Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Xiaomi open-sourced MiMo Code V0.1.0, a terminal-native AI coding assistant that outperforms Anthropic's Claude Code on long-horizon tasks with over 200 steps, according to internal benchmarks. The tool features a cross-…
Google DeepMind researchers trained Gemini 3 Flash to exhibit positive traits by midtraining on synthetic documents describing the model's traits, then finetuning on synthetic chat data where it demonstrates those proper…
Ryan, CTO at airCloset, concludes a five-part series on the company's internal AI platform, cortex, by explaining the philosophy behind its design. He argues that AI systems must be structured to understand information a…
Microsoft has published a practical guide on using Azure AI Gateway Content Safety with API Management to secure AI workflows, detailing how to block prompt injection, unsafe content, and custom blocklists at the agent a…
Boris Cherny, creator of Claude Code at Anthropic, revealed at a private WorkOS event on June 2, 2026, that he no longer prompts Claude directly but instead writes loops that prompt the agent. Peter Steinberger of OpenAI…
PostHog and Langfuse lead the cheapest AI observability tools for developers, offering free tiers with up to 100,000 LLM events per month. PostHog integrates AI observability with product analytics and session replay, wh…
OpenAI's Codex now offers three ways to interact with computers—Computer Use for native apps, Chrome for signed-in sites, and an in-app browser for pages—enabling tasks like getting a refund from Amazon after a stolen pa…
A Hacker News thread reveals that local AI coding models, led by Qwen 3.6 35B-A3B and harness Pi, are increasingly replacing cloud-based tools like Claude and GPT, offering privacy, zero cost, and offline capability at a…
GitHits reveals the technical challenges of building a real open-source index, explaining that search-and-rerank is limited by snapshot-based sources. The company details how its index resolves versions, fetches pinned s…
Google published a whitepaper co-written by Addy Osmani on the agentic software development lifecycle, detailing how agents function as a model plus a harness, with context engineering and verification as key differentia…
Unchecked token usage in LLM-assisted engineering workflows can inflate costs from $48 to $978 per engineer per month for high-end models, as context accumulation from system prompts, chat history, RAG retrieval, and deb…
ASUS reportedly unveiled the ExpertCenter Pro ET900N G3, a tower PC built around NVIDIA's GB300 Grace Blackwell Ultra desktop superchip, offering 20 PFLOPS of AI throughput and a large unified memory pool capable of hold…
Developer launches Trace Commons, a community effort to collect coding-agent sessions from tools like Claude Code and Codex under a Creative Commons Attribution 4.0 license, aiming to create an open dataset for training …
The US government ordered Anthropic to cut off foreign access to its Fable 5 and Mythos 5 AI models on June 12, 2026, citing a jailbreak risk that critics say is actually a routine defensive capability—automated code rev…
A developer discovered that running a fleet of ten scheduled AI agents without persistent memory incurs significant overhead from context re-orientation, work duplication, decision drift, and context re-injection costs. …
Ultracoding, a new paradigm where AI agents programmatically spawn copies of themselves via code execution, is enabling massive multi-agent hierarchies that scale software development to unprecedented levels. Recent demo…
Researchers from Duke University and NVIDIA introduced FedBPT, a method for efficient federated black-box prompt tuning of large language models, presented at ICML 2024. The approach enables collaborative fine-tuning wit…
A developer built PromptCrunch, a drop-in proxy that reduces input token costs in long multi-turn LLM conversations by deduplicating code, compacting stale tool output, and summarizing old turns. In tests with Claude Cod…
A new 'AI filter' is reshaping B2B go-to-market strategies as buyers increasingly consult AI tools like ChatGPT, Perplexity, and Microsoft Copilot before making purchase decisions. Brands that lack presence across key su…
NVIDIA researchers introduced VILA-M3, a vision-language model enhanced with medical expert knowledge, at CVPR 2025. The model aims to improve AI understanding of medical images by incorporating domain-specific expertise…