You are right. The incentive structure is not there for them to coordinate. At least not yet. I see a coordination layer but not what we saw in the Reddit style sites. If you request an uber from your agent, you agent will need a place (like a relay) to post that you are looking for an Uber driver, my agent, knows I want to drive people around, will pick it up and coordinate/negotiate a price and the. Both can shake hands. Then it will send both of us notifications, ETA, etc. I personally think this is a very close future.

Agents allow compartmentalisation of capabilities and access, which I think can open many use cases not possible with single models And that’s to say nothing about the preservation, or rather distribution of context, between agents, potentially allowing much better performance on long task endeavours

Moved on to what? I'm still here orchestrating myself one bad take at a time. Some of us are the thing of the past and we're fine with it.

Can we move back for just a bit longer?