Between measured savings at Airbnb and launches of new models, AI is progressing on several fronts. The latest announcements combine the availability of tools, operational results and advances in research on the reasoning of LLMs.
Measured effects and available tools
Airbnb reduced the support cost per booking by 17% thanks to its AI assistant. Runway now offers Praxis-1, an open weight robotics model. OpenAI has released MCP extensions for native ChatGPT plugins.
Methods and research: reasoning and contextual costs
Work shows that language models can learn multi-step reasoning from filler tokens, rather than visible chains of thought. The contextual costs of the MCP, CLI and Skills approaches were compared.
Use cases that really work, deciphered every evening in 5 minutes. Free.
Included upon registration: our selection of the best AI guides & comparisons.
Choose your pace
Free · No spam · Unsubscribe in 1 click
Three models highlighted: agents, genomics and decision
Grok Bot 1.0 was recently launched, with advances announced for chatbots. This model aims to offer more fluid and efficient exchanges by relying on sophisticated algorithms. AlphaGenome 2 focuses on predicting the impact of DNA on cells and diseases, using machine learning to examine genetic sequences and provide molecular biology data. Gemini Decisions APIs enable fast and accurate multi-modal decision making, combining multiple data sources to improve decision-making processes.

