I'm Rahul Nath.

My bet: local models will dominate.

I think local models will eventually dominate, even at the level of AGI. As more intelligence runs on our own machines, we'll need a durable way for humans to interact with it, supervise it, and decide what it can do.

That's the direction behind aidashos: plans, decisions, work, and feedback that outlive any one agent session. The history of those interactions could become useful in many ways to each person running the system, from remembering why a decision was made to learning from past work and shaping how their agents help them.

There's a lot of work to do on this. If you're interested, DM me on Instagram (@rah_juul), Twitter (@rahulkindarules), or LinkedIn.