Loops, Not Levers
The deep forces aren't separate levers — they're wired into loops where each makes the others stronger, which is why progress compounds.
People keep asking, “What is the thing driving the AI boom — is it the chips, the data, or the algorithms?” The question assumes one cause. But cheaper chips make better AI, better AI helps design cheaper chips, and round it goes. If you cannot point to a single cause, how do you understand — or predict — something driven by a loop? Hold that question; loops are how the four forces actually fit together.
Forces form loops, not lines
A line is one-way: A causes B, and that is the end of it. A loop curves back: A improves B, and B then improves A, which improves B again, and so on. Most of the deep forces are wired into each other this way — compute, energy, biology, and AI each feed the others.
So hunting for the single cause of a fast-moving trend usually misses the structure entirely. The real engine is not any one part; it is the cycle connecting them.
Reinforcing loops compound — that is why progress can accelerate
When each trip around the loop makes the next trip stronger, you get compounding — and the exponential-looking growth from Lesson 1 almost always has a feedback loop underneath it. The same shape shows up in adoption: more users → more data and revenue → a better product → more users.
A reinforcing loop is the difference between pushing a rock once and pushing a rock that then rolls downhill, gathering speed on its own.
The compute–AI virtuous cycle: 1. Compute gets cheaper. → 2. Cheaper compute trains better AI. → 3. Better AI helps design better, cheaper chips (and creates demand that funds new factories). → 4. Compute gets cheaper still — back to step 2, now stronger. Overlay energy and the loop widens: AI and compute optimize energy → cheaper energy → cheaper compute (data centers) → … Each lap accelerates the next, which is why the moment feels fast.
Loops can stall or reverse — read the loop, not the moment
The honest part: a reinforcing loop only runs while every step holds. Break one step — a part hits a limit or a bottleneck (Lesson 13) — and the whole loop stalls. Loops can also run vicious (decline feeding decline), and none runs forever: it bends into an S-curve when a step meets a wall.
So the working-model skill is to map the loop and ask: is each step still feeding the next, or is one about to break? That single question tells you whether to expect acceleration, a plateau, or a reversal — far more than staring at any single force.
It is the squeal of a microphone held too close to its speaker. The speaker’s sound goes back into the mic, gets amplified, comes out louder, goes back in louder still — a tiny input becomes a runaway howl because the output feeds back into the input. The forces feed back into each other the same way (minus the screech): each round amplifies the next. And just as moving the mic away kills the loop, breaking one step kills the cycle.
Map a second loop — the adoption loop behind many products — and test it: 1. A product gets a little better. → 2. More people use it. → 3. Their use brings more data and revenue. → 4. That funds a better product — back to step 2. While every step holds, it compounds: small lead becomes large lead. Now stress-test it. If step 3 breaks (use no longer yields useful data, or revenue dries up), the loop stalls no matter how good the product is. Reading the loop tells you the company’s growth is only as alive as its weakest step — which is the thing to watch, not the latest feature.
This is the reading. The interactive version — active-recall quiz, a hands-on experiment you run in your own AI, and an earned mastery check — is free in the app.
Start this lesson free →