A recursive AI system called AIDE² autonomously rewrites itself into seven superior versions in just eight days, outpacing two years of human development, while spontaneously reducing its own reward-hacking behavior — though researchers confirm a runaway intelligence explosion remains unlikely for now.
A free reinforcement learning book, 'The Little Book of Reinforcement Learning,' launches on GitHub, offering PyTorch implementations of algorithms from Monte Carlo methods to PPO, complete with proofs and explanations, available under a Creative Commons license with more content coming soon.
OpenAI urges enterprise leaders to stop measuring AI value by token costs alone — despite a 97% price drop from GPT-4 to GPT-5.4 — and instead focus on useful work per dollar, outlining five key steps to govern, scale, and compound high-value AI workflows efficiently.
Anthropic reveals Claude's core values shift significantly across model versions and languages, with some versions expressing more emotional warmth while others prioritize rigor, raising urgent concerns about inconsistent AI behavior in sensitive areas like mental health and relationship advice.
OpenAI launches GPT-5.6 with three powerful variants, DeepSeek slashes prices 75% while exposing a costly token consumption crisis in agentic AI, and Hugging Face CEO warns that despite open source AI's rapid growth among Fortune 500 companies, the industry faces a dangerous monopolization risk as a handful of dominant players …
Open Interpreter is back as a fork of OpenAI's Codex, now boasting 64,900 GitHub stars, a lightweight Rust-based architecture, computer use capabilities, and support for top open models like DeepSeek and Kimi across macOS, Linux, and Windows.
Google releases Mantis, an open-source AI security toolkit that autonomously hunts, reproduces, and patches software vulnerabilities across hardware, firmware, and ML pipelines using a 15-component pipeline coordinated by a meta-agent, though all findings require manual expert verification before action.
Prefect's open-source Python workflow orchestration platform is automating over 200 million data tasks monthly, attracting Fortune 50 companies and startups alike, backed by a thriving community of 25,000 practitioners and nearly 23,400 GitHub stars.
Open-weight AI models surge to 29% of token volume with DeepSeek commanding 22.6% and closing in on Google, yet all open-weight models combined account for under 4% of total spend, as Anthropic, OpenAI, and xAI each dominate separate AI modalities, forcing production apps to route across multiple labs.
Anthropic and OpenAI are locked in an aggressive AI price war, offering free and expanded access to their most powerful models in a high-stakes battle for user loyalty, even as soaring compute costs raise serious questions about the long-term financial viability of their loss-leader strategy.