic and OpenAI Each Built Wall Street JVs This Week to Sell Their Enterprise AI On May 4, Anthropic announced a joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs to build "a new enterprise AI services company.
2. "Notepad++ has never released a macOS version" — its creator had to say it out loud Don Ho posted a clarification this week that he should not have needed to write.
3. The study used to prove ChatGPT helps students just got retracted. Same week, OpenAI, Google and Microsoft pushed AI into schools. An influential study claiming ChatGPT improves student learning was retracted this week over data red flags, according to Ars Technica.
In Brief
- Cerebras files for IPO at $26.6 billion valuation with OpenAI as anchor customer The AI chip maker is heading to public markets with a commercial relationship that includes compute supply contracts and joint engineering work with OpenAI. The filing follows a year of surging demand for inference-optimized hardware.
- Image model launches drive 6.5x more app downloads than chatbot upgrades Appfigures tracked install spikes after major AI releases and found visual models pull far heavier traffic than text updates. Most apps fail to convert the spike into paying revenue.
- OpenAI published the WebRTC stack rebuild behind its Voice mode The post details how the team handles turn-taking, audio routing, and global edge deployment for real-time conversation. It is the first technical write-up since Voice Mode rolled to all paying users.
- OpenAI submitted Musk texts saying Altman and Brockman would be "the most hated men in America" The court filings show Musk sent the messages after asking for settlement talks days before trial began. OpenAI is using them to argue Musk tried to coerce a deal.
- Musk's only AI expert witness urged governments to restrict frontier labs Berkeley professor Stuart Russell took the stand warning of an AGI arms race. Russell has spent years arguing for binding caps on frontier training compute.
- Google added webhooks to the Gemini API for long-running jobs The push-based notifications replace polling for batch generations and async tasks. Developers receive a callback when a job finishes instead of repeatedly hitting the status endpoint.
- DoorDash shipped AI tools for merchant onboarding and dish photo edits The features let restaurants set up storefronts faster, retouch food photography, and generate websites from existing menus. The rollout targets independent operators that lack in-house marketing.
- DeepClaude wraps the Claude Code agent loop around DeepSeek V4 Pro The open-source project lets developers run Claude Code's terminal interface against DeepSeek's model instead of an Anthropic key. The repo hit Hacker News with working code over the weekend.
- Roomba creator Colin Angle launched a dog-sized companion robot His new company Familiar Machines & Magic shipped its first product as a furry pet rather than a cleaner. Angle previously put 50 million household robots into homes.
- Canadian election database caught a leaker with a canary trap Investigators planted intentional small errors unique to each authorized recipient. When a copy leaked, the fingerprint identified the source and led to charges.
Don't miss what's next. Subscribe to AI News Digest: