Everybody's Faking It: Why You Should Distrust Anyone Who's Certain About AI

Stand near the bar at any tech networking event right now and you'll hear it within about four minutes. Somebody, usually holding a drink they haven't touched, telling a small circle exactly what's coming. Which jobs are gone in eighteen months. Which industries are "over." What model is about to make every other model obsolete. All of it delivered with the flat, unblinking confidence of a person reading off a spec sheet nobody else got to see.

Scroll the comment section under any AI post and you'll find the same energy. Total certainty, both directions. "This changes everything" right next to "this is all hype and it's going nowhere," both stated as settled fact by people who, as far as I can tell, are working off the same publicly available information as everyone else in the thread.

I've started treating that certainty itself as the thing to be skeptical of. Not the claim. The tone.

The people actually building this rarely talk like that

Here's what I keep noticing. The researchers and engineers who are actually inside these labs, actually training the models, actually running the evaluations, tend to talk in a completely different register than the loudest voices online. Ask them what's coming in two years and you'll get some version of "it depends," or "we're seeing some things that surprise us," or, refreshingly often, "we don't fully know yet."

That's not evasiveness. That's what honesty sounds like when you're close enough to a fast-moving, genuinely uncertain field to know how much you don't know. The people with the most actual information tend to hold their claims the loosest, because they've seen how often last year's confident prediction turned out wrong.

Compare that to the guy at the bar. He's not closer to the information. He's further from it, and covering the distance with volume.

Why the confident ones get the biggest audience

This isn't really a mystery once you think about the incentives. Certainty is easier to consume than nuance. "Here's exactly what's going to happen" gets shared. "It's genuinely unclear and depends on a dozen variables" does not. The algorithm doesn't reward accuracy. It rewards a clean, quotable, confident line.

So a specific kind of creator has emerged, and you've probably noticed them without naming the pattern. Someone who was deep in crypto eighteen months ago, deep in some other trend before that, and is now speaking with total conviction about what AI means for your business, your job, your industry. The topic changed. The certainty didn't. That should tell you something about where the certainty is actually coming from. It's a performance built for engagement, not a conclusion built from evidence.

None of this means every confident voice is faking it, or that expertise doesn't exist. It means confidence alone was never proof of anything. It's just easier to listen to than "it depends," which is usually the more honest answer in a field that's still this new.

What to trust instead

I'd rather hear from the person who says "here's what I've actually seen work, for me, in my business, this month" than the person who tells me with total certainty what's happening to an entire industry six months from now. The first person is giving you a real data point. The second is giving you a guess dressed up as a forecast.

Ask yourself, next time someone speaks in absolutes about AI: how would they actually know that. Are they inside the labs building this. Are they running the studies. Or are they someone with a large following and a strong incentive to sound sure, because sounding sure is what gets shared.

This field moves fast enough that anyone claiming to have it fully figured out is telling on themselves. The honest position, most of the time, really is "here's what I've seen, here's what I think, and I could be wrong." That's not weakness. That's just what paying close attention to something genuinely new actually sounds like.

Distrust the certainty. Trust the person still willing to say "it depends."