The file
Sourced AI briefs in publish order — newest first. Citeable originals live on the homepage under From Signal Desk.
Jacob Coxon leaves Anthropic warning labs are gambling with self-improving superintelligence—and alignment lead Evan Hubinger backs him: earnest belief AI could kill everyone, personally above 10% this decade.
Google's AlphaGenome Atlas scores every possible single-base swap in the human genome—precomputing hints for non-coding variants that used to mean months of specialized guesswork.
On Equity, ControlAI's Connor Leahy argues alignment and containment aren't enough—stop building superintelligence—and says ideas that sounded fringe months ago now have legislative wind.
At IBC 2026, NVIDIA is packaging authenticity detectors, frame generation, and sports-specific fine-tuning into one media stack—pushing live production toward software-defined AI without pretending the trust problem is solved.
Google DeepMind's "Love, Rendered" turns generative image and performance models into a memory aid for a 70-year marriage—beautiful when guided by the couple, uneasy when the product pitch arrives.
On TechCrunch's Equity podcast, ControlAI's Connor Leahy argues superintelligence is an adversary to stop building—not a weapon to align—and points to new U.S. and U.K. bills as proof the Overton window moved.
IBM's Granite PatchTST-FM-r2 tops permissive-license zero-shot forecasting on GIFT-Eval—~385M parameters, Apache/OpenMDW dual license, and enough training transparency for enterprises to actually review.
A pre-training researcher walks out of Anthropic and says the quiet part out loud: the labs already know the stakes of self-improving AI — and they’re racing anyway.
Instacart's Clementine is less a chatbot novelty than a bid to own the household meal-planning decision before the cart is filled—and before ChatGPT gets there first.
Prime Video's AI lip-sync for human-dubbed audio starts narrow—English Maxton Hall—but points to a streaming arms race where localization quality becomes a retention feature, not a subtitle footnote.
OpenAI's Chris Lehane says the policy window is open: mandatory federal frontier safety rules, four new California bill endorsements, industry monitoring standards, and global bars on when development should slow.
Suno's licensed-data v6 launch is a legal and commercial reset: retire the contested stack, productize label deals, and sell new revenue toys while Sony and Universal cases still burn.