Daily TEA – AI Code Exploits, Quantum Racing, and Nuclear Powering AI
Meta AI overhaul, slopsquatting attacks, AI scribes, Bitcoin quantum threat, Rigetti 108-qubit, SMR data centers
Hello, dear TEA-mates! I’ve made a few changes to the Daily TEA newsletter! From now on, I’ll be curating the headlines and experimenting with a new format to better highlight the stories I find most interesting. You might notice some tweaks here and there, and I’d love any feedback as this evolves. Hope you enjoy the refreshed look starting today! :D
Here is what you need to know today:
~2 min read
🤖 AI
Meta debuts Muse Spark — Meta Superintelligence Labs’ first model under Alexandr Wang marks a “ground-up overhaul” of Meta AI across all its products.
Anthropic locks in gigawatts of next-gen TPU capacity from Google and Broadcom — The deal launches in 2027 as Claude Code hits a $1B annual run-rate in just six months, signaling massive scaling ambitions.
OpenAI launches Safety Fellowship for external alignment researchers — Program funds outside scholars in AI safety from September 2026 through February 2027, a push beyond internal research.
Synchronicity Labs releases Sync-3, its most advanced AI lipsync model — New model delivers state-of-the-art lip synchronization for production video use cases across multiple languages.
Slopsquatting: hackers are exploiting AI-hallucinated package names as supply chain attack vectors — 19.7% of AI-generated package names are hallucinated, and attackers register those names to deliver malware to developers who never verify them.
AI scribes are driving up healthcare costs — and neither side has a fix — Insurers and hospitals privately agree AI scribes inflate medical billing by generating more detailed documentation, but publicly blame each other with no regulatory solution in sight.
💰 Crypto / Fintech
Nobel physicist warns Bitcoin is quantum computing’s earliest and easiest real-world target — John Martinis, who helped build Google’s quantum computers, says a sufficiently advanced machine could crack Bitcoin encryption in minutes with a 5–10 year window to act.
Solana Foundation launches Accelerate AI, an exclusive May 6 Miami summit for AI and blockchain builders — Invite-only gathering signals Solana’s aggressive positioning as the settlement layer for agentic AI payments.
x402 on Solana hits 35M+ transactions and $10M+ in volume since summer launch — The AI agent payment protocol’s early traction confirms machine-to-machine micropayments are production-ready, not theoretical.
⚛️ Quantum + Energy
Rigetti launches Cepheus-1-108Q: the industry’s largest modular quantum computer — Now available via cloud, the 108-qubit system achieves 99.1% two-qubit gate fidelity using a proprietary chiplet architecture.
Quantum computing’s biggest reliability problem can now be tracked 100x faster — Norwegian researchers cut quantum error measurement time from one second to 10 milliseconds, enabling real-time monitoring of qubit stability.
Researchers run a fault-tolerant quantum algorithm with zero mid-circuit measurements — A Nature Communications study shows Grover’s search running on three logical qubits without pausing to measure, a new paradigm for quantum error correction.
NextEra and TerraPower announce a landmark SMR fleet to power Google and Microsoft AI data centers — A $15–20B partnership deploys Natrium sodium-cooled reactors in Iowa and the Southeast US, each producing 345–500 MW dedicated to AI workloads.
🛠️ Trending Tools Today
google-ai-edge/gallery — [GitHub] Google’s showcase app runs open-source AI models entirely on Android and iOS — no cloud, no internet, no data leaving your device; 853 stars today.
forrestchang/andrej-karpathy-skills — [GitHub] Karpathy-inspired CLAUDE.md guidelines that reduce AI overcomplexity and overengineering, installable as a Claude Code plugin; 702 stars today.
TheCraigHewitt/seomachine — [GitHub] A specialized Claude Code workspace for researching, writing, and optimizing long-form SEO content with GA4 and Search Console integration; 649 stars today.
⚡ Skill of the Day
Capability Evolver — The most downloaded skill on ClawhHub (35K installs), it gives your OpenClaw agents the ability to autonomously identify and improve their own capabilities over time.
Try it →
TEAHEE Moment
Stay sharp, stay informed. See you tomorrow.
If you enjoyed this TEA, follow along on social for more:
Twitter/X