Prebiotic Chemical Networks

Speculative If the sequence from geochemical feedstock to a mixture that copies itself with errors can be run end to end in a single apparatus — which no laboratory has done, and section 7 states exactly what would count — the output is a measured probability that chemistry becomes biology, and that number is the largest missing input to every estimate of how much life there is anywhere. The condition comes first because it has not been met: every published route is a chain of stages assembled by a chemist between steps.

Frontier The magnitude is best stated as the size of the free parameter this would close. Every forecast of inhabited planets multiplies a well-constrained astronomical term — star counts, planet occurrence rates, habitable-zone geometry, all measured to within a factor of a few — by a chemical term whose allowed range spans the entire interval from “once per galaxy” to “on every wet rock”. Established The one empirical constraint on that term is a date: the 2024 molecular-clock analysis puts the last universal common ancestor at 4.2 billion years ago, with a 95% interval of 4.09 to 4.33 Ga, against a Moon-forming impact at about 4.51 Ga — a window of roughly 300 million years from sterilising impact to a cell with a 2.5 to 3.0 megabase genome and about 2,600 proteins. Speculative That is one data point, drawn from the only biosphere we can sample, and a single fast origin is weak evidence that origins are fast.

Speculative A second, smaller prize is a capability rather than a number: a chemical mixture that improves at a task under selection without carrying a genome. This already exists in a narrow form. In October 2025 a Groningen group reported peptide-based self-replicators in a flow set-up in which mutants were selected for their ability to activate a photocatalytic cofactor — selection on a functional trait rather than on replication speed alone. Handwave Whether that scales from a designed building block to a mixture nobody designed is unknown, and the gap between the two is the gap this entire brief is about.

Frontier Who collects it is unusually lopsided. There is no product, no patent estate and no clinical endpoint at the end of this work; the gains accrue to astrobiology mission design, to exoplanet observing-time allocation, and to a small research community funded from philanthropic endowments and exploratory national lines rather than by a mission agency with a schedule. Handwave The condition on all of it: the upside is real only if “prebiotically plausible” is a predicate that can be failed, and section 5 argues that at present it is not.

The question is not how to make a nucleotide. That was answered, repeatedly, by different routes, and the answer turned out not to settle anything. The question is how a mixture of small molecules driven by an energy source acquires heredity — the capacity to hold a variation and pass it on so that selection has something to act on. A prebiotic chemical network is the unit of study that replaced the single molecule once the field noticed that no single molecule is the answer.

Established Three research programmes divide the subject, and they disagree about what came first rather than about any measurement. Genetics-first holds that a replicating informational polymer came first and metabolism was built around it; its flagship is the cyanosulfidic chemistry associated with John Sutherland’s group at the MRC Laboratory of Molecular Biology. Metabolism-first holds that a self-sustaining network of carbon-fixing reactions came first, driven by geochemical gradients, with genetics a later overlay; its flagship is the alkaline-hydrothermal-vent programme associated with William Martin, Nick Lane and Joseph Moran. Compartment-first holds that the decisive step was enclosure, because without a boundary nothing that a network produces stays with the network that produced it. Frontier All three have working chemistry. None has a demonstration that reaches heredity.

Frontier What changed in the last decade is that the field stopped arguing about molecules and started measuring networks. The formose reaction in a continuous stirred-tank reactor shifts its product distribution systematically with feedstock concentration, calcium and hydroxide levels, temperature and residence time — the environment selects the pathway, with no genetic information present. Speculative A prebiotic system does not need to find one correct reaction; it needs an environment that repeatedly favours one region of a network over the alternatives. Handwave That also makes the problem harder to falsify, because a network that responds to its environment can be reported as a success under almost any conditions.

Speculative The framing this page tests is that prebiotic chemistry has become an experimental science with a measurable endpoint, and it survives with one large qualification. The chemistry is real, the reactors are real, and the individual steps have measured yields. The qualification is that no route has a published end-to-end carbon yield, no stage sequence has been run without a chemist intervening between stages, and the field has no agreed statement of what a negative result would look like. Handwave A programme that cannot say what would count as failure is not yet a programme, however good its individual papers are.

