…Related I stopped treating Claude like a chatbot, and it became my best co-worker Claude gets smarter when conversations become collaborative. Great models make better guesses, not miracles Initially, I assumed…
…Any signs of slacking or lagging behind may permanently set the company back relative to its peers, so everyone is doing their best to release new models regularly. OpenAI is no exception…
…diagram where students can explore each component interactively D A physical model mailed to the student's address Correct! Claude can generate structured, interactive diagrams with labeled components students can explore in…
…But the recent release of Google DeepMind’s Gemma 4 12B challenges that status quo. This 12-billion-parameter mid-sized model handles multi-step reasoning and a massive (and efficient) 256K…
…Mistral 7B v0.3 was one of the first open-weight models with native function-calling tokens, and it still works. It's old, mind you, releasing in May 2024, but it…
…When you call a skill directly with $, the full content gets injected, just like it would on a cloud model. When a skill is attached to a model, only the name and…
…explicit AI features," with explicit tools being more akin to the agentic models like Claude Code, and the implicit ones improving what Ubuntu already does: One exciting example of this is bringing…
…likes of Claude Code. Each skill is represented by metadata in /fatfs/skills/skills_list.json and one or more Markdown documents/scripts on FATFS. The Markdown tells the model when and…
…The main model behind it, Gemini 3.1 Pro, is much faster now and handles huge amounts of information well, but it still feels a step behind Anthropic’s Claude when it…
…releases, with GPT‑5.4‑Cyber as the model and an expanded Trusted Access for Cyber (TAC) as the framework. Related I cancelled my ChatGPT, Perplexity, and Gemini subscriptions for Claude — and…