Insights
Practical analysis and guides to help you buy, build, and work with AI more effectively.
Kimi K3 Weights Go Live Today: The World's Largest Open Model, But It Makes Things Up Half the Time
Today Moonshot AI dropped the full weights of Kimi K3 on HuggingFace. 2.8 trillion parameters, Modified MIT license, free to download, modify, and self-host. It is the largest open-weight model hum
AMD Helios Rack-Scale AI System Debuts, with Microsoft On Board to Challenge Nvidia
AMD just pulled its rack-scale play out of the bag On July 20, AMD showed off Helios for the first time to press — not another graphics card, but a full rack. GPUs, CPUs, networking and a software st
Google Gemini 3.5 Pro Delayed Again: Coding Falls Short, Alphabet Sheds ~6% in Two Days
From "give us one more month" to "coming soon" At Google I/O on May 19, Sundar Pichai stood on stage and said: give us one more month. Everyone bought it. A month later, nothing shipped. The new lin
Mira Murati's First Move: Inkling, 975B Open Weights with Native Audio
Mira Murati's first card Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, pulled the wraps off its first foundation model on July 17 — Inkling. 975 billion parameters, Mixture-of-Exp
Kimi K3 Drops at Midnight: 2.8T Params, the Largest Open Model Yet
2.8 trillion. The open-weights ceiling just broke again. At midnight on July 17, Moonshot AI dropped Kimi K3. 2.8 trillion parameters, Mixture-of-Experts, 16 experts fired per token — the first open
Google's Gemini 3.5 Pro Drops Tomorrow: A Month-Long Retrain to Beat Fable 5 at Frontend
Google Is Making a Big Bet Tomorrow, July 17, Google DeepMind's flagship model, Gemini 3.5 Pro, finally drops. It was supposed to land in June. Instead it got pushed back over a month. And this wasn
Microsoft Quietly Replaced OpenAI with Its Own MAI Models in Excel and Outlook
Microsoft Is Pulling Away from OpenAI Bloomberg dropped a scoop on July 14: Microsoft has been quietly swapping OpenAI and Anthropic models for its own MAI series inside Excel and Outlook. Tens of t
Zhipu Internal Letter Revealed: Betting on AI Operating System and Self-Designed Chips
An Internal Letter July 11. Zhipu founder Tang Jie sent an internal letter named Touch High — the Reach-High Plan. Its weight isn't in slogans but in laying out Zhipu's two-year direction clearly: n
Meta Super Sensing Glasses Exposed: Always Recording, and the Light Stays Off
Meta's Glasses Want to Listen Forever: "Super Sensing" Plan Exposed, and the Light Stays Off On July 8, the Financial Times dropped a bombshell. Meta is internally prototyping AI smart glasses coden
Meituan Open-Sources 1.6T LongCat: Trained on Domestic Chips, Zero NVIDIA
Meituan open-sources 1.6T model: trained entirely on domestic chips, zero NVIDIA July 2, Meituan open-sourced LongCat-2.0. 1.6T parameter MoE, 59.5 on SWE-bench Pro, pricing at $0.038 per millio
Meta's First Paid API: Muse Spark 1.1 Targets GPT-5.5, Open-Weights Path Shifting?
Meta just did something big: its first paid API is live, and the open-weights path is shifting July 9, Meta released Muse Spark 1.1. The specs alone aren't shocking — 1M token context, agentic c
Anthropic Found J-space in Claude: Interpretability Breakthrough That Catches Models Lying
Anthropic found a "workspace" inside Claude's mind On July 6, 2026, Anthropic dropped a striking paper — "A global workspace in language models." Inside Claude's neural network they found a small spe