Analysis of the AI industry: companies, markets, buyers and the economics behind them. Recent posts look at Character.AI's bankruptcy, 47 stalled US AI data center projects, how Big Law buys AI, and Microsoft 365 Copilot vs Gemini in Workspace.
Siemens showed two AI tools built with NVIDIA at its Erlangen plant: inspection trained on digital twins before prototypes exist, and an AI planning agent. Siemens claims up to 85% less engineering effort and up to 20% cost savings. Here is how they work and what is missing.
World launched World ID for banking and finance at Money20/20: a Zoom Human Badge that matches an Orb image, a selfie and the live video, three levels of proof, and beta tools that let a verified person delegate proofs to an AI agent. Bain Capital and VanEck Funds are design partners.
Huntington Bank launched The 1866 Studio, an AI system built on Google Cloud's Gemini platform that helps small business owners launch, market and get loan ready. Here is what it does, how it is built, and how it compares with banks blocking outside agents.
Anthropic says junior roles fell from 27% to 9% of its technical hires as agent hours grew. What the report found, what other labor data shows for young workers, what hiring managers now screen for, and practical steps for anyone trying to land a first job in AI.
Santander's AI Lab says smaller, specialized models cut serving costs from around €5 to about €0.03 per interaction in selected experiments. Here is what it built with NVIDIA, what the three-year deal covers, and what the numbers do not show.
MiniMax says the anonymous Space Bunny model was an early version of M3.1 Flash. It has 1M context, always-on thinking, and a flat $0.15 input and $0.60 output price per million tokens. Here are MiniMax's benchmarks, how the price compares with DeepSeek, and what is still missing.
NOAA expects a strong-to-very strong El Niño through early 2027. Here is how AI weather models like GraphCast, AIFS and WeatherNext 3 work, who runs them, and why record-breaking conditions outside their training data are the real test.
Suno and Udio cost about the same, but they now differ on what you can do with a song. Suno v6 keeps metered downloads with commercial rights on paid plans. Udio has disabled downloads while it builds a licensed fan app. Here are models, prices, rights and legal status as of October 2026.
CyberSN counted 235,786 US cybersecurity job postings from January to September 2026, a four-year low, while AI Security Engineer postings rose 268.3% to 2,409. What the role involves, the skills it asks for, and verified resources to get there.
In a Financial Times poll of 94 economists, 54% expect AI to bring significant gains to many or nearly all G20 countries over the next decade, while 32 warned that weak infrastructure puts African economies at risk of falling behind.
Google paid Character.AI $2.7 billion in 2024. Disney is paying the bankrupt company $20 million while budgeting about $180 million for its staff. The reverse acqui-hire works for buyers and hires, and leaves users and creditors with the shell.
GPT Image 2.5, Nano Banana 2.1, Midjourney V8.2, Ideogram 4.5, Recraft V4.1, FLUX 3 Image Klein, Krea 3 and more: what each image generator is built for and what vendors publish about price, as of October 2026.
Apple HomeView is a $299, 7-inch smart display with Siri AI that recognizes up to six people by face. Here is what it does, what the AI features need, the privacy claims, and what to check before it ships on November 13.
OpenAI's Dean Ball now thinks distillation explains about 50% of how close Chinese models are to US ones, up from his earlier 30% guess. Here is the public evidence, from Anthropic's February report to the NSA, CISA and FBI advisory, and China's response.
Anthropic surveyed 1,334 staff and 10,980 US adults. Run through its economic model, the median staff answers imply 17.1% GDP growth in 2030 versus 5.4% for the public. A companion report shows junior technical hires falling from 27% to 9%.
Hogan Lovells Cadwalader signed with both Harvey and Legora, kept its 100-person in-house tech unit, and created an AI Board to decide what to build or buy. Here is what the firm announced, how its lawyers use Harvey, and what it says about legal AI.
Vercel's switching data, Replit's new default, DeepSeek's silent upgrade, and OpenAI's shutdown calendar point the same way: the model is now a replaceable part, and the power sits with whoever controls which part gets used.
Mistral's Timestral is a family of four open-weight time-series forecasting models, 5M to 209M parameters, that run on a CPU. Mistral says the 209M model ranks first on GIFT-Eval, TIME and fev-bench, by narrow margins. The weights license excludes companies with monthly revenue above $20 million.
Nvidia asked Samsung, SK hynix and Micron to build NVHBM base dies on TSMC 3nm and standardize on 1c DRAM, the Seoul Economic Daily reports. Samsung agreed, its first HBM manufacturing tie-up with TSMC, but keeps its in-house model for standard HBM and AMD.
Vercel's October Production Index shows open-weight models handled 71.5% of gateway tokens in September but about 23% of spend. Total spend rose 13x since December as token prices fell 47%, a third of teams switched labs, and Anthropic's spend share fell below half.