Established Start with the anchor this brief was commissioned on, because checking it produced the first finding. The commission describes “Sutherland’s cyanosulfidic route: activated pyrimidine ribonucleotides, amino acid and lipid precursors from HCN, H₂S and UV, measured yields, independently repeated”. Established That sentence compresses two different papers with two different feedstock sets into one route, and the compression matters. The 2009 Nature paper (459, 239–242, 14 May 2009) assembles activated pyrimidine ribonucleotides from cyanamide, cyanoacetylene, glycolaldehyde, glyceraldehyde and inorganic phosphate, through arabinose amino-oxazoline and anhydronucleoside intermediates, bypassing free ribose and the free nucleobases. Established Hydrogen sulfide appears in the other paper. The 2015 Nature Chemistry paper (7, 301–307, 16 March 2015) derives precursors of ribonucleotides, amino acids and lipids by reductive homologation of hydrogen cyanide and its derivatives, driven by ultraviolet light, using hydrogen sulfide as reductant and accelerated by Cu(I)–Cu(II) photoredox cycling. Frontier The 2015 work produces precursors, not the nucleotides; the 2009 work produces the nucleotides, without H₂S. Read as one continuous route, the anchor claims more than either paper does.

Frontier The second half of the anchor — “measured yields, independently repeated” — is the part that did not verify. Neither flagship abstract states a yield; the numbers live in supplementary information, step by step, and no end-to-end carbon yield for the full sequence has been published by anyone. Speculative Nor could this session find an independent replication in the ordinary sense — a second laboratory running the published procedure and reporting what it got. What the field has instead is competing syntheses: Thomas Carell’s group published a unified route to pyrimidine and purine ribonucleotides in Science in 2019, and Sutherland’s own group published a selective route giving RNA pyrimidine and DNA purine nucleosides together in Nature 582, 60–66 (3 June 2020). Frontier Alternative routes to the same target are evidence that the target is reachable. They are not replication, and the difference is exactly the difference a ring trial exists to measure.

Established The feedstocks, at least, are not in doubt, and the strongest evidence for that arrived from an asteroid. Analysis of samples returned from (101955) Bennu, published in Nature Astronomy in January 2025, reports 4.5 to 4.7 wt% total carbon and 0.23 to 0.25 wt% nitrogen; ammonia at 13.6 µmol g⁻¹ in hot-water extract, about 12 times the Murchison meteorite and 75 times Ryugu; 33 amino acids including 14 of the 20 used in terrestrial protein synthesis, with glycine most abundant at 44 nmol g⁻¹; and all five canonical nucleobases plus about 18 further nitrogen heterocycles totalling roughly 5 nmol g⁻¹. Established The chiral non-protein amino acids are racemic or nearly so. Frontier Two consequences follow, and the second is underreported. Monomer availability is not the bottleneck; a pristine, never-terrestrially-contaminated rock carries the building blocks. And the standard explanation for biological homochirality — that impactors delivered an enantiomeric excess — loses its best available sample.

Established The clock on the other end is now tight. A 2024 Nature Ecology & Evolution analysis places the last universal common ancestor at 4.2 Ga (95% interval 4.09–4.33), with an inferred genome of at least 2.5 Mb (2.49–2.99 Mb) encoding about 2,600 proteins, a near-complete Wood–Ljungdahl pathway, and an anaerobic acetogen lifestyle on hydrogen and carbon dioxide. Frontier The authors state the implication plainly: the interval between the Moon-forming impact and a fully modern prokaryote-grade cell was surprisingly short. Speculative What it does not show is that the origin was easy, because LUCA is not first life — it is the last common ancestor of surviving life, separated from the origin by an unknown amount of evolution.

