Analysis — page 6
Nvidia's real moat is CUDA, not silicon
A matrix multiply that takes three lines in PyTorch takes 50-plus in CUDA — and that gap is why AMD and Intel keep losing.
Barry Diller backs Sam Altman but warns 'trust is irrelevant' as AGI nears
The IAC chairman says OpenAI's CEO is sincere, but the real risk is that AI's creators don't understand what they're building.
Schmidt office researchers lay out blueprint for AI's role in democracy
Andrew Sorota and Josh Hendler argue personal AI agents are reshaping how citizens form beliefs — and that today's institutions are not ready.
Eric Schmidt's office publishes a blueprint for AI in democracy
Andrew Sorota and Josh Hendler argue personal AI agents are about to mediate citizenship — and US institutions are not ready.
IAC just shut down a $321M business with a 30-year brand. Perplexity is worth $21B.
Ask.com closed on May 1, 2026, instead of being relaunched as an AI search wrapper. By any reasonable read of the comp set, that was the biggest unforced strategic error in consumer AI this year.
Jack Clark: 60% chance AI builds its own successor by end of 2028
Anthropic co-founder lays out the benchmark trail — SWE-Bench from 2% to 93.9%, METR horizons from 30 seconds to 12 hours — and calls the takeoff.
Reuters warns investors are running out of time to brace for an oil shock
A Reuters column argues markets are underpricing the odds of a genuine crude disruption — a risk that lands directly on AI data center economics.
Platformer's Casey Newton: the AI bubble looks more like railroads than crypto
A new framing argues today's AI overbuild will leave behind durable infrastructure, even if the equity returns disappoint a generation of investors.
Reid Hoffman says doctors not using AI for second opinions border on malpractice
The LinkedIn cofounder, speaking at WIRED Health in London, argues frontier models from OpenAI and Anthropic should sit beside every clinician.
Casey Newton overhauls Platformer's schedule as AI eats link aggregation
Platformer drops its thrice-weekly cadence for original reporting, citing AI-generated briefings encroaching on newsletter work since January 2026.
AI's underpants gnomes problem: Mercor study finds agents fail most of 480 workplace tasks
A February 2026 Mercor study tested top agents from OpenAI, Anthropic, and Google DeepMind on real banker, consultant, and lawyer work. They flunked.
Databricks and Infosys say enterprise AI now needs 92% precision to ship
Bavesh Patel and Rajan Padmanabhan argue fragmented data, not model choice, is what stalls enterprise AI deployments.
Stay ahead of everyone in AI.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at