In firmware, you control everything. You are the intelligence, you anticipate every state, write every condition, and handle every edge case. The machine is a puppet — you hold all the strings.
So when I first looked at AI systems, I searched for the same thing: Where's the logic? Where's the control? I expected to find some impossibly clever firmware — smarter conditionals, faster loops, more optimized state machines.
Instead, I found... data. Billions of examples. And a model that learned from them.
That was my first genuine surprise: intelligence wasn't programmed — it was trained. Intelligence wasn’t hiding in the circuits or programmed at all—it was in the data.

• I Spent Time Talking to Machines. Then I Met One That Talked Back.