---
title: "Its Own Metaphor"
url: https://raskell.io/articles/review-how-life-works/
author: Raffael Schneider
date: 2026-11-14
description: "Philip Ball's How Life Works is the popular science book I wish every other one were: deep, clear, and honest about how we came to know what we know."
tags: [life-sciences, genetics, emergence, complex-systems, philosophy-of-science, synthetic-biology]
categories: [reviews]
review_of: "How Life Works"
review_of_author: "Philip Ball"
review_of_kind: book
review_of_year: 2023
rating: 9/10
---

# Its Own Metaphor

## In short

Philip Ball's *How Life Works* is a tour of the new biology, from genes to bodies to medicine, and the best popular science book I have read; I give it 9 out of 10. Its argument is that the genome is not a blueprint and life is not a machine or a computer: genes are resources the cell draws on, and every level of life has its own rules. It explains the science deeply enough to get a real grip on it, and it tells how each piece of knowledge was won, by whom, and with which technique. Ball keeps describing life as if it were engineered, then attributes everything to evolution; I read the same evidence the other way. Some of its arguments against the computer metaphor aim at an old picture of computing, and its take on AlphaFold has aged, but none of that changes the recommendation.

On June 26, 2000, Bill Clinton announced the first draft of the human genome and said that we were learning "the language in which God created life." Philip Ball opens *How Life Works* with that sentence and answers it in the next line: Clinton was wrong, but not (just) in the way you might think. The problem, Ball says, was never the reference to God. It was the idea that the genome is an instruction book at all.

I read the book in April 2025 and only now realised it was missing from [my reading log](https://raskell.io/reading/). It belongs near the top. It is the best popular science book I have read, and the one I would hold up as the model for all the others. I give it 9 out of 10.

## What the book argues

The book is a user's guide in the literal sense: it walks through life one level at a time, from genes and RNA to proteins, networks, cells, tissues, and bodies, then to agency, medicine, and the engineering of new living things. At each level Ball shows a picture most of us learned at school, explains why the research of the last few decades has broken it, and describes what is taking its place.

The thread that ties the levels together is a single claim. The genome is not a blueprint, a program, or a code that an organism executes. Genes, in his phrase, are "not instructions, but resources": a toolkit the cell draws on, switched on and off by committees of molecules, read differently in every context. The decisions that make a body are taken higher up, by cells and tissues that sense, signal, and settle into a small number of stable states. No level is in charge, and each has rules that do not depend on the details of the level below.

| Level | The picture we learned | Ball's picture |
|---|---|---|
| Genes | A blueprint; a gene for each trait | Resources the cell draws on; most variants matter only in combination |
| RNA | A messenger carrying the gene's instructions | Where much of the regulation happens; a large part of the genome is read for control, not for proteins |
| Proteins | Machines with a fixed shape: lock and key | Many are partly disordered and take their role from context and partners |
| Networks | Circuits of switches | Regulation "by committee": loose, combinatorial, noisy, and robust because of it |
| Cells | Factories following orders | Agents that decide their fate together, settling into stable states |
| Tissues and bodies | Built to a plan | Generated by rules: chemical gradients, Turing patterns, mechanical forces |

Ball's conclusion is that living things "have to be their own metaphor." No technology we have built works the way a cell does. The closest analogy he finds, in the epilogue, is language: something that works on many levels at once and loses its meaning when you reduce it to its words.

## How it is told

What makes the book great is the writing. Ball is curious, generous, and a little flashy, in the best sense: he enjoys his material and wants you to enjoy it too. He explains how microbiology works in plain language without flattening it, so that you come away with a real grip on what is happening inside a cell rather than a pleasant feeling of having been told. When the science gets complicated he does not hide it; he shows you why it is complicated and what the complication is for.

He also tells how we came to know what we think we know. Each chapter is a history of a technique and the people who used it: Conrad Waddington drawing cell fates as balls rolling down a landscape of valleys, Alan Turing working out how spots and stripes can arise from two chemicals, Shinya Yamanaka turning adult cells back into stem cells with four factors, single-cell sequencing that finally let researchers watch the valleys Waddington imagined. He includes the false starts, the reversals, and the researchers who took months to believe their own data. It is science shown as an iterative, human process, which is how it actually works, and it is far more convincing than a list of results.

Some of the episodes stay with you. Cut a flatworm into a hundred pieces and each grows into a whole worm. Five fingers is not written anywhere in the genome; it is what a Turing-type pattern happens to fit into a limb bud, and the same mechanism, tuned differently, makes fins. The side of the body your heart ends up on is decided, early in the embryo, by tiny hairs stirring fluid in one direction.

## Designed, and then not

There is one thread through the book that Ball never resolves, and I think it is the most interesting thing in it.

