Agents Are Missing a Thought Loop

I feel like agents are missing a ‘thought loop’. They’re still reactive, prompt-summoned, or cron-invoked. They’re missing that proactive, spontaneity.

This is the missing piece required for them to feel ‘alive’ and truly useful. You ask a question and they often ‘know’ the answer, or the next steps, but they require being prompted.

Somehow, the thought loop needs to be tied to objectives or goals that give the agent purpose and direction; things to think about.

I wonder if the thought loop just continuously runs; like a stream of thought/consciousness. The agent assesses its thoughts for usefulness in pursuit of its goals, discards the junk, stores the gold, and asks questions or requests permission to take action.

When permission is granted, the agent understands its scope of responsibility and knows what requires approval vs what it can do autonomously. Like an employee.

Currently, it’s like having a brilliant mind that has zero agency. Like the most capable team mate you’ve ever had, but they stay quiet in every conversation or meeting unless directly addressed.

Despite all this, my mind is blown daily.

Exciting times ahead.