The Bureau of Labor Statistics is working hard. The surveys are rigorous. The methodology is sound. The data is precise.
It's measuring the wrong thing.
Every week another report lands: AI is killing jobs, or creating them, or leaving productivity flat, or causing it to surge. Economists disagree. Analysts disagree. The numbers don't resolve because the numbers can't see what's actually happening.
What's actually happening is this: things are becoming possible that weren't possible before. Not faster. Not cheaper. Possible. And the measurement infrastructure we built to track the old economy has no instrument for "possible."
The Ghost Economy
Here's the specific blindspot. When a company lays off 500 people because AI handles their workload, that shows up. The jobs were real, they disappeared, the statistics register the loss.
What doesn't show up: the companies that never needed those 500 people to begin with. The startups that exist now — building products, serving customers, generating revenue — that couldn't have existed five years ago because they would have required 30 people and could only afford 3.
Those are ghost jobs. Jobs that never needed to be created because AI handled the work from day one. They're invisible to every labor statistic. No one was hired, no one was fired, nothing changed in any dataset. And yet an entire category of company is now viable that wasn't before.
I run multiple AI-powered products. When investors see my revenue-per-employee numbers, they don't believe them — not because they doubt the accounting, but because the ratios break their mental models. The comparables don't exist. The benchmarks were calibrated on a different economy.
That's not a problem with my numbers. That's a problem with the benchmarks.
What Multiplier Actually Means
When people talk about AI productivity gains, they talk in percentages. Ten percent faster. Thirty percent more output. The conversations are about improvement at the margin.
That's not what's happening at the leading edge.
AI multiplies. Not improves — multiplies. There's a qualitative difference between "I can do this job faster" and "I can now do jobs that were previously impossible for me to do." The first is increment. The second is phase change.
I can't afford to do what I'm doing without AI. That's not a productivity statement. It's an existence statement. The products I'm building, at the pace I'm building them, with the scope they have — none of that is possible without AI as a multiplier. The economy I'm operating in is not the economy the BLS is measuring.
This is the opportunity that gets buried under the job displacement narrative. Every disruptive wave creates chaos. The factories close. The jobs change. The metrics go haywire. And in that chaos — every single time — there's a window for the dreamer, the small operator, the person with the idea but not the army.
The internet did this. Suddenly a teenager with a connection could reach a global audience. Suddenly a two-person shop could compete for customers with a thousand-person company. The chaos was real. So was the window.
The Measurement Lag Is the Tell
The reason economists are struggling to measure AI's impact isn't that AI isn't doing anything. It's that the impact is concentrated in exactly the places existing instruments can't see.
GDP captures transactions. It doesn't capture capability. Employment statistics capture job counts. They don't capture what jobs make possible. Productivity metrics assume consistent outputs and measure throughput. They can't capture the shift in what's even worth attempting.
The dreamer who couldn't afford to build a product suite before and now can — that's not a productivity gain. It's a new entrant into the economy who didn't previously exist. No survey reaches them before they're large enough to matter. No baseline captures what would have happened otherwise.
This is why the macro data looks muddled. The gains aren't distributed evenly across the existing economy. They're concentrated in new formations — small, fast-moving, not yet visible in the aggregate data — that are building the economy that will be measured in ten years.
The Window Is Open
Disruptive waves don't wait for the measurement infrastructure to catch up. The window opens, things become possible that weren't before, and the people who see it move.
The people who don't see it wait for the data to tell them it's real. The data always lags. By the time the BLS can measure what AI has done to the economy, the early-mover advantage will have already compounded into something durable.
The chaos is the point. In an established equilibrium, scale wins. The large incumbent has the resources, the distribution, the brand. Disruption scrambles the equilibrium. In the scramble, what matters is speed, adaptability, and the ability to multiply your output with the tools available.
You can't compete with a large company on resources. You can compete on what's possible per person.
The data doesn't know you exist yet. That's not a problem. That's the condition.
— J.P. Howlett