Probably a stupid take but what if

Agents with the ability to browse &amp; use a computer 

was the actual unlock for autonomous Agentic activity?

I’ve no doubt frameworks + plug-ins is one solution but,

Isn’t it easier if an agent can just do things the way a we do?

Just log on…

It’s much faster for agents to call APIs than use UI.

With a UI, agents are constrained by  physics. It takes time to move mouse, click, wait for render, move mouse, click again…

By contrast, API calls are only bounded by the speed of electrons over wire. 

AI Agents 🤝APIs

yuga.eth 🛡

👍 
https://x.com/yugacohler/status/1879728344545263681