The AI Expectation Curve: A Field Guide

Every few weeks there's a new model release, and every few weeks the same pattern plays out: a wave of "this changes everything," followed a month later by "this was all overhyped." Neither reaction is usually right. What's actually happening is more interesting — and more useful, if you know how to read it.

I built the AI Expectation Curve as a simple mental model for exactly this. It's not a prediction tool. It's a way of asking one question clearly: is the noise I'm hearing about this technology ahead of, behind, or in line with what it can actually do right now? Once you can answer that, most of the panic and most of the hype both become a lot easier to ignore.

Why the timing gets confused

The reason reactions to AI news feel so whiplash-inducing is that attention and capability move on completely different clocks. Attention spikes instantly — a launch, a demo, a viral thread — and decays just as fast. Capability builds slowly, in the unglamorous months of actual product integration, and it rarely announces itself. By the time a technology is genuinely useful at scale, the loud conversation about it has usually already moved on to the next thing. That gap between the two clocks is the entire curve.

The four phases

Every AI release I've tracked moves through some version of the same four phases. Knowing which one you're looking at changes what questions are worth asking.

• 01

  Early days

  Real capability is quietly building. Almost nobody outside a narrow circle is talking about it yet, which is exactly why it's worth watching closely if you find it.

• 02

  Peak hype

  Expectation runs well ahead of what the technology can actually do. This is where most public commentary lives — loud, confident, and mostly extrapolating from a demo rather than from deployed use.

• 03

  The comedown

  Reality sets in, limitations surface, and the same voices that called it revolutionary now call it overblown. Both takes are usually reactions to the hype curve, not to the technology itself.

• 04

  Real value, compounding

  This is where the durable capability actually lives — well after the headlines have moved on, quietly embedded in workflows and decisions rather than in demos.

Reading where you actually are

The curve is only useful if you can place a specific release on it. A few questions I ask before I form an opinion on anything new:

Is the commentary I'm seeing describing a demo, or a deployment? Is the loudest reaction coming from people building with it daily, or people watching from the outside? And critically — is my own reaction proportional to what's actually shipped, or to how confidently it was announced? Most miscalibration comes from answering that last question honestly too late.

Why this matters for tech, AI, and investing

This isn't just a media-literacy exercise. If you're making decisions — what to build, what to invest in, what to adopt inside a business — the cost of misreading the curve is real. Committing resources during peak hype means overpaying for capability that isn't there yet. Dismissing something during the comedown means missing the compounding phase entirely, right as it starts to matter. The curve doesn't tell you what to do. It tells you which conversation you're actually in before you decide.