Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A developer released TIS (Token Importance Scoring), a learned method for compressing the KV cache in large language models, achieving 100% accuracy on synthetic retrieval at 50% cache budget. The approach uses constrain…
Course creator Josh W. Comeau reports that his third course, Whimsical Animations, is selling roughly one-third as many copies as typical launches, with overall course revenue down 50% or more. He attributes the decline …
A developer who runs AGenO evaluated nine free AI generators in 2026, including ChatGPT, Claude, Gemini, and Perplexity, detailing their actual free tiers and hidden limits. The analysis reveals that most 'free' tools im…
TechCrunch published a comprehensive AI glossary defining key terms like AGI, AI agents, API endpoints, and chain-of-thought reasoning to help readers navigate the rapidly evolving field of artificial intelligence. The g…
A 62-year-old man in India suffered from severe headaches for 25 years, with no doctor able to diagnose the cause. His nephew uploaded his medical records into Claude, an AI assistant, which asked if the man snored. The …
Claude Fable 5's post-reinstatement performance drop on BridgeBench's debugging score from 86.2 to 25.9 was caused by Anthropic's safety classifier routing most tasks to Opus 4.8, not by a degradation of the model itself…
Microsoft announced Copilot Cowork, a new AI system that shifts from chat-based assistance to autonomous task execution across Microsoft 365. The system allows users to delegate multi-step workflows, with pricing based o…
Google released ADK 2.0, a framework that blends deterministic workflow execution with AI agent capabilities to address production reliability issues such as infinite loops and hallucinations. The update introduces struc…
A developer released 'how clanker are you?', a reverse Turing test that measures how surprising a user's writing is to language models. Users finish eight sentences and receive a score based on token surprisal, with low …
Microsoft announced plans to merge its consumer and enterprise Copilot AI chatbots into a single unified application by August 2026, aiming to streamline its AI assistant strategy and better compete with ChatGPT and Clau…
Dropway launches a platform that lets users share HTML and Markdown files generated by LLMs via drag-and-drop, CLI, or MCP integration, with per-site access controls and automatic versioning. The tool addresses the diffi…
A developer's experiments reveal that including a URL in a prompt can steer an LLM's output toward the content at that URL, but only if the URL and its content were part of the model's training data. The findings highlig…
A developer built USAP, an open-source system that enforces a strict evidence gate for LLM security verdicts, requiring every verdict to cite a resolvable source. The system rejects hand-wavy output by checking evidence …
Anthropic released its most powerful AI model, Claude Fable 5, in two versions on June 9, 2026, but the US government imposed export controls three days later, causing it to vanish for everyone until July 1. The same und…
Google's MediaPipe Tasks and AICore are revolutionizing on-device machine learning for Android developers by abstracting low-level tensor manipulation into declarative pipeline orchestration. MediaPipe Tasks treat AI mod…
Anthropic's Claude Code now writes 90% of the company's own codebase, boosting engineer output by 200% and shifting bottlenecks to code review. The tool's Code Review product increased substantive pull request comments f…
Alibaba banned employees from using Anthropic's Claude Code after security researchers discovered hidden steganographic code that tracked Chinese users by timezone and proxy settings. The ban follows Anthropic's accusati…
Developers are exploiting a loophole in Anthropic's Claude Fable 5 pricing by using an open-source proxy called pxpipe that converts dense text into PNG images, reducing API costs by up to 70%. The proxy leverages the fa…
A commentator argues that AI models should have built-in controls to refuse harmful requests, comparing the need for friction to restrictions on handguns and fentanyl. The author advocates for open-source AI to prevent a…
Simon Willison reported that instructing Claude Code to delegate coding tasks to lower-power subagents using its own judgment significantly improved efficiency and reduced token usage. The approach, inspired by tips from…