One to Zero: Every “Innovation” Is Just Math

Every startup deck promises “revolutionary innovation.” The story is always creation — value out of nothing. It is a flattering story, and it is wrong. Innovation does not create from nothing. It compresses: the same outcome, from radically less.

Peter Thiel framed the prize as zero to one — make something that did not exist. The thing that actually wins is one to zero: take what an outcome costs and drive it toward nothing.

The function

A 10× improvement is f(x) = 10. A great one is 100. And every innovation law you have heard of is this one ratio wearing a costume: Wright’s Law is f(x) growing with cumulative production, Moore’s Law is f(x) doubling with time, the experience curve, network effects, even the S-curve of adoption — all views of the same number. Measure the compression ratio and you have measured the innovation.

The zero is a floor, not a wall

The denominator never actually reaches zero — physics forbids it. Computation itself has a minimum energy cost; every outcome has an irreducible floor. So the real game is not “reach zero.” It is close the distance to the floor.

The purest version I know is in computation. Filter a billion database rows the ordinary way and each row costs ~2,000 bit-operations — string pointers, cache misses, mispredicted branches, electrons commuting from RAM over paths 200× longer than necessary. Do the same filter as a single integer compare and it costs 32 bit-operations per row. Identical answer, ~60× fewer electrons moved, seconds collapsing into milliseconds. Nothing was invented. The waste was deleted.

The fastest computation is the one that moves the fewest electrons.

That sentence generalizes to everything. The best process of any kind is the one that moves the least — atoms, hours, dollars, decisions — while still producing the outcome. “One to zero” is not doing the old thing slightly cheaper. It is asking: what is the physical minimum this outcome could cost, and how much of the gap have we closed? Most systems run at a few percent of their floor. That gap is the entire opportunity.

History was the same function

The printing press (1440) — the classic. Gutenberg is remembered as a genius inventor. What he actually did was apply the function to books. The outcome — one mind’s words in another’s hands — cost a scribe-year, roughly 2,000 hours of copying. The press collapsed it to hours; at print-run scale, something like 20,000× per copy. Nothing about the outcome changed. Its cost fell toward the floor — and books went from aristocratic furniture to everywhere, and literacy, the Reformation, and modern science followed. That is what a compression ratio that size does: it doesn’t improve a market, it rewrites a civilization.

Facebook (2004). Social connection is older than agriculture; Zuckerberg did not invent it. He compressed what maintaining it costs — travel, postage, coordination — to approximately zero. Sharing a photo went from print-and-mail, days and dollars, to seconds and nothing. The outcome was ancient. The denominator collapsed.

Amazon (1994– ). Bezos did not reinvent retail. An unlimited digital shelf, third parties funding the inventory, robots running the warehouse, your own front door as the store — each move deleted a cost that stood between wanting a thing and having it.

And the bigger the ratio, the faster the world switches. The telegraph took 56 years to reach half of us. Personal computers took 16. Smartphones took 5. AI tools took about 3. Adoption speed is not a mystery — it is a function of f(x).

Compression compounds

Independent improvements multiply; they do not add. Ten separate 2× gains are not 20× — they are 2¹⁰ ≈ 1,000×. This is why stacks beat point solutions, and why the current wave is unlike anything before it: AI inference is the steepest compression curve ever measured — costs fell ~50× in three years, a learning rate roughly triple that of semiconductors, the previous record holder.

The paradox at the bottom

Here is the counterintuitive part: as cost approaches the floor, total spending goes up. Cheap outcomes uncover demand that expensive outcomes suppressed — Jevons saw it with coal in 1865, and it is happening now with intelligence: per-token prices fell ~1,000× while total spend on inference exploded. When demand is elastic, total value scales like f(x)². Zero-adjacent costs do not shrink markets. They detonate them.

The honest part

Better math does not guarantee a win. Most breakthroughs were stumbled into, not planned; a 10×-better idea still dies without the capital, skill, or timing to ship it; and soft forces matter — Facebook was wired into a primate instinct long before it was efficient. But the pattern holds where it counts: the businesses that win have better compression by default. When one option delivers the same outcome at a fraction of the cost, it pulls demand the way lower ground pulls water. The math is not destiny. It is gravity.

What to do with this

Stop hunting for “innovation.” Pick the outcome your business actually sells, estimate its floor, and measure your distance from it honestly. Then delete waste until the gap closes. If you don’t, the function is patient — someone else will, and the market will flow downhill to meet them.