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 argues that the term 'AI agent' is overused and causing engineering mistakes, defining a true agent as a system with an objective that decides next steps, handles failure, and knows when it's done. The develo…
An open-source MCP server called codebase-memory-mcp, built by DeusData, cuts the token cost of structural code queries by up to 121x by using a persistent knowledge graph. The tool, trending #1 on GitHub, indexes codeba…
OpenAI restricted its new GPT-5.6 Sol AI model to customers approved by the Trump administration, citing government requests amid heightened national security scrutiny. The move follows similar actions against Anthropic,…
Llmaker, an open-source platform for self-hosting a complete LLM stack including models, vector databases, embeddings, caching, observability, and an agent layer, launched on Hacker News. The platform provisions and mana…
Tau is an educational Python coding agent designed to teach how agents are built by exposing every component—provider adapters, harness, coding environment, and terminal UI—in readable layers. The project emphasizes a cl…
Context7 released a CLI tool for current documentation lookup of libraries, frameworks, SDKs, CLIs, and cloud services. The tool resolves library names and queries their documentation, with a limit of four commands per r…
A developer team reduced their LLM API costs by 60% through systematic optimization, starting with instrumenting every call to log token counts per request type. The biggest savings came from semantic caching, which cach…
An independent research project analyzed the internal dynamics of small and medium-sized language models, revealing that functional properties become linearly decodable in hidden representations and that models cluster i…
A developer outlines practical patterns for building multi-agent systems with Python, including the ReAct architecture, tool selection, memory hierarchy, and error recovery. The guide emphasizes starting with narrow task…
Smolagents, a Python library from Hugging Face for building AI agents, prioritizes code-based actions over JSON tool calls, enabling loops, conditionals, and tool composition. However, this design shifts the safety focus…
A developer inquired about DeepSeek, asking if anyone has used it and whether it is really good. The post received no comments.
A Seoul-based backend developer built k-cosmos, an interactive 3D music space that maps K-pop tracks using 768-dimensional vector embeddings, running entirely on the OCI Free Tier. The system uses a self-reinforcing pipe…
Anthropic's Claude Fable 5 is on track to return soon, possibly this week, according to a report from Axios. The AI model's return follows a period of unavailability, and its relaunch is anticipated by the AI community.
Researchers at Anthropic discovered that language models can transfer hidden behaviors through fine-tuning on seemingly meaningless data, a phenomenon called subliminal learning. They identified 'entangled tokens'—such a…
A developer discovered a workaround for a bug in Codex Desktop where requests fail with the error 'This model is not supported when using X-OpenAI-Internal-Codex-Responses-Lite.' The issue occurs after updating Codex Des…
HackerRank open-sourced its applicant tracking system (ATS) scoring tool, but testing reveals extreme score variability—ranging from 66 to 99 on the same resume—due to LLM non-determinism. The tool's scoring is inconsist…
A developer built an AI-powered smart contract auditor designed to help developers catch vulnerabilities early in the development process. The tool analyzes Solidity contracts for issues like reentrancy and access contro…
Developers integrating Qwen3-VL vision-language models into production pipelines face three silent failures: the thinking budget trap where the model consumes output tokens on internal reasoning, leaving empty content; a…
An independent researcher analyzed hidden-state dynamics across seven open-weight LLMs and found recurring functional patterns encoded in representation geometry rather than individual neurons, with functional proxy stat…
A developer demonstrates running a complete RAG agent offline using LangGraph, Ollama, and an embedded Qdrant instance, requiring zero API keys or Docker. The system uses a provider-swap design that allows switching betw…