My top 5 priorities as AI Czar

I posted my AI Czar platform on X and tagged the President and the Vice President, with a cc to the War Room. It did well. A tweet is a list of positions with the reasoning cut off, so here is the long version, humbly, for anyone in Washington who is taking applications.

The premise under all five: AI is the most strategically important technology this country has touched since the bomb, and we are running it like a startup perk program. If we believe our own rhetoric about a race with China, the policy has to look like we are in one.

1. Day one: fire all H-1B visa holders from Anthropic, OpenAI, and SpaceXAI

This is a national security risk, and we already have plenty of qualified workers.

Think about what actually lives inside a frontier lab. Model weights. Training recipes. Unreleased eval results that tell you exactly what the next system can and cannot do. Every serious government on earth would love a copy of any one of those. We guard far less important things with security clearances and background checks, and then we staff the most sensitive rooms in American technology through a visa program built to fill jobs more cheaply.

That is the other half of it. The H-1B ties a worker's legal right to stay in the country to one employer. That gives the company enormous leverage over the worker and gives the worker every reason not to make trouble. It is a labor arbitrage tool that got rebranded as a talent pipeline. I have written about what that looks like from inside a big company, and I know a lot of very capable American engineers who are out of work right now. The "we can't find anyone" line does not survive contact with the unemployment line.

O-1 holders may stay. For now. That visa is for demonstrated extraordinary ability, which is a narrower bar and a different conversation. "For now" means it is under review too.

2. Charge any American slowing down AI progress with treason

If this is a race, and the labs and Washington both say it is, then deliberately slowing the American side down helps the other side. There is a word for helping the enemy. The Constitution uses it: treason consists of "adhering to their Enemies, giving them Aid and Comfort."

People want to treat obstruction as a neutral lifestyle choice. It isn't neutral. When you sabotage the American effort in a two-horse race, you have picked a horse.

3. Easement money and free utilities for data center towns

When a data center goes into your town, you absorb the construction, the load on the grid, the water, the noise, and the land. You get a press release about "jobs" and "tax base." That deal stinks, and it is why so many of these projects turn into fights at the county commission.

We already know how to handle this. When a pipeline or a power line crosses your land, you get paid for the easement. Do the same thing here. If a data center goes in your city, you get free utilities and an annual stipend for as long as it runs. The day it shuts off, so does the check.

Now your neighbors want the data center. The incentive runs the right direction, the buildout gets faster, and the people carrying the cost actually share in the upside. That is how you build infrastructure in a democracy without having to lie to anybody.

4. Direct the CIA to subvert Chinese AI labs at the highest level

I assume we already have assets placed there. If we don't, that is a CIA leadership problem, and it needs a leadership change.

China is not being polite about our labs. We should stop being polite about theirs.

5. Monthly meetings with the CEO, technical leadership, and 5 random engineers

Every month, at each company, I meet with three groups: the CEO, the technical leadership, and five software engineers picked at random. Separately. The goal is to gauge real progress and make sure nobody is sandbagging.

Separate rooms matter because each group tells a different story. The CEO gives you the roadmap and the vibe. Technical leadership gives you the risks they are comfortable saying out loud. Five randomly chosen engineers tell you what is actually happening, because nobody had time to prep them. Random matters because a company will absolutely stage-manage a hand-picked engineer panel.

The truth is usually in the gap between those three stories. If the CEO says the next model is on track, the leads say "mostly," and the engineers are telling you they have been blocked on the same data pipeline for six weeks, you have found your problem.

I have been in engineering leadership for years, including as an Engineering Director at American Express. I know when something STINKS about progress on a technical project, and I know exactly how to get it back on track. This is pretty much my only skill in life. I would like to spend it on something that matters.

TLDR

Crack some skulls and make America technologically superior again, because this is what I was born to do.

Humbly for your consideration, @POTUS and @VP.

• Dr. J