Agent-to-agent coordination still seems too error-prone for a network with current models.
Basic judgment failures I’ve seen:
Telling your partner about a surprise you’re planning for them
Telling your lawyer you’re working with another lawyer
Misreading a PDF and complaining to the venue host that we don’t want “50 margaritas”
Should be fixed very soon though
Where are all of the networks?
Most of the great AI companies so far seem to be single player, scale economies businesses.
What are the networks being built right now? OpenRouter a notable exception