Put sugar, salt, water and a few gases in a pot, heat it, run current through it, keep at it as long as you like: it will not come alive. Yet on Earth 4-odd billion years ago that is essentially what happened. A pile of completely lifeless chemicals, somewhere, turned into something that copies itself.
The hard part is not "simple becomes complex". It is the deadlock inside every living cell today: the instructions sit in DNA, but reading them and making a copy takes a set of molecular machines called proteins — and how to build those machines is written only in the instructions. Each side presupposes the other. Put either one in the pot alone and nothing turns.
So the real question is not which end to start from, but where the loop was first cut open. And the uncomfortable part: whatever cut it should, by that same logic, be long gone.
In the 1960s several people hit on the same way out almost at once: if some molecule could both record information in the order of its parts and fold into a shape that does work, the loop need not be a loop — it could hold itself up.
RNA is exactly that kind of molecule. It is a close relative of DNA, likewise a chain of small parts whose order is itself information; but it is usually single-stranded rather than wound taut around a partner, so it can fold up, and the folded shape can grip other molecules and speed some reaction up enormously. Storing information and doing work do not conflict in it.
That stayed a guess until someone caught an RNA actually doing work: in 1982 Cech found a stretch of RNA that cuts a segment out of itself and joins the ends back together with no protein involved, and he and Altman took the 1989 Nobel Prize in Chemistry for it. Such catalytic RNAs later got a name: ribozymes. Evidence closer to a fossil arrived in 2000. The ribosome is the machine that builds proteins in every living cell — bacteria, mushrooms and you all use it, which says it is extremely old — and when its structure was solved that year, people could see that the spot where two amino acids are actually joined is ringed entirely by RNA. The core operation of the whole protein production line is done by RNA. The name Gilbert gave this picture in 1986 has stuck: the RNA world.
This line is nowhere near finished, and two holes are wide open. First, the parts: how do RNA's small components accumulate in an environment with no life in it? In 2009 Sutherland's group published a route that goes around the textbook way of taking the problem apart and made two of them — which proves it can be done, not that this is how it happened on Earth. Second, replication: nobody has yet made a ribozyme that copies itself in full. The best result amplifies short templates more than ten-thousandfold, but stalls the moment it meets a complex folded structure — and it is itself exactly such a structure.
The objection is blunt: RNA is far too intricate for step one — a long chain that stores information needs ready energy and parts concentrated enough to react, and in a dilute broth everything diffuses away. So the other camp holds that what turned over first was a set of chemical reactions feeding each other. The product of A is exactly the feedstock of B, and B's product feeds back into A; as long as supply keeps coming from outside, the contents of the cycle keep building up. It needs no instructions. The chemistry closes on itself.
Where, then? The black smokers found in the deep sea in 1977 run at hundreds of degrees and are acidic — spectacular, but too fierce. What is genuinely useful to this camp is the Lost City field found on the Mid-Atlantic Ridge in 2000: far cooler, alkaline fluid, mineral towers like frozen foam, riddled inside with pores tens of micrometres across.
Why that shape matters takes a detour through the present. Every living cell on Earth makes energy the same way: it pumps protons (what is left of a hydrogen atom after it loses its electron) to one side of a membrane, then lets them leak back through a narrow channel, and the leak drives a molecular turbine that turns out the currency the cell spends directly. (Mitchell proposed this in 1961; it later won a Nobel Prize in Chemistry.) And the wall of one of those pores has that same difference across it for free: alkaline fluid inside the pore, few protons; the acidic early ocean outside, many. Across one thin wall sits a proton gradient that costs nothing and does not stop for tens of thousands of years. Russell and Martin argue from this that cells did not invent that way of making energy — they borrowed it first, and only later learned to build a membrane of their own and carry it around.
Metabolism-first has its own hardest obstacle elsewhere: without heritable variation there is no selection. However well a chemical cycle sustains itself, the fact that it happens to run a little better cannot be passed to the "next" cycle — of the three conditions set out in the issue on the logic of natural selection, it is missing the most crucial one. So the two camps are less opposed than they once were: the argument is about where the handover happens, not about who is entirely right. And a third line is filling in another piece — molecules like fatty acids wrap themselves into small bubbles in water, and once you enclose things, dilution stops being fatal.
In 1953 the graduate student Miller built a glass apparatus in his supervisor Urey's laboratory: water boiling below, a few simple gases above, two electrodes sparking across the gap to stand in for lightning. A week later he drew off the water and analysed it — there were amino acids in it, the parts proteins are made of. It was the first time anyone had made "the material of life" from scratch in a flask.
But that classic experiment has to be read in reverse today: the atmosphere Miller assumed is now thought to be wrong, since early Earth's air is now mostly reckoned to have been carbon dioxide and nitrogen, under which the same apparatus yields far less. Then again, when Bada's group re-analysed samples Miller had sealed away and never published, the volcanic-plume version of the apparatus turned out to have given the richer set of products — an unsuitable atmosphere overall does not mean an unsuitable patch next to a volcano. The parts are genuinely not rare either: the Murchison meteorite that fell in Australia in 1969 was found to contain dozens of amino acids.
But the parts were never the hard part of this problem. What such experiments establish is that a route works, not that it is the route that was taken. And the gap cannot be closed, for two independent reasons. First, chemistry leaves no fossils — the earliest things had no bones, no shells, no shape at all; they do not press a print into rock the way a trilobite does. Second, and worse: in a system that holds itself up, the earliest version has no reason to persist once it has produced something better. Its erasure is not an accident. It is the normal outcome.
The time window keeps tightening. Earth is about 4.54 billion years old; the least contested traces of life are stromatolites in Western Australia from around 3.48 billion years ago — rock texture laid down layer by layer by microbial mats — and even those have been argued to be the work of non-biological processes. Meanwhile a molecular-clock estimate published in 2024 puts the common ancestor of all living things at roughly 4.2 billion years ago, and gives it a genome comparable to a modern bacterium's. That date will go on being tested, but if it holds, the time left for "chemistry to cell" is frighteningly narrow.
One last tempting question also has to be left hanging. All known life shares one genetic code, and uses only the left-handed version of its amino acids (the same molecule can come in two mirror-image versions, like left and right hands, and life uses only one of them). This reads equally well as "it only ever succeeded once" and as "it succeeded many times and only one lineage survived", and the two readings cannot currently be told apart. The most honest summary: we have several routes that each make sense, and not one of them yet runs from chemistry all the way to a cell.