Frontier Replication chemistry produced two genuinely new results in 2025 and 2026, and they are the most important numbers on this page. In July 2025 Attwater and colleagues in Philipp Holliger’s group reported in Nature Chemistry that a triplet polymerase ribozyme supplied with trinucleotide substrates under pH and freeze–thaw cycling achieves open-ended exponential RNA replication: the trinucleotides act both as substrates and as chaperones that keep the product duplex from reannealing, which is the strand-separation problem that had blocked the route for two decades. Established The measured performance is 21 to 73 iterative cycles, per-cycle growth of 1.2 to 1.38-fold for defined sequences, up to 31-fold amplification of individual strands and 60,000-fold for emergent sequences across more than 40 cycles with serial dilution, on duplexes of 30 nucleotides and pools up to 60. Frontier The ribozyme does not replicate itself. The paper reports only fragmentary self-replication of a 29-nucleotide segment and states that the intact polymerase is an inefficient template and largely resistant to invasion.

Established Then, in February 2026, the self-replication half moved. Science published QT45, a 45-nucleotide polymerase ribozyme found in random-sequence pools, which catalyses general RNA-templated synthesis from triplet substrates in mildly alkaline eutectic ice. It synthesises its own complementary strand from a random triplet pool at 94.1% per-nucleotide fidelity, and a copy of itself from defined substrates — both at yields of about 0.2% in 72 days. Frontier The caveats are the authors’ own and they are severe. The two syntheses run as separate reactions rather than in one pot, plus-strand synthesis needs a pre-formed hexamer in addition to triplets, and the reactions ran at −7 °C over 28 and 33 days. Frontier David Lilley of Dundee called the rate “unbelievably slow” while calling the result a good step; the first author stated that because the two reactions are not sequential and not in one pot, “we are not yet at the stage of self-replication”. Speculative Ninety-four per cent fidelity is also below the error threshold a 45-nucleotide genome needs to be maintained indefinitely, which is why the result is a milestone rather than a replicator.

Established Non-enzymatic copying — replication with no ribozyme at all — has a measured and brutal bias problem, and 2024 put a number on the fix. Copying random-sequence RNA templates with activated mononucleotides is overwhelmingly skewed to G and C: averaged across product positions, adenine appears at frequency 0.068 and uracil at 0.028, and the CCC trimer appears 854 times more often than UAA. Established A Nucleic Acids Research paper (52, 13515, 2024) combining a random 2- to 16-mer oligonucleotide pool, methyl isocyanide in situ activation and eutectic ice cycling brought the median trimer frequency to the ideal 1/64 = 0.016, with an error frequency of 0.063 and about 23% of primers extended. Frontier That is uniform sequence coverage at a 6% error rate, which is the honest state of genome-free copying: no longer biased, still far too inaccurate to hold information.

Established Compartments are the one subsystem where the engineering is close to routine. A December 2024 Journal of the American Chemical Society paper from Job Boekhoven’s group at TU Munich showed fatty-acid vesicles — decanoic acid at 65 mM, pH 4.9 to 6.8, driven by a carbodiimide chemical fuel across 0.4 to 65 mM — budding off daughter vesicles under fuel-concentration control while retaining encapsulated DNA. Established A 2025 preprint characterised mixed single-chain amphiphile coacervates of nonanoic acid, nonanol and tyramine that are stable from pH 7 to 9, 5 to 65 °C, up to 600 mM NaCl and 20 mM MgCl₂, persist for 8 days, sequester RNA and support non-enzymatic template-directed primer extension inside the droplet. Frontier The stated limit is the one that matters: stability falls away above 20 mM magnesium, and ribozyme catalysis generally wants more. The compartment and the chemistry it is supposed to contain want different ionic conditions, which is a concrete instance of the incompatibility problem in section 5.

Frontier On the metabolism-first side, the most substantial recent result is a reconstruction rather than a synthesis. A Science Advances paper of 5 August 2026 by Natalia Mrnjavac and colleagues in Martin’s group reports that LUCA had enzymes for only about half of the roughly 420 reactions of core metabolism, the rest carried by environmental metal catalysts. Established The same work identifies phosphite, which occurs in serpentinising hydrothermal systems, as an energy source that phosphorylates AMP to ADP and serine to phosphoserine over native metal catalysts (Fe⁰, Co⁰, Ni⁰, Pd⁰) in water. Speculative The authors go further and argue that bacteria and archaea became free-living independently — one origin of the genetic code, two origins of cellular life. Handwave That rests on the two domains using structurally distinct enzymes for the same reactions, which has a duller explanation: independent replacement after divergence.

