artificial-intelligence

Mid-2026 headlines flipped from AI job apocalypse to “overblown.” The fight underneath is not whether anyone loses a job — it is who keeps the money if the surplus shows up, and whether “AI” is the cause or the cover story.

Seventy years of US national-lab data have never been uploaded to a model. The White House bet is that unlocking the archive doubles American scientific productivity — not building a better chatbot.

Keep a bot working all day by waking on a clock, reading durable state, doing one atomic unit in a fresh session, writing to git, then exiting — not by keeping one session alive.

Grok App Builder auto-lists .grok/skills. Cursor, Cloud Agents, and Grok Bot do not. A committed symlink is what follows the clone — and Paper still has to sit in the same session as the repo.

Models spike on benchmarks and crater on the next problem. Math is where verification loops work — and where the gap between MCTS credit assignment and LLM RL shows up sharpest.

Software agents can loop all day on a laptop. The next decade-long frontier, in the panel-class read, is bodies — humanoids in factories and streets, and a US–China race over who owns the stack.

Developers feel faster with AI. Controlled trials often measure the opposite. Then December 2025 arrived and four independent practitioners named the same inflection — from Karpathy to Benioff.

Anthropic named four partner-eval incidents and asked METR to look. OpenAI designated Astra Critical on cyber. The Hugging Face report is still a pointer, not a fetched document.

In June the executive branch sat in the commercial release loop. In August a lab paused its own training. In September the labs gated the next SKUs themselves.