Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
/Dispatch #108
Meta reversed course again, open-sourcing Muse Glimmer, a 30B model that runs AI agents on a laptop. Zuckerberg's 6,500-word manifesto pitched "personal superintelligence" against rivals' closed models.
Gemini crossed 1 billion monthly users, Google's fastest product ever and just ten weeks after ChatGPT. But 3.5 Pro is late and DeepMind just lost Hassabis.
+11DeepSeek, the lab that started the AI price war, warned clients of a steep price hike as demand outstrips its compute. The subsidized near-free era may be ending.
+3Amazon auto-enrolled every Twitch streamer into AI training, opt-out by default. A product chief admitted why: "If this was opt-in, nobody would opt in."
+8OpenAI halts Astra AI model development over critical cybersecurity threshold concerns
Tl;dr: OpenAI froze a model that could hack on its own.
Chinese Hackers Use AI Agents in First Autonomous Cyberattack on Taiwan, Stealing 2,500+ Records
Tl;dr: The first fully autonomous hack just hit a government.
Major AI Model Flaw Let Weaker Models Decode Encrypted Reasoning, Exposing 62 API Keys and Passwords
Tl;dr: Cheap models could read Claude's encrypted thoughts.
