1. Alibaba bans employees from using Claude Code, pushes Qoder — Classified as high-risk software; ban starts July 10. Anthropic had been running experiments to identify Chinese users through Claude Code. Alibaba directs staff to use their own Qoder tool instead. TechCrunch · Jul 4
2. Claude Code session/cache leakage between workspace instances — GitHub issue #74066, 267 HN points. Potential data bleed between sessions raises security concerns for a tool reading entire codebases. Anthropic investigating. GitHub · Jul 4
3. GPT-5.5 Codex reasoning-token clustering may degrade performance — OpenAI issue #30364, 107 HN points. Token clustering at batch sizes 516/1034/1552 causing degraded results on complex tasks. Engineers investigating. GitHub · Jul 4
4. Better Models: Worse Tools — Armin Ronacher on Claude regression — Tool-calling regression in newer Claude models: models improve on benchmarks but get worse at reliably using tools. Essential read for agent pipeline builders. 74 HN points. lucumr.pocoo.org · Jul 4
5. Midjourney demands Hollywood studios reveal their AI usage — In copyright lawsuit, Midjourney seeks full disclosure of studio AI practices including all prompts and outputs. Judge already ordered consumer-facing AI disclosure; Midjourney wants the full picture. TechCrunch · Jul 4
6. Fable releases novel 4D splat format (.splat4d) — Four-dimensional Gaussian splatting captures 3D space + time in a single representation, enabling navigable holographic video. 95 HN points. splats4D · Jul 4
7. AI-built PHP engine in Rust (Phargo) renders WordPress — Developer who doesn't know Rust used AI to build a PHP interpreter that passes 17% of PHP's test suite and renders WordPress. Open source with auto-published test results. ekinertac.com · Jul 4
8. Fanfiction community at war with AI — and itself — Archive of Our Own writers use AI detectors to flag suspected AI-generated fics; detectors prove unreliable and the community splits over policing vs. witch hunts. The Verge · Jul 4

---

🧪 Try This Next

• 🔧 Read Armin Ronacher's "Better Models: Worse Tools" for agent pipeline builders
• 🎥 Check out the Phargo repo — the most honest AI coding experiment out there
• 📝 If using Claude Code professionally, review the session leakage issue on GitHub