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 on the Mattrx platform replaced blind document chunking with the Open Knowledge Format (OKF) and a Context Engine, reducing hallucination rates from 18% to 3% and stale-answer rates from 11% to 1.5%. OKF stor…
A developer argues that manual test cases should be stored as YAML files in version control rather than in SaaS databases. The approach leverages AI for case generation, enables PR-based review via git diff, and provides…
Anthropic released Claude Sonnet 5, a new AI model with advanced autonomous capabilities, available immediately across all plans and APIs. The model offers improved reasoning, tool use, and coding at a lower cost than th…
SemiAnalysis surveyed over 50 enterprises and found that widespread token overspending, as reported at Meta and Uber, is overstated and not representative of most organizations. Companies are now adopting token budgets r…
Meta introduced Brain2Qwerty v2, an AI system that decodes brain activity into text using non-invasive MEG scans, achieving 78% word accuracy in experiments with healthy volunteers. The research aims to help patients wit…
Transformer models develop outlier channels—feature dimensions with unusually large values in weights and activations—due to the softmax normalization in attention layers, which forces tokens to assign importance even wh…
Agentic AI is transforming information from static documents into responsive entities that adapt to the reader's understanding process. This shift moves beyond search and generation to create a feedback loop where the re…
Anthropic launched Claude Science, an AI research workbench integrating databases, tooling, and compute into a single environment. The platform includes over 60 scientific databases, renders 3D protein structures and gen…
Anthropic's Claude Sonnet 5 is now available on Snowflake Cortex AI in private preview, offering enhanced reasoning, coding, and agentic capabilities within Snowflake's secure perimeter. The model approaches Opus-level p…
Autotune, a new open-source tool, optimizes local large language models by automatically right-sizing KV cache buffers, tuning precision, caching system prompts, and managing model keep-alive, freeing over 300 MB of RAM …
A developer known as rynaqrtz created an open-source Node.js library called chatday that provides programmatic access to the Chatday.ai API. The library handles anonymous authentication, model listing, and streaming chat…
A developer created mcpgen, a tool that converts any OpenAPI specification into a ready-to-run MCP server in seconds. The generated server runs over stdio and can be plugged into Claude Desktop or any MCP client. The too…
A developer identified a structural cause for coding agents ignoring system-prompt rules mid-session: attention dilution as context grows. The rule becomes an old, low-weight token among thousands of newer tokens, leadin…
Trix Cyrus introduces context engineering as a discipline that goes beyond prompt engineering to curate everything a language model sees during inference, including system prompts, retrieved documents, conversation histo…
Anthropic released Claude Sonnet 5, a new AI model offering top-tier coding and tool use performance at Sonnet pricing with a 1M context window. It is now the default in Claude Code for Pro users and available across the…
A developer argues that feature engineering remains critical for machine learning performance, even in the era of large language models. Raw data, such as dates and order counts, often fails to capture meaningful pattern…
Anthropic is rolling out new real-time cyber safeguards on Claude Opus and Sonnet models to automatically detect and block prohibited or high-risk cybersecurity usage, while launching a Cyber Verification Program for leg…
Giles Thomas built a JAX training loop for an LLM from scratch, using NNX and Optax, as part of a project to replicate GPT-2 small quality without referencing Sebastian Raschka's book. He trained a base model on 3.2B tok…
Answer Engine Optimization (AEO) is reshaping digital marketing as zero-click searches approach 65% and Google AI Overviews trigger an 83% zero-click rate. ChatGPT processes 2.5 billion prompts daily, with LLM-referred t…
A developer shares three MCP servers they use daily with Claude Desktop: server-filesystem for reading project files, server-github for code reviews, and server-postgres for database queries. Each server takes two minute…