Two questions I avoid asking these days are:

• “What’s scarce now that intelligence is abundant?”

• “What’s the new moat?”

Both are imo mind-killing dead ends.

The first is the reflexive incumbent response to disruption and will cause you to retreat to an essentialized “upmarket humanism” which you’ll harvest for a while and then die. This is a thinly veiled version of climbing a sacredness gradient when the right move is in fact to chase the profanity gradient. Notice the next time you see this “scarcity” frame people don’t actually make a scarcity argument. They make a sacredness argument.

The second question will have you fracking a series shrinking pockets of transient advantages as you retreat to smaller and smaller besieged islands. Moats have always been an incoherent business metaphor, but historically they’ve been actually used to mean “big monopoly rent based on preferential attachment effects” which is a barrier to entry against asymmetrically smaller adversaries, not a defense against peers or larger adversaries. With individuals navigating AI, using the “moat” metaphor is a self-inflicted siege condition against a larger force. If you must retreat, don’t retreat to a castle behind a moat. Flee into the hills at least. It’s less stupid and you’ll survive longer.

I don’t expect to influence anyone who loves those questions and the lofty, self-affirming thoughts they spark, because they reveal a very strong humanist identity. But if your fundamental reaction to both is meh because you don’t orient around scarcity and don’t like to retreat behind moats, I have 2 different questions you can ask:

• “How can I waste the abundance now?” — looks similar to the first question, but subtly different framing.

Yes, obviously tokens/intelligence is the thing to waste, but the key idea is to pursue waste rather than utility. Silliness, triviality, self-indulgence, absurdity — these are the signals to follow in deciding how to use your $20/mo worth of tokens (or whatever your spend level is — $20 is already enough to waste). This is much harder than it looks. I keep getting distracted from my waste projects by utility projects.

This is true even for big corps suddenly hit with millions of dollars of bills. That’s not true waste but inefficient pursuit of value. You want true waste. Which means the goals have to be somewhat silly.

In the og “waste transistors” heuristic of Alan Kay, the waste was stuff like GUIs in the PC paradigm. In the PDP-1 era before this waste principle got enshrined, the wastage was to use even scarcer compute to play early video games. In the phone era we learned to waste photography to replace typing.

How to waste tokens is a very non-trivial question.

My substitute for the “moat” question (I mean, really, do you want to pull up a drawbridge and cower fearfully with a besieged mindset as your AI “strategy”?) is:

• “What’s the new network primitive?”

Moats imply a containers-and-walls mindset. Networks imply a nodes and edges mindset. The baseline answer here is of course stigmergic localized contexts with agents backed by different models crawling around them. What I’ve called intelligence media. But it’s not yet obvious what the networking primitives are, the way social graphs and interest graphs based on follow and like relationships were with social media. One I like is model-sampling, where an output is a blend of multiple weighted finetuning as with TITLES. There are others.

I’m working on a top-secret one myself where the primitive is based on the changelog/diff. I’ll share more in a few months, if Fable turns out to be as smart as it claims.