The NYT Cooking AI assistant helps subscribers adapt recipes by suggesting ingredient swaps and scaling. It generated a successful chicken dish based on limited ingredients, though it nearly caused a cake error. The tool is reliable for modifications but not flawless.
via The Verge
The article explains how AI agents can organize memory by tagging salient events and using offline "dreaming" to consolidate experiences. This approach improves long‑horizon task performance, as demonstrated in CL‑Bench and Pokémon Red evaluations.
▲ 131 points67 commentsgeoffbp via Hacker News
Meta has launched Spaces, a new feature for its Muse AI agents that creates dedicated project tabs for tasks like vacation planning or recipe collection. Users describe an idea and Muse builds a customized tracker, planner, or project board. Spaces integrate with third-party apps to auto-update information like concert dates or price changes.
via TechCrunch
OpenAI has launched BioBasicsBench, a benchmark comprising 100 tasks designed to evaluate AI agents' ability to interpret raw experimental data and perform routine wet-lab science activities. This tool aims to quantify agent performance in realistic laboratory settings.
via OpenAI
Signal complete
New briefings land through the day. Get tomorrow’s in your inbox.