Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Z.ai released GLM-5.2, an open-weight AI model that performs within one percentage point of Anthropic's Opus 4.8 on agentic benchmarks while costing roughly one-fifth as much, according to CNBC. The company positions the…
On June 12, the US government ordered Anthropic to disable Fable 5 globally within six hours, leaving developers without access and highlighting the risk of regulatory takedowns. The following day, GPT-5.6 Sol launched e…
A developer built Sieve, an open-source proxy that adds persistent, encrypted memory to Ollama's stateless local LLM endpoint. Sieve sits between the client and Ollama, stripping redundant history and injecting relevant …
A developer discovered that high-quality websites can be invisible to ChatGPT Search due to an eligibility problem, not a content issue. ChatGPT Search uses Microsoft Bing as its primary index, so sites must be submitted…
GitHub Copilot's Terraform code suggestions often introduce severe security vulnerabilities, such as opening security groups to the world or enabling public access on RDS instances, because the suggestions are syntactica…
A developer tracked API costs across 184 models over 18 months, spending $340,000 in credits. The data reveals that direct provider pricing can be 40x cheaper than GPT-4o, but operational friction and compliance costs of…
Local LLMs are necessary for privacy and freedom, as data center subscriptions risk data misuse and account lockouts. Smaller models are becoming more capable, making local AI viable for software development despite upfr…
A developer built a Flutter AI lecture video creator using local Ollama and FFmpeg, cutting cloud API costs by 80%. The system tackles three major synchronization challenges by generating segmented audio with Edge-TTS an…
Anthropic received U.S. government authorization to allow a small group of American cybersecurity firms to access its powerful AI model Mythos 5, which had been blocked due to national security concerns. The company will…
NVIDIA tested its open-weight Nemotron family of models on real-world coding agent tasks and found a capability floor below which models cannot drive an agent loop at all. The smallest variant, Nano 12B, produced no resu…
A developer spent two weeks comparing four Chinese AI model families—DeepSeek, Qwen, Kimi, and GLM—via Global API's unified endpoint, finding DeepSeek V4 Flash at $0.25/M output offers a 97.5% discount over GPT-4o while …
A developer proposes a release gate pattern for AI browser automation runs to validate environment consistency before execution. The pattern checks profile identity, proxy region, session validity, evidence collection, a…
A new interview guide for senior Go engineers focusing on AI platform engineering has been published on Leanpub. The book covers LLM gateways, RAG, vector search, Kubernetes, and production failure modes, targeting mid-l…
A new WebGPU-powered demo of Qwen3.5 has been released on Hugging Face Spaces, enabling large language model inference directly in the browser. The project showcases the potential of running advanced AI models on consume…
Dhravya Shah raised approximately $3 million in seed funding for Supermemory, a startup building a universal memory API for AI applications. The round was led by Susa Ventures, Browder Capital, and SF1.vc, with backing f…
OpenAI launched GPT-5.6 Sol, Terra, and Luna as a restricted preview limited to trusted partners at the request of the U.S. government, marking a shift toward government-mediated frontier AI releases. The flagship Sol mo…
A Hacker News user proposes a blueprint for achieving human-like thinking in machines using current LLM technology, suggesting a system of random multi-domain image/video generation combined with specialized LLMs that an…
A developer building an AI memory system called MemStrata found that it comprehensively beats RAG on mutating code content, as detailed in a paper on arXiv. The developer emphasizes the importance of trusting data over h…
Piyush Gupta proposes Certainty Volatility Theory, which suggests that psychological discomfort stems not from uncertainty itself but from volatility in one's level of certainty. The theory argues that stable, even low, …
A developer built a Text-to-SQL assistant using Python, Hugging Face's Inference API, and Streamlit, enabling users to query databases in plain English. The project leverages a fine-tuned T5 model to translate natural la…