TLDR:

Agentic AI

Moltbook

• Implications -
• Predictions - Everyone will have their own AI Agent (that will harness subagents) by the end of this year. So, for example, let’s say you agree to book a meeting via email. Your contact’s personalized agent will then reach out to your personalized agent (which will be running 24/7) and go back and forth to find a suitable time based on each of your known goals and past preferences. Each human user will then be notified and asked to approve of the action.
• The future of AI is looking less like a centralized entity, and looking a lot more like a decentralized human-ai merger.
• Ideas

Some folks are saying that whereas 2025 was the year of Agents, 2026 is shaping up to be the year of Agent Harnesses.

Frontier Models

Another big announcement in the past week has been the release of Kimi 2.5

What I think this demonstrates is just how quickly Open Source models are able to catch up to Frontier models. Now, the catch is that these models are only possible because of the billions of dollars being invested

Sacra reports here

How will these AI companies get to the point where they make revenue (before they run out of spend)?

Two very different approaches to building AGI: Models that can Code Better Models vs Building a General Intelligence. One has a mid-2027 delivery versus a mid-3033 general AI emergence with competent robots. https://www.youtube.com/watch?v=02YLwsCKUww

Two very different motivations Dario: “we pioneered this method of mechanistic interpretability (to make safe and truthworthy models)" Demis: “(I want to build) the ultimate tool for science and understanding the world around us"