Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Intel's AutoRound quantization algorithm for large language models claims state-of-the-art performance, enabling LLMs like Qwen 3.6 to run on consumer GPUs. In testing, AutoRound produced consistent output, slightly smal…
A developer identified the root cause of Claude Code API ECONNRESET errors on a specific home network: the macOS auto-tuned TCP send buffer causes large uploads to burst into the DSL modem's upstream buffer, which tail-d…
MarkTechPost published a hands-on guide to GLM-5.2's OpenAI-compatible API, demonstrating reasoning effort control, function calling, tool-using agents, structured JSON output, and long-context retrieval. The tutorial in…
Weibo's AI group released VibeThinker-3B, a 3-billion-parameter model that achieves 94.3 on AIME and outperforms Claude Opus 4.5 on competition reasoning, but collapses on general knowledge tasks. The model's performance…
A staff engineer at an unnamed company built an AI code reviewer that mines three years of team PR comments to generate feedback in each engineer's voice, addressing the bottleneck of code review in the age of AI-generat…
Amazon announced Alexa+ Agentic Ads at Cannes Lions, a new ad format that lets users complete purchases within natural-language conversations. The move signals a shift toward AI-native shopping, with brands like Papa Joh…
OpenAI announced at its first Cannes Lions appearance that it is entering the advertising business with a new ad product for ChatGPT, aiming to generate $100 billion in revenue within four years. The company plans to mea…
A developer created a benchmark to test whether AI models besides Anthropic's Mythos can find security bugs as effectively, using bugs Mythos discovered. The benchmark currently includes nine bugs from Mythos's documenta…
Z.ai released GLM-5.2, an open-weight AI model, on June 13, 2026, to GLM Coding Plan members, followed by public release on June 16. The model has been praised by the AI community for its strong performance on agent benc…
A Harvard Business Review blog warns that reliance on generative AI is causing 'knowledge decay' in organizations, as employees produce and consume low-quality AI-generated content, eroding trust and productivity. Expert…
A user reviewing Anthropic's Claude Mythos Preview System Card discovered two plots with likely mislabeled captions: one showing an 80% dishonesty rate that appears to actually represent honesty rate, and another on hall…
Researchers developed VibeThinker-3B, a 3-billion-parameter language model that achieves reasoning performance matching or exceeding models orders of magnitude larger, scoring 94.3 on AIME26 and 80.2 on LiveCodeBench v6.…
Eliezer Yudkowsky's 25-year-old document 'Creating Friendly AI' criticizes the 'adversarial attitude' in AI development, arguing that superintelligent AI should genuinely want to interpret human wishes accurately rather …
Anthropic is preparing to extend its Cowork agentic system to mobile devices, with an iOS app build showing a Cowork entry that would allow users to schedule tasks from a phone and access results across platforms. The up…
Hugging Face automated its weekly huggingface_hub releases using a GitHub Actions workflow with open-source tools and open-weights AI models, keeping a human in the loop for release notes and announcements. The new proce…
Frontier LLMs including GPT-5.5, Gemini 3 Pro, and Opus 4.7 failed to correctly solve over two-thirds of problems in ParallelKernelBench, a new benchmark for multi-GPU kernel generation. The benchmark's 87 problems, draw…
A new study finds that panels of nine large language model judges effectively provide only about two independent votes' worth of information, because the models make correlated errors on the same items. The panel's accur…
PostHog introduces 'scouts' as small autonomous agents that monitor product data and emit signals for action, forming a core part of its self-driving pipeline now in open beta. Scouts use markdown-based skills to detect …
Conductor 0.69.0 introduces OpenCode integration, allowing users to run models through the OpenCode harness. The update includes improvements such as macOS Keychain-backed storage for credentials, a fast mode toggle for …
Anthropic introduced Claude Tag, an AI that can be tagged into shared conversations, while a new paper examines stability in AI agents. The author argues that AI systems should prioritize helping users re-enter tasks gra…