Ball describes life, again and again, in the language of design. Its mechanisms are "ingenious." Combinatorial signalling and causal emergence are "design principles." Bodies form through rules that are designed to produce something generative rather than a fixed product. In chapter 9 he says it outright: all these processes behave as if serving some overall plan, and "Biology looks uncannily teleological," that is, as if directed toward a goal. He then takes on William Paley's argument from 1802 that contrivances like the eye point to a designer.

And every time, he steps back. The appearance of a plan is real, he argues, but it comes from evolution working with self-organisation, physics that gives structure "for free," and the agency of organisms themselves. In the epilogue he even asks whether that agency might give evolution something like a direction, and hurries to add that this is "not a backdoor for intelligent design." In the prologue he offers readers who believe in God the consolation that God comes out of the new biology looking smarter than the Human Genome Project implied, and invites everyone else simply to be enchanted by the genius of life.

I read the same evidence the other way. I believe life was engineered, though not necessarily in a theistic sense: I do not need the designer to be a god for the design to be real. What struck me is how much of Ball's own material fits that reading better than the one he chooses. A small, shared toolkit that produces coherent bodies from loose rules. Generative systems that correct their own errors. A cell computing close to the physical limit of efficiency. Layers of control that delegate decisions upward, the way a good engineer builds a system to adapt rather than to follow a script. Ball spends four hundred pages explaining that life is not a machine because it is far cleverer than any machine we have made. That is a strange argument against design.

To be fair to him: he is a science writer explaining the consensus of his field, and he says plainly which questions remain open. The book is honest about the limits of what is known, and it does not pretend the design question is settled by the data. It simply declines to ask it. I would have liked one chapter that did.

## What a builder takes from it

For someone who builds systems, the book is full of ideas worth stealing:

- **Noise as a resource.** Cells do not fight molecular randomness so much as use it to make decisions, explore options, and keep populations diverse.
- **Structure for free.** Cells let physics do work they never specify: phase separation, surface tension, buckling. The equivalent in software is using what the platform already provides.
- **Generative, not prescriptive.** A body is not assembled from a plan but grown toward a target shape by rules that correct themselves. A cut zebrafish tail regrows, stripes included.
- **Prediction on a budget.** Ball reports an estimate that a cell computes within about ten times the theoretical minimum energy, while our computers use around a million times more. Life, as he puts it, takes care not to overthink the problem of survival. It fits my conviction that [intelligence is compression](https://raskell.io/notes/intelligence-is-compression/): an organism stores only what helps it predict.

Two places where I would push back. First, the computer Ball argues against is an old one: the tape reader, the fixed circuit, the program as a list of steps. Modern distributed systems are noisy, fault-tolerant, and full of emergent behaviour, and a program never builds anything without the runtime it runs on. "Source plus runtime" survives his argument; only "source alone" fails. Second, his verdict on AlphaFold, that it "does not so much solve" the protein-folding problem "as sidestep it", undersells what it does. Predicting the result without simulating the path is exactly what makes it useful. The book was finished before the Nobel Prize for that work, which I wrote about in my review of [*The Infinity Machine*](https://raskell.io/articles/win-everything-control-no-one/).

## Medicine and making

The last chapters turn the argument into practice, and they are where the book is most hopeful. If disease is a state of the whole body rather than a broken part, medicine should work at the level where the cause sits, and aim, as Ball says, less at curing disease than at curating health. He is sceptical that genetic screening or gene therapy will help much with common diseases, which involve hundreds of variants each, and hopeful about immunotherapy, about nudging cancer cells back to a benign state, and about the body's own small number of ways of failing and recovering.

Then come the things that sound like science fiction and are not: organoids, synthetic embryo models with beating hearts, and xenobots, tiny living machines made from frog cells, designed by an evolutionary algorithm in a computer, and then assembled by hand. Ball's view is that we will engineer life not by writing genes but by steering how cells organise themselves. Whichever way that turns out, it confirms what I believe about the atom, the bit, and the gene: the life sciences are at the very beginning of their revolution.

## Who should read it

Anyone who wants to understand how living things work as we know them now, not as the textbooks of thirty years ago taught it. You need no background in biology, but you will need attention: it is a long book, and dense in places, and it rewards being read slowly.

A note on how this review came about: I am writing it a year and a half after I finished the book, so the details are not as fresh as they would be for something I read last week. I went back to the text to check every quotation and example above, but the judgement is the one I carried away in 2025, and it has not faded. What prompted me to write it now is the book I am reading at the moment, Walter Isaacson's *The Code Breaker*, and the contrast is instructive. Isaacson tells the story of gene editing through the people who invented it; Ball explains why the gene was never the whole story. A few chapters in, Isaacson feels slow and a little patronising next to Ball, who trusts his reader with the real complexity and makes it a pleasure. That trust is what I would wish for every popular science book.