Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
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…
Uber engineers burned through $500 to $2,000 per month each on AI tokens, with some teams exhausting their entire quarterly AI budget within four months. The cost explosion is driven by agentic workflows that make multip…
Business owners using AI tools must decide whether to build custom solutions or buy existing SaaS. A two-criteria framework—whether a good-enough solution exists and whether the capability differentiates the business—hel…
OpenAI has launched ChatGPT Scheduled Tasks, a feature for Plus, Pro, and Team subscribers that automates recurring prompts such as daily news briefings and email summaries. Users can set tasks to run at specific times w…
Google DeepMind published a paper in 2025 outlining four pathways from artificial general intelligence (AGI) to artificial superintelligence (ASI): scaling, algorithmic innovation, recursive self-improvement, and group a…
Zhipu AI's GLM 5.2 coding model can be used as a cost-effective backend for AI agents via OpenRouter, offering strong code generation and reasoning at 10-20% of the cost of Claude Sonnet or GPT-4o. This guide provides st…
A new automated workflow enables developers to run an overnight documentation sweep loop that reviews codebases for documentation gaps, updates outdated docs, and opens a pull request for human review. The system uses th…
Claude Code, an AI coding agent, lacks persistent memory across sessions, forcing developers to re-explain context repeatedly. A new guide outlines building a memory system with short-term session memory, long-term seman…
Half of US adults now use AI chatbots, yet more Americans expect negative societal impact than positive, according to Pew Research. The paradox reflects a split between personal utility and systemic risk, with usage doub…
A new automated workflow enables teams to build a nightly production error sweep loop that scans logs, traces errors to root causes, generates fixes, and opens pull requests without human intervention, compressing hours …
Zhipu AI released GLM 5.2, a 744-billion-parameter open-weight language model with an MIT license, achieving coding benchmark scores that rival Claude Fable 5 at a fraction of the API cost. The model's permissive license…
Researchers have developed adaptive speculative decoding, a method that dynamically selects draft lengths at runtime to optimize token generation efficiency in large language models. The approach addresses the changing e…
Open-weight MoE models Kimi K2.7 Code and GLM 5.2, released in June 2026, target developers and agentic coding workflows but differ in design. GLM 5.2 leads with a 1M-token context window and lower hallucination rates, w…
Perplexity launched Max, a $200-per-month AI search tier with advanced tools like Model Council, Deep Research, and a Computer agent, but everyday search still relies on content mills and most users do not need the premi…
Google promoted its Interactions API to general availability on June 22, 2026, making it the default interface for Gemini models and agents. The new API introduces typed steps, managed agents with Linux sandboxes, backgr…