Stay informed with the latest trends, tutorials, and breakthroughs in artificial intelligence from the DAIR.AI team.
February 26, 2026
We tested Google's new Nano Banana 2 image generation model (gemini-3.1-flash-image-preview) against Nano Banana Pro across six experiments. Here's what improved in text rendering, object fidelity, search grounding, and what it means for your image generation workflows.
February 25, 2026
ETH Zurich evaluates whether AGENTS.md and CLAUDE.md context files actually improve coding agent performance. Human-written files help (+4%), LLM-generated files hurt (-2%), and all files add 20%+ cost. Introduces the AGENTbench benchmark.
February 17, 2026
LCM introduces deterministic context management for LLM agents, replacing model-written recursion with engine-controlled primitives. Its coding agent, Volt, outperforms Claude Code on the OOLONG long-context benchmark. Here's how it works and why it matters.
February 16, 2026
The next big thing might be recursive language models (RLMs). Instead of cramming text into the context window, RLMs treat text as an external environment that the model programs against. Here's why an 8B model can rival GPT-5 on long-context tasks.
February 3, 2026
A hands-on tutorial on using the image generator plugin in Claude Code. Learn how to generate images from text prompts, edit existing photos, create blog covers, product mockups, logos, and more - all from your terminal.
February 1, 2026
We're excited to launch DAIR.AI Academy, a new learning platform designed to help you master AI through expert-led courses, hands-on projects, and a vibrant community of builders.