- Simon Willison: ggml.ai joins Hugging Face to ensure the long-term progress of Local AI (Feb. 20, 2026)
ggml.ai (llama.cpp) is joining Hugging Face to secure long-term progress of local AI, aiming for tighter transformers compatibility, better packaging, and improved user experience. - Simon Willison: Gemini 3.1 Pro (Feb. 19, 2026)
Google released Gemini 3.1 Pro, priced like Gemini 3 Pro, cheaper than competitors, and boasting improved SVG animation. - SemiAnalysis: Claude Code is the Inflection Point (Feb. 5, 2026)
Claude Code writes 4% of GitHub commits, may hit 20% by end-2026. As a CLI agent, it reads, plans, and executes code, reshaping work, and fueling Anthropic. - WSJ: Long-Running AI Agents Are Here (Feb. 5, 2026)
Long-running AI agents, like Anthropic’s Claude and Opus 4.6, sparked a tech selloff. They run multi-step tasks, use plugins, and threaten SaaS, forcing firms to rethink models, roles. - Tyler Cowen: GPT as a Measurement Tool (Feb. 20, 2026)
“We find that GPT as a measurement tool is accurate across domains and generally indistinguishable from human evaluators. ‘ - Michał Podlewski: Cardiologist wins 3rd place at Anthropic’s hackathon (Feb. 20, 2026)
A cardiologist placed third among 13,000 applicants at Anthropic’s hackathon, building postvisit.ai in seven days, coding day and night. The Opus 4.6‑powered platform acts as an AI care companion after visits, consolidating medical history, devices, and evidence‑based resources. - NY Times: Can A.I. Already Do Your Job? (Feb. 18, 2026)
AI tools like Claude Code let non-coders build software by supervising autonomous agents that plan, write, and test code. They’re automating many white‑collar tasks, changing software work, and raising concerns. - Jan Tegze: Your Job Isn’t Disappearing. It’s Shrinking Around You in Real Time (Feb. 2, 2026)
AI agents are eroding knowledge workers’ roles; learning tools, deepening expertise, and leaning on soft skills don’t solve it. Create new roles that orchestrate agents, remove human limits. - NY Times: Can an A.I. Productivity Boom Clear a Path for More Rate Cuts? (Feb. 20, 2026)
Kevin Warsh, President Trump’s Fed pick, says an A.I. productivity boom could raise growth without inflation, opening space for rate cuts. - Margaret-Anne Storey: How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt (Feb. 8, 2026)
Generative and agentic AI shift focus from technical debt to cognitive debt, the loss of shared understanding that blocks changes. - NY Times: These Mathematicians Are Trying to Educate A.I. (Feb. 7, 2026)
First Proof tests LLMs on unpublished research math problems, finding they give hand‑wavy, inconsistent answers and struggle without human oversight. The authors seek objective benchmarks, curb hype, and protect students. - WSJ: OpenAI Employees Raised Alarms About Canada Shooting Suspect Months Ago (Feb. 20, 2026)
OpenAI flagged ChatGPT messages by Jesse Van Rootselaar about gun violence, debated alerting police, and banned her account without notifying authorities. - Transformer: White House war on Utah AI bill could backfire (Feb. 20, 2026)
Voters across parties worry chatbots threaten children’s mental health, safety, and welfare, and favor state AI safeguards. The White House opposed Utah’s modest transparency bill, sparking concerns about federal preemption, industry influence, and a possible political backlash. - NY Times Opinion: The Left Needs a Sharper A.I. Politics (Feb. 10, 2026)
The left needs to craft a clear response to A.I. Weigh job preservation, universal basic income, and human exceptionalism, not dismissal or signaling. - WSJ: Google Is Exploring Ways to Use Its Financial Might to Take On Nvidia (Feb. 20, 2026)
Google is expanding the market for its TPU AI chips by funding data-center and neocloud partners, including talks to invest about $100 million in Fluidstack. - CNBC: Silicon Valley engineers charged with stealing Google trade secrets (Feb. 19, 2026)
Three Silicon Valley engineers were indicted for allegedly stealing Google and other firms’ trade secrets, copying SoC processor files, and sending them to Iran. - WSJ: How to Stay Sane in the AI Skills Race (Feb. 4, 2026)
Don’t panic about AI—relatively few job listings demand it. Choose targeted training, build a portfolio, and explain how AI improves your work, not chase flashy certificates. - NY Times Opinion: In search of grown-up movies for kids (Feb. 24, 2026)
Popular culture needs a middle ground, more adult than Y.A., less explicit than HBO, that lets 10–16-year-olds encounter grown-up themes gradually. Older PG-13 films and 1950s–60s movies offer models that suggest sex, violence, and maturity, rather than showing them graphically.