Stay ahead on AI.
All in ONE place.

Google has made AlphaEvolve generally available to all users. AlphaEvolve is a code optimization and discovery agent built on Google's Gemini model, designed to help organizations, researchers, and developers tackle some of their most difficult business problems. The tool uses AI to automate and improve code-based problem solving at scale, broadening access beyond any earlier limited release. Its general availability marks a significant step in Google's push to bring advanced AI coding agents to a wider audience.

Microsoft AI has partnered with poet and Poetry Pharmacy author William Sieghart and creative studio Gravity Road to launch Ode Poetry, a digital experience at odepoetry.ai that uses MAI audio models to replicate Sieghart's one-on-one poetry consultation sessions at global scale. The app uses MAI-Transcribe-1.5, which achieves a 4.86% Word Error Rate on the FLEURS benchmark, and MAI-Voice-2 to recreate Sieghart's distinctive vocal style from a short audio sample.

Anthropic has reset the hour and weekly rate limits for all Claude users. The reset affects both limit types simultaneously across all user tiers. Rate limits on Claude typically restrict how many messages or tokens users can send within a given timeframe, and resets like this temporarily restore full access for users who had hit their caps. No reason was given for the reset, but it offers immediate relief to heavy users who had reached their usage ceilings.

Anthropic has launched Reflect, a beta tool that lets Claude users track and visualize their AI usage patterns across the past 1, 3, 6, or 12 months. Available in Settings on Claude's web and desktop apps, the dashboard shows key topics, usage patterns, and task types, and uses a 4D AI Fluency Framework covering delegation, description, discernment, and diligence. Users can set quiet hours or break reminders. The feature is available to Free, Pro, and Max users with Memory enabled.

OpenAI has audited SWE-Bench Pro, a widely used coding benchmark, and found that approximately 30% of its 731 tasks are broken. Using a pipeline combining automated screening, Codex-based investigator agents, and review by experienced software engineers, OpenAI flagged 200 to 249 broken tasks. Issues include overly strict tests, underspecified prompts, low-coverage tests, and misleading prompts. OpenAI is retracting its earlier recommendation to adopt SWE-Bench Pro and is urging the evaluation community to build new benchmarks designed specifically to test model capabilities.

OpenAI has published its National Security Principles, outlining how its AI technology can and cannot be used in government and defense contexts. Key restrictions include no mass domestic surveillance, no autonomous weapons direction, and no high-stakes automated decisions. The principles cover existing partnerships, including with the U.S. Department of War. OpenAI has also established Trusted Access for Cyber partnerships with nine allied nations under its Daybreak program, and expanded access to its GPT-Rosalind model for biodefense missions.

Mistral AI has launched Robostral Navigate, an 8B model enabling robots to autonomously navigate complex environments using only a single RGB camera and no depth sensors or LiDAR. It achieves 76.6% success on the R2R-CE unseen benchmark, beating the best multi-sensor systems by 4.5 points. Built entirely in-house and trained on 400,000 simulated trajectories across 6,000 scenes, it runs on wheeled, legged, and flying robots. Token-efficient prefix-caching reduced training time from months to days.