More Perfect

Author: Christian Keil URL: https://www.moreperfect.tech/ Format: a 346-slide visual essay, published as a web app — no article body, no narration track. Self-described as “The State of the Union, as written by a technologist.” One-line: A data-dense, two-part report arguing that the last 250 years of American technology have quietly delivered on the Constitution’s preamble, and that the next 250 depend on energy and intelligence — closing on the claim that the union can become more perfect only if people make it so.

What it is, structurally

This is not an article with pictures. It is a deck built to be read, organised as a formal report and navigated slide by slide, each slide carrying a single claim in a header bar with a sourced chart or image beneath it and a citation line in the footer. The opening frame — “This is More Perfect. A report on how technology shaped — and will reshape — the American union” — sets the conceit, and the “Inaugural Edition (2026)” label suggests it is intended to recur.

Part one is organised by the preamble to the Constitution. Keil reproduces the preamble early and then uses its clauses as section headers, asking of each whether America is living up to it. The running sections are:

• General Welfare — infant mortality, avoidable deaths split into preventable and treatable, poverty rates, housing demand composition.
• Common Defense — cultural power (English-language web share; 18 of Europe’s top 20 films in 2024 American-made; streaming share of global TV), missile production over fifteen years, Chinese manufacturing value-added against the US, new home construction. One slide is a joke that lands the section’s argument: “There’s nowhere to run. We have only colonized one of the 6,000 confirmed exoplanets.”
• Justice & Domestic Tranquility — violent crime, recidivism (one in three prisoners back within three years), rule of law, community, and trust in institutions: local newspaper closures, confidence in the military, confidence in the president.
• The Blessings of Liberty — the section that turns. “Americans are unhappy,” and only a quarter believe their children will be better off economically than they were.

That structural choice is the report’s best idea. It refuses to grade the country on vibes or on GDP, and instead grades it against the document’s own stated objectives — which produces a genuinely mixed verdict rather than a boosterish one, since the material clauses do well and the tranquility and liberty clauses do not.

Part two — “Technology in 2276” — asks what the next 250 years hold. It opens with a Yogi Berra epigraph on the difficulty of prediction and then works through the physical stack:

• Energy, framed as a personal budget: think of daily electricity as 35 kWh/day and “buy” what you want, itemised (EV driving 15.0, air conditioning 7.5, refrigeration 2.5, laundry, lighting, devices). Coal has fallen ~5.4%/yr since its 2005 peak. Oil is “a more complicated story,” with horizontal drilling and fracking taking America from feared shortages to the world’s largest producer in about a decade, and a careful explanation that “proved reserves” (17.7 tcm) is a term of art distinct from “unproved” (+66.8 tcm). Nuclear gets the sharpest chart — global reactor construction starts since 1951, with China’s bars dwarfing everyone else’s since 2010 — plus the Department of Energy’s Reactor Pilot Program (Aalo, Antares, Deep Fission, Last Energy, Natura Resources, Oklo, Radiant, Terrestrial Energy, Valar Atomics) and an argument that on a 250-year horizon fusion is hard to discount. Solar is still a tiny share of the grid, with a Berkeley Lab OpEx chart; a memorable slide computes that covering the entire contiguous United States in panels would yield ~380 TW, or 300x current consumption.
• Intelligence, via METR’s task-horizon curve — the length of task an AI can complete has doubled roughly every seven months for six years — plus the generalist claim that “feeding tons of data and compute to a general purpose algorithm is better than trying to specialize,” illustrated with Waymo’s EMMA driving on a language model and Physical Intelligence operating robots on one. Then the buildout: xAI’s Colossus 2 as the world’s first 1 GW data center; Epoch AI’s analysis of whether a ~100M-H100 ramp is feasible, with Keil noting he finds their 20%-of-capacity assumption pessimistic. The section’s epigraph is Niels Bohr to Edward Teller on the Manhattan Project — that it could not be done without turning the whole country into a factory, and that they had done just that.
• “Life in the Year 2276” — today/tomorrow pairs. Water: treatment consumes 2% of global energy today, and cheap electricity allows direct potable reuse tomorrow. Labor: the marginal cost of physical labor is a person’s wage today, and collapses toward the price of electricity tomorrow as robots proliferate.

It closes with three framing questions — what can companies do that individuals cannot; what if every physical object were better at its job; what if every individual had every degree and knew every trade — and answers the third with “Talent arbitrage is over,” arguing that the vertical SaaS boom was a bet on Stanford CS grads knowing software better than the incumbents in industry X, and that all such arbitrage trends to zero in a highly-intelligent world.

The final slide is the thesis: “I am fundamentally optimistic about the future. Our union can become more perfect — if smart, driven people make it so.”

The argument, stated plainly

Strip the charts and More Perfect makes one claim in two halves. First, that the American record against its own founding objectives is far better than the national conversation admits, and that the gap between the material facts and the felt mood is itself the thing to explain. Second, that the next 250 years are governed by two inputs — abundant energy and abundant intelligence — and that their arrival makes many current economic positions (specialist labor, talent arbitrage, point-solution software) worth much less than they look.

The conditional in the last sentence is load-bearing, and Keil states it explicitly in the companion essay: he does not believe things automatically get better, only that they can.

Archived material

Key slides captured locally, plus the deck’s structure and the full slide manifest: ../attachments/more-perfect/more-perfect.md

Connections

• Christian Keil — the author. He invests in startups that promote the national interest as a partner at Andreessen Horowitz, and writes as @pronounced_kyle.
• What I Would Tell the Founders About American Technology — the companion piece, and they should be read as a pair. The Arena essay is the argument in prose at a few thousand words; More Perfect is the same argument as a 346-slide evidence base. The essay says so directly, pointing readers to “my report, More Perfect.”
• Historical Futurism — the method exactly: use the 250-year record to reason about the next 250, rather than forecasting from the present alone.
• Progress Studies — the intellectual neighbourhood, and More Perfect is one of the most complete data artifacts the movement has produced.
• American Dynamism — the investment thesis the report is arguing for, without naming it.
• Nuclear Energy — the reactor-construction-starts chart and the DOE Reactor Pilot Program slate are the report’s strongest single section.
• Energy — framed as the binding constraint on everything in part two, using the 35 kWh/day personal budget as the unit of account.
• Compute and Deep Learning — the METR task-horizon curve and the generalist-beats-specialist claim.
• Robotics — the labor today/tomorrow pair: the marginal cost of physical labor collapsing toward the price of electricity.