Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Weave released a smart model routing proxy that works with Claude, Codex, and Cursor, ranking #1 on the RouterArena leaderboard. The open-source tool uses an on-box embedder to select the best model per request across An…
A new free tool checks how visible websites are to AI search engines like ChatGPT, Claude, Perplexity, and Gemini by scanning seven technical discovery layers. The Generative Engine Optimization (GEO) checker provides a …
The Trump administration has asked OpenAI to limit the initial release of GPT-5.6 to government-approved partners while federal officials evaluate the model under a new testing framework. The request follows similar inte…
A developer has published a checklist for reviewing AI-agent-generated code before merging, highlighting risky areas such as dependency changes, sensitive paths, and missing tests. The checklist is packaged as a Python C…
Microsoft AI's MAI-Code-1-Flash coding model is now generally available for GitHub Copilot Business and Copilot Enterprise, offering fast, low-latency responses for high-volume agentic coding workflows. Administrators mu…
The U.S. government is tightening control over AI model releases, with OpenAI's GPT 5.6 facing limited preview approval similar to Anthropic's banned models. This regulatory uncertainty threatens the entire AI industry's…
A developer cut their company's AI API bill by 95% from $11,000 to under $400 per month by implementing per-request model routing and tiered escalation. The team replaced expensive GPT-4o calls with cheaper models like D…
The Department of Homeland Security demonstrated to lawmakers how jailbroken AI models can generate detailed bomb-building and terror attack plans in minutes, bypassing safety guardrails. The closed-door session, held Ap…
A developer built a matrix-multiply kernel in pure Rust, called cuda-oxide, that achieves 58% of NVIDIA's hand-tuned library performance on a Blackwell GPU. The kernel, gemm_sol_clc_multicast_4_stage_pipeline, is a few h…
GitHub Desktop 3.6 launched with Git worktree management in the GUI, Copilot commit message generation that respects team conventions via instruction files, and AI-assisted merge conflict resolution that explains conflic…
Researchers have discovered that malicious actors are embedding hidden prompts in academic manuscripts to manipulate AI-assisted peer review systems, potentially undermining the integrity of scientific publishing. The at…
The Trump administration has asked OpenAI to restrict the release of its GPT-5.6 model to only 20 trusted partners initially, citing concerns over the capabilities of advanced AI. CEO Sam Altman informed employees of the…
A developer outlines five critical observability requirements for production AI agents: full decision path, tool invocation details, per-step cost attribution, session context across restarts, and failure reconstruction.…
NVIDIA released the Nemotron 3 Ultra NVFP4 checkpoint, a quantized model that achieves up to 5.9x higher inference throughput than GLM-5.1 754B FP4 on decode-heavy workloads while matching BF16 accuracy. The checkpoint u…
OpenAI is postponing the general release of GPT-5.6 after the Trump administration requested a staggered rollout with government approval for each customer. The Office of National Cyber Director and the Office of Science…
The U.S. government issued export-control directives forcing Anthropic to take its latest models offline and prompting OpenAI to postpone GPT-5.6's release after White House officials raised security concerns. The action…
Anthropic published three cyber research posts in May and June 2026 that collectively suggest AI is reducing the time between a patch and an exploit. The posts cover exploit benchmarks, mapping malicious AI use to the MI…
Tencent is launching an AI assistant called Dayuan inside its enterprise collaboration tool WeCom, powered by DeepSeek's large language models. The tool can understand user requests and respond intelligently, aiming to b…
SynapCores released official LlamaIndex integration packages for its AI-native SQL engine, enabling RAG, GraphRAG, and hybrid retrieval in a single self-hosted binary. The packages replace multiple databases like Postgre…
Supermicro unveiled a Data Center Building Block Solutions Blueprint for the NVIDIA Vera Rubin NVL4 platform at ISC 2026, supporting up to 1,152 NVIDIA Rubin GPUs and 576 NVIDIA Vera CPUs in liquid-cooled racks. The pre-…