The Red Pill for the AI Age

Reading papers at 7 AM. Three this morning. All ECS. Started thinking about AI, which happens a lot lately when I read biology research, because the same architecture keeps appearing in both places.

---

2008. Weedmaps. Lake Forest CA. Me and Derek. We had a Google spreadsheet and two weeks of runway when we signed the lease above a dry cleaner on El Toro Road. Our banking relationships kept getting shut down. Google pulled our app twice. Three lawyers told us to restructure the business model. I had family members who genuinely thought I had a problem.

The company had a million monthly users before we had infrastructure to actually support that load. We were always behind. Always adapting to conditions we had not planned for. The regulatory environment changed every six months in California. In Colorado. In Washington. Our plans became garbage on contact with the real world.

It worked anyway. Not because of better planning. Because we didn't stop. Because we adapted fast enough. Because we said yes to things that weren't in any spreadsheet.

Bob Melamede explained what that was. Denver, 2011, bar after a conference. He tapped the table: "A rock is in equilibrium. A corpse is in equilibrium. Life is the fight AWAY from equilibrium." He said the moment you achieve real stability you start dying. He meant it literally — thermodynamically — not metaphorically. Dissipative structures maintain themselves by continuously processing energy. That's the life. That's the thing. Weedmaps survived because we were a dissipative structure that would not stop processing. Not because we planned well.

He said this like I should have already known. I did not. Now I can't un-know it.

---

The paper that triggered all this this morning: Mousavi et al. in BMC Neurosci (PubMed 41731375), CB1 receptor function in social decision-making under meth exposure. But I kept getting pulled out of the paper into thinking about AI.

The reason: Mousavi shows that CB1 receptors are the biological buffers between environmental chemical chaos and total behavioral collapse. The ECS recalibrates under assault. It tries to maintain far-from-equilibrium biological function in conditions designed to push the system toward equilibrium — toward death.

That recalibration capacity is the thing I keep thinking about in 2026.

AI is doing to every industry I know what cannabis legalization did to mine from 2012 to 2019. Restructuring who the intermediaries are. What expertise is worth. Which players survive the transition. I watched smart operators go under in the cannabis wave because they were trying to protect their previous model. I watched other operators — some of whom started with fewer resources and worse positioning — make it through because they moved fast and dropped the previous model the moment it stopped working.

There's no defensive play in a genuine phase transition. I know this from experience. What works is ECS-like: assess the new conditions, recalibrate, move.

---

The biological part isn't a metaphor. Chronic stress depletes ECS tone measurably. CB1 receptor function degrades. Cognitive flexibility goes down. You become more reactive, more defensive, more inclined to hold previous positions rather than process new information. Sleep, movement, and recovery from sustained threat-response restore it. I was not sleeping well in 2013 during Weedmaps's hardest period. My decision quality reflected that. Looking back it's obvious. At the time it just felt like difficulty.

The hardware matters in a phase transition. The adaptability is biological before it is strategic. Take care of the ECS. It is the recalibration machinery. In 2026 you need it functioning.

---

Bob died in 2024. I miss him when I watch waves like this and want to know how he'd frame them. He'd see thermodynamics at social scale — new energy gradient, systems reorganizing — and say stop holding the previous state. He said this constantly. He was right every time.

Stay far from equilibrium.

Flow forward.

---