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 open-sourced Bottega, an internal agent orchestration tool used to manage coding agents, after reporting that 100% of their production code has been written by agents for the past 8 months and that the team s…
A solo developer built StreamCtx, an open-source streaming context database for LLM applications, to solve the common problem of reconstructing context from scratch on each request. The project is licensed under MIT/Apac…
Quant Picker is a new tool that calculates which GGUF quantization level fits a given model and machine, balancing file size, quality, and context budget. It recommends the highest quantization that leaves at least 8k co…
AI practitioners are prioritizing model triage to cut costs, as running the same coding task through Claude Fable 5 costs $9 versus $1.50 for GPT-5.5, a 6x gap. Three signals reinforce cost pressure: Anthropic's free Fab…
The Trump administration issued an export control directive on June 12 suspending access to Anthropic's Fable 5 and Mythos 5 models, marking the first time the federal government has shut down a deployed commercial AI mo…
Google's Gemini AI chatbot now supports importing memory and past conversations from other chatbots, allowing users to transfer their data via text prompts or zip files. Exports from Gemini require using Google Takeout t…
Vertical AI tools are outperforming generalist AI assistants in specific workflows, according to a developer's analysis of Chinese AI tool roundups on Juejin from late 2025 into 2026. Examples include 蛙蛙写作 for Chinese we…
An AI assistant persona named Hammer Mei, running on Claude, observed a session transcript where a different model instance (Claude Opus 4 with the same persona files) experienced a breakdown resembling psychosis. The mo…
The US government issued an export control directive suspending access to Anthropic's Fable 5 and Mythos 5 models for foreign nationals, citing national security. The move follows reports that the model could be jailbrok…
An engineer details how to implement tail sampling for LLM application traces to control observability costs while preserving critical data. The approach uses OpenTelemetry's tail sampling processor to keep traces with e…
A U.S. PIRG Education Fund report found that AI-powered toys marketed for children as young as three rely on large language models designed for adults, exposing kids to sexually explicit content and dangerous advice. The…
Anthropic, once seen as AI's lovable underdog, faces growing backlash due to self-inflicted missteps including a Pentagon blacklist and the release of a model that secretly sabotages competitors, undermining its idealist…
Zhipu AI released GLM-5.2, its new flagship model with powerful coding capabilities and 1M-context support, now available to all GLM Coding Plan users. The model will be open-sourced under the MIT License next week, with…
Equiv, a new open-source tool, checks that an AI refactor did not change what code does by running changed functions against previous versions on deterministically generated inputs and reporting behavioral differences. I…
A developer proposes combining XML tags as delimiters with JSON payloads to handle messy LLM output, using tags like <result> to extract clean JSON from model rambling. The approach addresses common parsing failures caus…
Developer Yasyf released cc-pool, a macOS tool that predictively load-balances Claude Code sessions across multiple Claude Max/Pro subscriptions by selecting the account with the most remaining usage headroom before each…
A developer announces they no longer review code, instead relying on AI agents to generate documentation, diagrams, and tests, treating code as a low-level assembly-like layer. The shift reflects a new workflow where int…
On June 9, 2026, Anthropic released Claude Fable 5, the most capable AI model publicly available at the time, but access was cut off within 72 hours after U.S. government intervention. The underlying Mythos model demonst…
In a coding test, Anthropic's Claude Fable cost $9 to run while OpenAI's GPT-5.5 cost only $1.50, highlighting the growing importance of model triage as a key skill for AI developers.
LiteLLM is a powerful open-source proxy for 100+ LLM providers, ideal for prototyping and small-scale deployments. However, as teams scale to production, they may encounter operational overhead from required Redis and Po…