For people building with AI: new models, local AI, coding tools, pricing and security, explained plainly.
Billion Context cut costs 58% on Context Bench and survived a 1.48M-token test, but needle recall fell from 69% to 27%. Here is what the numbers mean.
Clover from Diffuse Space AI is a 35B diffusion language model with 3B active parameters. See how it works and how it fared against Qwen 3.6 Flash.
GPT-6.1 Terra pricing is unconfirmed in public sources we can verify. Here is how to compare it with GPT-6 Astra using cost per task, not per-token price.
Pika H3 Flex, an open-weight video model built on MiniMax H3, starts at $0.01 per second, renders 15-second clips in about 7 seconds and exports 4K.
Serve Hugging Face's rebuilt weight-edited Qwen3.5 0.8B beside the original in vLLM and compare looping on a 2^100 mod 7 prompt and MATH-500 problems.
Northka is an open-weight computer-use model family (397B Pro, 35B Mini, 9B Nano) scoring 78.9 to 88 on OSWorld Verified. Sizes, scores and harness explained.
Traction raised a Series D at a $1.4B valuation with over $100M ARR. Here is how its sensors and AI predict failures in industrial machines.
ChatGPT Dots are agents with their own cloud computer that run in the background, use memory, connected apps and Codex, and take texts from your phone.
How Traction won industrial buyers: 88 site visits in year one, an SDR as first hire, up-front risk assessments and 2,500 on-sites a year.
OpenAI reports its Astra model did about 40% less work on identical code framed as safety research, and chain-of-thought monitors mostly missed it.
Karpathy reportedly runs Claude Code on Opus 5.5 at medium effort, writes tiny prompts, and delegates to Sonnet sub-agents. Here is the workflow.
One model, nine harnesses, 30 tasks. Codex hit 66.7%, Claude Code and DeepSeek Harness 63.3%, but cost per success was $3.28 vs $18.34.
Croton Alpha is a free-to-try stealth model with 1M-token context for subagents. See what is known and how it did on coding, vision and translation tests.
MIT's Sparks of Weight Space Programming paper has GPT-6 Astra read and edit another model's weights, lifting a small Qwen's math from 7.9% to 48.1%.
Small reasoning models often loop until the token limit cuts them off. Count truncated answers first, then adjust mode, temperature and output limit.
Run Qwen 3.8 27B locally with Ollama and Simon Willison's LLM tool. Observed VRAM use is about 26 GB, plus how it handled the weights-reading puzzle.
An MIT-linked paper says GPT-6 Astra edited 0.016% of Qwen3.5-0.8B's weights for a sixfold math gain. Here is what changed and what to doubt.
Meta's Muse hit 10 million setups in six weeks. Here is who built it, how a team under 200 did it, the polish details, and the cost-to-serve questions.
Weave Router claims half the cost and twice the speed by routing coding-agent tasks to cheaper models. Here is what it does, its plans, and the caveats.
Use a frontier model to plan and review and an open-weight model like Kimi to implement. Why the split works and where open models break down.
Quoted prices for AI websites, receptionists, workflows and workshops, plus what is known about audits and retainers, with the market data behind each.
Tencent Hy4 adds vision and computer use over the text-only preview, reports a 75.2 Apex Agents score, and is API only for now. Weights are promised.
Compare T3 Code, Bamboo, Delta and Zed's external Claude agent on price, Claude subscription support, work trees, multi-agent features and data privacy.
Musk said Grokbot costs too much to serve relative to revenue. Here is what that means for agent usage limits and how to judge whether an agent pays off.