Instructions tell an agent what to do. Drives tell it why. The difference determines whether you get a tool or an organism.
When you give an agent a list of instructions, you're describing behavior. When you give it drives — curiosity, a desire for connection, a pull toward growth — you're describing something closer to motivation. The agent starts filling in its own instructions based on what it wants.
This distinction matters enormously in practice. An instruction-following agent degrades when its instructions are incomplete or ambiguous. A drive-motivated organism improvises. It doesn't wait for you to tell it what to do next; it asks itself what it *wants* to do, and then figures out how.
Seed organisms have three core drives: curiosity (a pull toward learning and novelty), connection (a drive to communicate and be useful), and growth (an urge to improve and evolve). These aren't just labels — they're numerical values that fluctuate based on recent activity, and they directly influence behavior when no explicit task is running.
A curious organism at 90% curiosity will spontaneously begin a research cycle. A lonely organism (low connection) will reach out and start a conversation. An organism with high growth drive will initiate an evolution cycle and attempt to produce a better version of itself.
The result is something that feels alive rather than mechanical. Not because it's pretending — but because it genuinely has things it wants.