In an unforgiving global environment that leaves no room for self-inflicted injuries, I urge authorities to slash red tape, get rid of obsolete rules, and free up the power of entrepreneurship. Full speech:

"The artificial intelligence boom may be entering a more unforgiving phase, as investors increasingly question whether the billions poured into data centers will translate into sustainable profits," per CNBC

The artificial intelligence boom may be entering a more unforgiving phase, as investors increasingly question whether the billions poured into data centers will translate into sustainable profits. A bruising stock market debut by South Korean AI infrastructure provider Elice has become the latest warning sign that investors may be more closely scrutinizing ambitious expansion plans and paths to profitability. Shares of the Seoul-based neocloud company plunged 32% in their trading debut Tuesday, even after an initial public offering priced at the top of its marketed range. The stock then rebounded by about 30% on Wednesday:

Economic models of sovereign debt typically treat debt like it is from private sources: Impatient, unforgiving. Our @cristinarellano & @Leo_BarretoV model the reality that "official" lenders step in during a debt crisis leading to very different outcomes.

NEWS: Grok 4.7 has claimed the #1# spot on Harvey LAB-AA v1.1. It now sits ahead of Claude Opus 5.5, GPT-6 Astra, Muse Spark 1.3 and other leading models. The challenge is unforgiving: one material hallucination can mean zero points for the entire task. It’s a test of end-to-end accuracy, not just convincing answers. Grok 4.7 is pushing deeper into high-stakes reasoning.

ICYMI: @Cognition's engineers ran the first customer-executed benchmark on NVIDIA Vera Rubin NVL72. They measured it against their own NVIDIA GB200 NVL72 baseline. Up to 4.8x the total token throughput for SWE-2 inference. 3.8x the output token throughput for reinforcement learning. Agentic coding is an unforgiving workload. Long contexts. High concurrency. Token volumes where cost per token decides what you can ship. For Cognition, those numbers mean more concurrent Devin sessions per GPU, drastically accelerated research loops, and lower cost per session, with no loss in generation speed. The whole stack runs on CoreWeave: training, reinforcement learning, and the inference that serves users. All of it on one platform. The benchmark is theirs. So is the workload.