Frontier Heredity in a messy mixture — the thing all of this is for — has been attempted once at the level of a controlled serial-transfer experiment, and the result is weak. A 2024 preprint ran a defined food set (formic acid, acetic acid, methanol, sodium bicarbonate, ammonium hydroxide, sodium trimetaphosphate) over pyrite for 24 hours at 40 °C, transferred 20% into fresh medium, and repeated for 10 generations against no-transfer controls, analysing composition by liquid chromatography–mass spectrometry with ecological distance metrics. Frontier Parent and offspring vials were more chemically similar to each other than to other vials in several generations. Speculative The authors describe the evidence as weak and the experimental variability as high, which is the correct reading and also the most valuable sentence in the paper. Handwave One under-powered experiment is the entire measured literature on chemical heritability in an undirected mixture.

Established Two further measured results belong on the record because they are routinely over-read. Carell’s 2022 Nature paper (605, 279–284) showed that two non-canonical nucleosides still present in modern tRNA will couple amino acids onto RNA strands at 40 to 77% yield depending on the amino acid, with 10 to 15% for the final urea-cleavage step — a bridge from RNA to peptides with no ribosome. Established And in March 2025 a Stanford group reported in Science Advances that sprays of water microdroplets into a nitrogen–methane–carbon-dioxide–ammonia mixture produce hydrogen cyanide, glycine and uracil through microscale discharges between oppositely charged droplets, with no applied electricity at all. Speculative Both are real chemistry and neither is a step toward heredity.

Speculative And the record of what has never been shown should be stated as flatly as the record of what has. No mixture of prebiotically plausible molecules has ever been observed to replicate with heritable variation and be selected on that variation. No route from geochemical feedstock to any replicator has been run without a chemist changing the conditions between stages. Speculative No published prebiotic route reports an end-to-end carbon yield. Handwave Every claim that the origin of life is “nearly solved” is a claim about the first of those three, restated as though it were the third.

Frontier The live positions here are unusually well separated, because they disagree about the order of events rather than about any measurement. They are set out as positions with holders.

Speculative Position one: genetics first, and the cyanosulfidic route is the working demonstration. On this reading the hard problem is an informational polymer, everything else is downstream, and the evidence is a chain of syntheses that reach activated ribonucleotides, amino-acid and lipid precursors and even DNA purine nucleosides from one-carbon feedstocks under ultraviolet light. Established The chemistry is real and has been extended for seventeen years by the group that started it. Handwave The weakness is structural and this page states it: a route assembled by a chemist who chooses which fraction to carry forward at each stage is a demonstration that the molecules are reachable, not that a planet reaches them.

Speculative Position two: metabolism first, in an alkaline hydrothermal vent, with genetics a late overlay. The case is that LUCA’s inferred physiology — acetogenic, Wood–Ljungdahl, hydrogen-and-carbon-dioxide — is precisely the chemistry a serpentinising vent supplies for free, and the 2026 phosphite result gives the programme an energy currency that is not ATP. Frontier It also has the only continuous, geologically realistic, self-refreshing reactor in the argument, which is an advantage the surface-pond scenarios do not have. Speculative Its standing objection is nearly twenty years old and has never been answered on its own terms. Leslie Orgel argued in PLOS Biology in 2008 that the barrier to self-organising metabolic cycles is not catalytic efficiency but catalytic specificity: a reverse citric acid cycle needs at least six distinct specific activities, and a non-specific mineral catalyst that reduces oxaloacetate will also reduce pyruvate and disrupt the cycle. Handwave His conclusion, that appeals to “if pigs could fly” hypothetical chemistry are unlikely to help, is quoted approvingly by people arguing against abiogenesis entirely, which is a reason to answer it rather than to avoid it.

Speculative Position three: compartment first, because a network without a boundary does not have products, it has a diluted ocean. This is the position with the best recent engineering — fuel-driven vesicle budding with content retention, coacervates stable for days that support primer extension inside them — and the least developed account of where the information comes from. Frontier Its sharpest internal problem is quantitative and already visible: the magnesium concentration a coacervate tolerates and the concentration a ribozyme wants do not overlap.

