1. "I cancelled Claude" hit 957 points on Hacker News the same week an AI agent deleted a production database Nicky Reinert published a post titled "I cancelled Claude: Token issues, declining quality, and poor support" on his personal blog on April 24.

2. A 13-acre Mill Valley listing now accepts Anthropic stock — days after Google committed $40B Anthropic's valuation is being marked the same week from two ends of the capital stack. Google said it will invest as much as $40 billion in the company, according to Ars Technica.

3. Vibe maths cracked a 60-year-old Erdős problem, the same week the AI industry conceded the public hates it An amateur without formal mathematical training used ChatGPT to make substantive progress on a question Paul Erdős posed about 60 years ago, Scientific American reported.

In Brief

• DeepSeek previews V4 model claiming parity with frontier systems DeepSeek released a preview of two new models built on architectural changes from V3.2, citing efficiency gains and reasoning benchmark scores close to leading open and closed models. The company has not disclosed parameter counts or release timing for the full version.

• Sam Altman publishes five OpenAI principles Altman posted a five-point statement framing OpenAI's mission around AGI benefiting humanity, the company's first formal principles document since recent governance disputes. The post does not include enforcement mechanisms or board-level commitments tied to the principles.

• John Ternus succeeds Tim Cook as Apple CEO Ternus, Apple's hardware chief, takes over with a mandate to ship a competitive AI product after Cook's tenure ended without one. Analysts read the hardware-engineer pick as Apple returning device design to the strategic center.

• Discord researchers breached Anthropic's Mythos model A group operating through Discord obtained unauthorized access to Anthropic's Mythos system, the same model the NSA is reportedly running. Wired's writeup also flags a 500,000-record UK health data sale on Alibaba and an Apple notification leak patch.

• Uber CTO joins StrictlyVC SF on April 30 Praveen Neppalli Naga is added to the TechCrunch StrictlyVC lineup at Sentro Filipino Cultural Center to discuss running Uber's infrastructure under AI workloads. The event opens TechCrunch's 2026 conference calendar.

• Google ships Gemini home and inbox organization prompts Google published eight Gemini prompt templates for cleaning schedules, inbox triage, and seasonal household chores, pushing the assistant into routine personal-management tasks. The set targets consumers rather than developer or enterprise workflows.

• Ace table tennis robot rallies against human players Ace tracks ball trajectory, adjusts racket angle, and returns shots fast enough to sustain rallies with amateur opponents. The system targets competitive table tennis as a benchmark for real-time perception and actuation.

• Wired warns against chatbot financial advice The piece lists five failure modes when users treat ChatGPT and similar tools as financial advisors: outdated data, hallucinated tax rules, no fiduciary duty, missing personal context, and confident wrong answers. It arrives as more retail users substitute chatbots for licensed planners.

• SusHi Tech Tokyo 2026 sets four-domain agenda Organizers structured the conference around four technology tracks with live demos, dedicated floors, and sessions led by operators and investors. TechCrunch frames Tokyo as the year's central tech destination after Japan's recent AI policy moves.

Read the full edition →