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 built repomix-rs, an open-source Rust-based infrastructure for transmitting codebase structure and content in an AI-readable format. The tool addresses token budget constraints by using tiktoken-rs for tokeni…
On May 31, 2026, DeepSeek made its V4-Pro API pricing permanent at a 75% discount, with output costs of $0.87 per million tokens—10x cheaper than GPT-4o and 5x cheaper than Claude Haiku 4.5. Meanwhile, Google launched Ge…
A developer implemented webmentions and a comments section for a static website using Claude Code, drawing on resources from the IndieWeb and Fediverse communities. The project, hosted on GitHub, successfully federated a…
An engineer discovered that 60% of a $312 monthly Anthropic bill came from a single retry loop in a Claude Code agent. The culprit was found by shipping Workers logs to R2 via Logpush and querying with DuckDB, revealing …
Shopify's Spring '26 release introduced Agentic Storefronts and the Universal Commerce Protocol (UCP), an open standard backed by major companies that allows AI agents to search catalogs, build carts, and complete checko…
A new study on Hacker News claims self-generated agent skills are useless, but the author argues the research is flawed because it merely asks agents to write about tasks before solving them, which is identical to thinki…
Researchers at Drexel University found that a majority of American teenagers regularly use AI companion chatbots, with a quarter relying on them as a primary source of emotional support. The study analyzed Reddit posts w…
Researchers trained Qwen3-8B natural language autoencoders with varying length penalties and found that explanation length can be significantly reduced without harming reconstruction fidelity, suggesting many tokens in s…
A developer running multiple Bedrock-backed agent workflows in production details defenses against the OWASP LLM Top 10. The agents are designed as mostly-read-only pipelines with no tool calls, limiting impact. Controls…
A developer identifies the 'index distribution problem' as a fundamental flaw in agent-based search systems, where retrieval indexes encode past relevance judgments rather than semantic truth, causing agents to produce c…
A developer discovered that adding excessive rules to an AI agent's context window degraded performance, as the agent cannot remember past corrections and treats rules as soft suggestions. After bloating a rules file to …
A developer built an MCP (Model Context Protocol) server in 200 lines of Go, which they claim made Claude 10x more useful. The server enables Claude to interact with external tools and data sources, significantly enhanci…
Harness engineering is not a fixed location in the stack but a property of code that involves specifying and verifying agent behavior, according to a developer. The durable core of harness engineering is specification an…
Phileas, a local-first long-term memory layer for AI chatbots, launched on Hacker News. The open-source tool uses MCP to let any LLM read and write persistent context across sessions, running two small models locally on …
Loop Engineering introduces a governance architecture for AI agents that prevents costly errors by adding structured control layers, including task definitions, state separation, tool guards, observation collectors, eval…
Twio, an AI SaaS for loan brokers, migrated its production RAG system from pgvector to Vertex AI Search after scaling challenges revealed that vector storage alone was insufficient for handling messy broker documents, em…
Developer Torgeir Helgevold fine-tuned a 600-million-parameter local LLM (Qwen 3:0.6B) to classify household questions into metadata categories, achieving 92% accuracy on a test set—up from 10% with prompting alone. The …
Hugging Face engineer Onur developed a real-time notification system for the OpenClaw repository using local open-weight models like Gemma and Qwen, running on an NVIDIA GB10 with 128 GB of unified memory. The system cla…
PostHog is launching a new AI-powered pipeline that automatically detects product issues, researches them, and opens pull requests with fixes, aiming to replace manual dashboard monitoring with autonomous code generation…
The US government forced Anthropic to pull Claude Fable 5 globally—the first forced shutdown of a commercial AI model. The hypothetical scenario highlights how existing regulations like the EU AI Act and US executive ord…