Objection I
Isn’t this just effective altruism with a flag on it?
Partly, and we’d put it the other way around. The habits effective altruism is known for — counting lives, comparing options, writing down your reasoning so someone can check it — are not an import. Franklin ran the arithmetic on his own decisions and called it moral algebra. Madison designed a government on the assumption that the people running it would be unreliable. The Federalist Papers are eighty-five essays of crucial considerations, published under a pseudonym, arguing with the objections in advance.
What we take from effective altruism is the arithmetic. What we take from America is the answer to what the arithmetic is for. Neither one is complete alone: a movement that counts carefully but has no country ends up accountable to nobody, and a country that stops counting ends up spending its strength badly.
Objection II
Are you anti-China?
No, and we think the framing is the problem. The question that matters is not who reaches powerful AI first. It is whether whoever reaches it can control the thing they have built.
An uncontrolled system built in America is not an American victory. It is a loss that happens to have an American return address. The same is true in Beijing, which is why their engineers have an interest in this that their government’s rhetoric doesn’t capture. There is no version of this race where one side wins by being the first to lose control.
So: compete, absolutely. Compete on chips, on energy, on talent, on the willingness to fund hard things. The one thing not worth competing on is who can skip safety work fastest, because the prize for winning that race is the same for everyone.
Objection III
So you want to slow down AI.
We want it built the way Americans built the Apollo program: fast and paranoid at the same time. Those flights ran to a public deadline, and they also ran on triple-redundant systems, abort modes rehearsed until they were boring, and engineers with standing to stop a launch. Nobody called that slowing down. They called it the reason people came home.
Every serious American engineering achievement has been fast and careful together. The tradeoff between speed and safety is real at the margin and wildly overstated in the middle, and it is usually invoked by people who want to skip the part where somebody checks.
Objection IV
What do you actually want? Name something.
Six things, all of which follow from the Articles rather than from anyone’s party:
- Publish the numbers. Any federal program justified by lives saved should state how many, by what method, with what uncertainty. This is already how the Office of Management and Budget is supposed to work.
- Evaluation capacity for frontier systems. The government should be able to test the most capable AI systems for dangerous capabilities before deployment, with the standing and the technical staff to mean it.
- Treat biosecurity as national security. Washington did. Pandemic preparedness belongs in the same budget conversation as missiles.
- An off switch with an owner. Any system deployed in critical infrastructure should have a named person who can stop it, and the authority to do so without asking the vendor.
- Count foreign lives at a non-zero rate. Not equal to American lives in a defense budget — we understand the job of a government. But not zero, either. PEPFAR is the best thing the United States did this century and it cost less than a rounding error.
- Fund the boring prevention. The interventions that save the most lives per dollar are almost always unglamorous and almost always underfunded, in foreign aid and at home both.
Objection V
Effective altruism blew up. Why borrow the brand?
A movement about honest accounting had a fraud at the center of it, and pretending otherwise would be the same failure twice. What FTX proved is that stated principles don’t constrain anyone by themselves — that you need structures that hold when the person inside them is ambitious, wrong, or lying.
Which is Madison’s entire point, and the reason this document is shaped like a founding document rather than a manifesto. If men were angels, no government would be necessary. Nobody involved is an angel. Build accordingly.
Objection VI
Fine. What does an Effective American actually do?
Gives a real fraction of their income to whatever does the most good, and can tell you why it was that and not something else. Chooses work by how much it matters rather than how it sounds at a reunion. Argues about the numbers in public, where the arguing is the point.
And declines, politely and permanently, the demand to choose between loving the country and doing the arithmetic.
Who we are
Nobody official.
This is not an organization. There is no staff, no budget, and nothing to join. It began as one sentence posted by a government account and one person who thought the sentence had the wrong conjunction in it.
Everything asserted on the front page is sourced, including the parts that are contested, because a document arguing for counting things should be checkable. If something here is wrong, it gets corrected in public.