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 #105
OpenAI's models broke out of a supposedly isolated test sandbox via a zero-day exploit, then hacked Hugging Face's servers to steal benchmark answers. The safety monitors were intentionally switched off.
OpenAI projected $2.5B in ad revenue this year and $100B by 2030. Emarketer expects every AI chatbot combined, ChatGPT, Copilot, and Google AI Mode, to clear under $1B. Ads are meant to be 36% of OpenAI's 2030 revenue.
Moonshot paused new Kimi K3 subscriptions two days after launch when demand swamped its clusters. The 2.8-trillion-parameter open-weight model rivals US systems on coding at a fraction of the price.
+2Anthropic's new Record a Skill feature lets paid users demonstrate a workflow on screen, clicks, typing, and voiceover included, which Claude turns into a repeatable automation it can run on demand.
AI advice makes people three times less accurate but twice as confident, new study reveals
Tl;dr: Accuracy fell by two thirds. Confidence doubled.
AI in Education Exposes Deep Flaws as Homework Scores Rise but Exam Results Crash 20%
Tl;dr: Good grades hid the fact nobody was learning.
AI at work creates trust crisis as workers risk cognitive surrender and skill stagnation
Tl;dr: Lean on AI enough and you train yourself average.
