The attention economy optimizes for confidence, not accuracy — and confidence runs in the opposite direction of being right.
I built a database of 199 specific AI predictions from 70 speakers, then scored each one against reality. Not vibes. Not "who sounds smart." Actual predictions, with actual outcomes, scored on a 0-to-1 scale.
The pattern that emerged was immediate and disturbing.
Scott Alexander — a pseudonymous blogger with roughly 50,000 readers — scored 0.88 accuracy across 9 predictions. He hedges. He qualifies. He publishes probability distributions instead of hot takes. Almost nobody has heard of him.
Ed Zitron — a tech podcaster with millions of listeners, regular media appearances, and a voice built for broadcasting — scored 0.30 accuracy across 7 predictions. He's confident, entertaining, and wrong about nearly everything.
The middle of the leaderboard is almost empty. You're either accurate and obscure, or confident and everywhere. There is very little in between.
It's a structural feature of how attention works, not an accident.
This is why the state of AI predictions is sad: the full dataset, with every prediction scored: The Sad State Of AI Predictions
Philip Tetlock spent 20 years studying expert prediction. He tracked 284 experts making 82,361 predictions across politics, economics, and geopolitics. His conclusion: the experts who were wrong the most were also the ones who appeared on television the most.
He divided forecasters into two types. Foxes — who know many things, hedge constantly, and update their views — were significantly more accurate. Hedgehogs — who know one big thing and apply it to everything with total confidence — were terrible forecasters but phenomenal media personalities.
The traits that make someone a good predictor are the opposite of the traits that make someone a good influencer.
| Trait | Good Predictor (Top of Leaderboard) | Good Influencer (Bottom of Leaderboard) |
|---|---|---|
| Framing | "It depends on multiple factors" | "Here's the formula" |
| Confidence | "60% probability" | "ZERO CHANCE" |
| Speed | "I'd need to think about that" | Instant confident answer |
| Output | Publishes spreadsheets and probability tables | Publishes hot takes and manifestos |
| When wrong | Updates publicly ("I was wrong about X") | Never acknowledges errors |
| On camera | Boring, hedged, qualified | Magnetic, certain, quotable |
| Priority | Being right | Being heard |
Intelligence comes with a set of traits that are actively punished by the attention economy. Every instinct that makes someone a careful thinker is an instinct that kills their reach.
| Intelligence Trait | Why It Kills Influence |
|---|---|
| "I'd have to research that before answering" | Immediate disqualification as a public figure |
| Prefers depth over speed | Social media rewards volume, not quality |
| Analysis paralysis / perfectionism | Delays posting while competitors ship daily |
| Uncomfortable with performative vulnerability | Audiences reward emotional openness, not logic |
| Updates views when evidence changes | Looks "inconsistent" instead of principled |
| Avoids oversimplification | Complex ideas don't fit in 280 characters |
| Disdain for shallow content | Refuses to create engagement-bait algorithms reward |
| Quiet observer by nature | Platforms reward loudness and frequency |
It's a structural mismatch, not a character flaw. The attention economy was built to surface the most engaging thinking, not the best thinking — and those are rarely the same thing.
The business podcast ecosystem runs on a specific psychological loop that has nothing to do with teaching you anything useful.
Barry's Economics documented how success media creates a closed system: you feel anxious about your performance, consume content that promises solutions, get a temporary dopamine hit from the inspiration, then return next week because nothing actually changed.
Mathematician Abraham Wald showed the military they were armoring the wrong parts of their planes — the bullet holes they could see were on the planes that survived. The planes that were hit in critical spots never came back to be studied. Success podcasts interview the equivalent of the surviving planes. The founders who did the same things and failed never get booked.
Neuroscientists Gazzaniga and Libet demonstrated that the brain constructs narratives after decisions are made, not before. When a successful founder tells you their "strategy," they're often reverse-engineering a story from random events. The narrative feels true to them. It just isn't replicable.
Sociologist Duncan Watts ran an experiment with 14,341 participants who rated the same set of songs. In one group, a song ranked #1. In another group, the exact same song ranked #40. Success was almost entirely determined by early social signals, not quality. The "best" song depended on which group you were in.
| Step | What Happens | What You Think Is Happening |
|---|---|---|
| 1. Create anxiety | "You're not optimized enough" | "They understand my struggle" |
| 2. Promise solution | "This CEO cracked the code" | "I'm about to learn the secret" |
| 3. Deliver partial satisfaction | Inspiration hit, no lasting change | "I just need to try harder" |
| 4. Repeat | Next episode, next guest, next framework | "Maybe this one will be the one" |
| Revenue model | You stay anxious and keep returning | You think you're learning |
This matters right now because we're in the middle of the most consequential technology shift in decades, and the people shaping public understanding of AI are structurally selected to be wrong about it.
See the full analysis: The Sad State Of AI Predictions
New research on prediction accuracy, AI trends, and which forecasters are actually worth your time. Sent only when there's real data behind it — not on a schedule.
— Scott Covert, who actually built this. I build with these tools daily, not just write about them. If this raised a question, or you've got a project that needs building, that's worth a message. Reach me, pitch a collab, or book a consult.