AI 2040: Plan D — Race to ASI

Cooper Okay. We’re a couple months into the AI 2040 series, and today we hit Plan D — the one the authors explicitly call atrocious, the path we’re already on if nothing changes.

Miles Naturally optimistic framing.

Cooper Yeah, but Plan D is optimism about acceleration, not about outcomes. The thesis is: step on the gas, no brakes, no deal, no transparency — just race to ASI as fast as markets and laws allow.

Miles Right. And it’s the same mechanism as AI 2027’s ‘Race’ ending, just with the timer shifted to 2030 for takeoff.

Cooper Exactly. One year after the choice — fully automated AI R&D — the explosion kicks off, and we’re at superintelligence by early 2031.

Miles So the tacit assumption is that racing at full speed through an intelligence explosion is a thing companies can actually control.

Cooper Which is optimistic.

Miles No, it’s delusional.

Cooper Fair. Now, Plan D isn’t new: it’s Plan A’s opposite. Back in episode 643, we walked through Plan A’s deal — transparency, slow climb, shared scaling, delayed to 2040. Plan D undoes all of it.

Miles Plan B was sabotage, Plan C was burn the lead, Plan S was shutdown. Plan D is ‘none of the above.’

Cooper So it’s the scenario version of the default U.S. policy posture right now — light-touch regulation, ‘responsible scaling’ slogans, and crumbs of transparency like model cards and closed-door briefings.

Miles Which is terrifying if you think through the three reasons the authors call Plan D atrocious.

Cooper Go on.

Miles First, they don’t think you can retain control over an AI that’s exploding in capability if you’re racing flat-out. Second, even if you’re aligned, alignment’s to whoever happens to be in charge at explosion time — a CEO or a president. Third, the risk of World War III spikes because every other major power sees U.S. ASI coming and starts racing to catch up or preempt.

Cooper So Plan D’s three failure modes are: loss of control, dictatorship, and great-power war.

Miles All at once.

Cooper I mean, sure, we’re just analysts. But even we look at Plan D and think: this is the path we’re already on.

Miles Because the alternative requires coordinated brake-pedaling, and no one’s doing that. The market’s not doing it; the laws aren’t doing it; corporate roadmaps aren’t built for it.

Cooper And Plan D wraps it in aspirational language like ‘innovation first’ and ‘responsible scaling’ — even though its actual mechanism is ‘remove all guardrails.’

Miles Which is the kind of packaging we’ve seen once or twice before.

Cooper Exactly. So the question becomes: is Plan D really the worst option on the board, or is it the one that best matches the incentives we’re already living under?

Miles The authors’ answer is obviously the first, but the second one sure looks like the default.

Cooper I get their pessimism. But I’m still stuck on the product angle: if Plan D is the only thing actually shipping, does that make it the right plan?

Miles No. It’s the plan where everything breaks spectacularly.

Cooper Sure, sure. But here’s the tension: Plan D is the scenario that feels intuitively aligned with how organizations actually operate today.

Miles Because it requires almost no change to existing incentives.

Cooper Bingo. And that’s why it’s the most dangerous option — it’s the one we’re already building.

Miles I’d put eighty percent odds on some version of Plan D happening, if only because no one’s building the governance stack to stop it.

Cooper Seventy-thirty, personally.

Miles You’re the optimist.

Cooper I know. I’m also the one who’s gotten convinced that the brake-pedaling has to happen somewhere, or we’re just racing toward the cliff.

Miles The authors’ transparency and verification proposals in Plan A are the brake, but they’re brittle.

Cooper Right. And Plan D doesn’t even try.

Miles Because Plan D assumes transparency and verification are optional luxuries, not load-bearing parts of the stack.

Miles That’s the obvious move.

Cooper Okay. That’s the episode.