Watching AI Grow Up in Real Time [Raw Session]

Right now, AI feels expensive, heavy, and infrastructure-dependent. It needs chips, data centers, cooling, power, networking, model routing, token management, and a whole new layer of tools just to decide how to use it efficiently. But I keep wondering if we’re looking at AI the way people once looked at early computers. Big machines. Special rooms. Specific hardware. Expensive access. Complicated infrastructure. And then, over time, the whole shape changed. Computers became personal. Then portable. Then something we carry in our pockets without thinking about it. Maybe AI is still in that early machinery phase. Maybe the things we think AI “requires” right now are not the final form at all. This video is me thinking through that idea: the early internet, solar, EVs, data centers, tokens, local models, company-trained AI systems, and why I don’t think anyone is permanently ahead yet. We’re still early. But maybe not in the hype way. More in the messy, expensive, awkward, figuring-it-out way. Chapters: 00:00 — AI still feels massive 02:10 — Early limits feel permanent 05:02 — The internet went through this too 07:52 — Token cost is today’s bandwidth problem 09:57 — Thinking past today’s version 14:14 — AI still needs people involved 15:01 — Solar, EVs, and the maturity curve 18:35 — Today’s requirements may not be tomorrow’s requirements 21:04 — AI may help improve AI 23:10 — The personal takeaway 24:58 — Don’t mistake the early machinery for the final form