How to take apart a tech prophecy in five questions

I pasted an interview transcript into the chat and asked one thing, before any thought of a post: does the logic hold? Not "is it true?", not "what do you think?". Hold.

Inside was the prophecy you've read everywhere: in essence, money won't matter by 2036, because AI and robots will produce more than we can consume. Every so often a very rich man steps onto a stage and announces the future with a precise date, and the date travels the world before anyone checks the arithmetic.

Checking the arithmetic has become a reflex. These are the five questions I use, in order, applied to this case.

1. Who is he really talking to?

Before the content: what's the audience, and what counts as winning for the speaker? A grand claim in an establishment interview isn't meant to convince the interviewer: it's meant to produce clips for one's own audience. Once you see that, you stop asking "is he right?" and start asking "for whom is this sentence a win?". The answer often explains everything.

2. Is the evidence systematic or anecdotal?

A strong forecast should rest on systematic data, not chosen examples. "Robots will produce everything" is an anecdote about the future dressed up as an economic law. When someone stretches one case (manufacturing becomes nearly free) into a universal rule (everything becomes free), the alarm should ring.

3. Is it real inevitability, or self-fulfilling?

Beware of anyone who describes the future as a force of nature — in essence, all roads lead there, might as well join the club — precisely while he is one of the engines of that race. It's the subtlest trick: erase your own agency to erase your own responsibility. Inevitability, often, becomes true only because the actors behave as if it were.

4. Are we arguing inside the frame, or refusing it?

The best prophecies move the conversation into a "post-normal" world where today's rules no longer apply — which means you can't disprove them today. A forecast about 2036 is very convenient: unfalsifiable for ten years. When a claim can't be falsified in the short term, it isn't a prediction: it's a rhetorical bet.

5. Does the conclusion exceed the premises?

This is the final test, and almost always the decisive one. It measures the distance between what's been shown and what's being concluded. Here the defensible premise ("manufacturing becomes nearly free") is plausible; the conclusion ("money won't matter") sits several miles further out. And the check is quick: if, in the same interview, the proposed remedy is that governments will hand out purchasing power, the premise has just contradicted the conclusion. You don't need to wait until 2036 to notice — you need to read two sentences in a row.

Why I'm writing this (and what I have to do with it)

Because it's exactly my job: to read big claims and look for the crack. And because anyone who really builds with AI knows something the stage can't see: AI isn't a force of nature, it's a thing you put inside a fence. In the companies I run there's a written rule: every one that goes to market has to be something I can switch off in under an hour. Builders install the brakes; prophets say "it's coming anyway".

The five questions work for Musk as for the next person who climbs a stage with a date and a number. Keep them in your pocket: they cost nothing and save you a lot of wonder.

A future with a precise date is almost always a clip. The real future gets built without announcing it, one fence at a time.

The logic didn't hold. Five questions, and you could see it.