Reality Check
And remember: this is a back-of-the-envelope calculation, built on numbers that don't match reality with any precision. The aim was never to nail down an exact figure — just to get some sense of the scale of what evolution actually faces. And even that is only the tip of the iceberg.
If you've made it this far, you've done the hard part. What's left is short, and it only makes the picture clearer. Let's stop being quite so charitable, and see what happens when the numbers get closer to what's actually been published.
Start with the change per species. This chapter assumed one protein's worth for each step from one species to the next. We've already seen that even the most cautious count gives more than a hundred — and that other ways of counting give many times that.
Not all of that change necessarily codes for proteins — much of it plausibly does other work instead: regulation, structural changes, the wiring of a developing brain, and any number of similar jobs. But there's no good reason to think that other kind of genetic information needs any less precision, on average, than the instructions for a protein. Some of it surely needs less; some, especially anything shaping the brain, likely needs more.
None of this necessarily generalizes to every lineage in the history of life — it's entirely possible that getting from an ancestral species to something as different as a modern human took more change than the average step elsewhere on the tree of life. But treating "one protein's worth of change per species" as a typical figure, the way this chapter did, now looks less like a fair estimate and more like extreme generosity.
The total species count was generous in the same direction. Six hundred million was described, deliberately, as a conspicuously low figure — and it was. Estimates of the real total over the same six hundred million years run into the tens of billions. Take just ten billion, more than sixteen times our figure, and combine it with even a hundred proteins' worth of change per species, and the pace at which life would have needed to generate brand-new, working, protein-sized genetic information stops looking like "about one a year." It comes to nearly two thousand a year — and on the largest count, tens of thousands. (The arithmetic: Human and Chimpanzee.) That demands a considerably better accounting than anything offered so far.
Moreover, every calculation in this chapter was built around a single protein size: three hundred letters, chosen for convenience, so we'd have one clean number to work with. But we already saw in the protein data that the real average runs somewhat higher, that a hundred amino acids is closer to a floor than a ceiling for anything that has to fold into a working machine, and that proteins many times longer than three hundred are common. What we haven't said yet is that the largest and most complex machines of all aren't oddities sitting at the margins — they're essential to life itself. The Krebs cycle and the ribosome, both of which later chapters look at directly, depend on machines assembled from many separate chains, some of them far longer than three hundred letters. Once the full range of what life actually needs is taken into account, the numbers here don't just get bigger. They get bigger by a margin that makes three hundred look almost quaint.
And there's another, equally important way in which this whole exercise was generous. No living thing produces a brand-new protein every time it reproduces. Real mutation works far more modestly than that: mostly point mutations, each changing a single letter of DNA, scattered across the genome, and only rarely landing in the small stretch that codes for any one particular protein. Counted realistically, this widens the gap between what's needed and what's available considerably further still.
Put it all together, and what life has actually had available to explore, across its entire history on Earth, is a minuscule fraction of everything that's possible. Not "minuscule" in the ordinary sense of one in a hundred, or even one in a million. The whole search space of three-hundred-letter protein chains takes three hundred ninety-one drums to count; all the tries in all of life's history, on our most generous count, take forty-five. The share that could ever have been tried is one chain in a number three hundred forty-six drums long. Picking out one particular atom from among all the atoms in the visible universe is an eighty-drum lottery. Our lottery is like picking the right atom four times in a row, each time from a fresh universe — and then some.
Of course, answers have been offered. But as far as I can tell, few of them tackle the specific issues raised by these numbers head-on — and most of them are aimed at explaining how life could improve once it already existed, not how it could have gotten started in the first place. These answers include things like:
- It isn't random, because natural selection does most or all of the real work.
- There's never any need to hit a working sequence directly — evolution proceeds one letter at a time, with each small step tested and selected on its own, so the search space never has to be searched all at once.
- Modular tools let evolution cut corners entirely — existing chunks can be stitched together into new proteins without ever searching the broader search space at all.
And there are more along these lines. The chapters that follow deal with these responses, one at a time, always keeping an eye on the actual arithmetic involved rather than settling for the plausibility of the idea alone. And through all of it, one question will stay squarely on the table, because none of these answers actually touch it: how did any of this begin in the first place?
Last revised 2026-11-17