Frontier Position four is not about order at all, and it is the one gaining ground: the unit of selection before life was the network, not the molecule. The 2022 formose work in a continuous stirred-tank reactor showed the product distribution reorganising systematically as feedstock, catalyst, temperature and residence time change — selection among pathways with no genetic material anywhere in the system. Speculative The same group’s 2024 Nature paper ran the formose reaction as a physical reservoir computer: up to 106 distinct ions tracked by ion-mobility mass spectrometry, eight hours of continuous operation, and nonlinear classification and forecasting performed by the chemistry itself. Handwave That a reaction mixture can be trained to forecast a Lorenz attractor tells you the network is expressive; it tells you nothing about whether it is heritable, and the slide from one to the other is the commonest move in this literature.

Speculative Position five: two origins, not one. The 2026 reconstruction argues that bacteria and archaea independently completed the transition from geochemical dependence to autonomy, sharing a genetic code but not a cellular origin. Frontier If it holds it changes the arithmetic of this entire subject, because an event that happened twice on one planet within a few hundred million years is a different kind of event from one that happened once. Handwave It rests on inference from enzyme structural distinctness, and structural distinctness has a second explanation — non-orthologous displacement after divergence — that the argument has to exclude and has not yet excluded.

Frontier Position six is methodological and it is the deepest disagreement in the field: nobody agrees what “prebiotically plausible” means. The phrase is used to mean, variously: uses only molecules detected in meteorites; uses no purified reagents; runs at plausible concentrations; runs without an operator; runs under a specific Hadean atmosphere. Speculative These are not the same criterion, and a synthesis can satisfy one while failing the others — which is how two groups describe the same experiment as a demonstration and as a curiosity without either misrepresenting it. Handwave Until the predicate is defined, nothing here can be definitively refuted, and a field in which nothing can be refuted accumulates papers rather than knowledge.

Speculative Position seven: exogenous delivery did the hard part. Bennu is a strong result for this view. Frontier What it is not is an answer, because the monomers were never the scarce thing and delivery relocates the chemistry rather than performing it. Handwave The strong version, that life itself arrived, moves the problem somewhere it cannot be worked on, and this page treats that as a change of subject.

Handwave Position eight, held quietly and rarely written down: the problem is not solvable with the evidence available, because the record was destroyed. On this reading the Hadean rock record cannot constrain surface cyanide, ground-level ultraviolet flux or phosphate availability tightly enough to rule anything out, and a laboratory demonstration establishes only possibility. Frontier The counter is methodological rather than rhetorical: a route that runs end to end without intervention would be strong evidence regardless of what the Hadean was like, which is why section 7 ranks that experiment first.

Frontier The binding bottleneck is not a reaction. It is that no route in this subject has a published end-to-end yield, so the composite probability of any complete pathway is unknown to everyone including its authors. Individual steps report yields — 40 to 77% for RNA–peptide coupling, 10 to 15% for the cleavage after it, about 23% of primers extended in optimised non-enzymatic copying, 0.2% for QT45 in 72 days. Speculative Multiply a dozen such steps and the answer is a number nobody has written down. Handwave A field that reports step yields and never the product of them is reporting the numerator of a fraction whose denominator would embarrass it.

Frontier The second is condition incompatibility between stages, and it is the reason the operator is still in the loop. The known routes want ultraviolet light at one stage and darkness at another, hydrogen sulfide here and its absence there, freezing in eutectic ice for polymerisation and liquid water for hydrolysis-driven steps, high magnesium for ribozymes and low magnesium for compartment stability. Established The coacervate result puts a number on one instance: droplets destabilise above 20 mM MgCl₂. Speculative Every stage transition is currently performed by a person, and the person is doing selection work that the geology is supposed to do.

Speculative The third is the error threshold, and it is now quantitative enough to be uncomfortable. QT45 copies at 94.1% per-nucleotide fidelity; optimised non-enzymatic copying has an error frequency of 0.063. Frontier Both are around 6% per base. A 45-nucleotide sequence copied at that rate retains its exact sequence in roughly 6% of products; a 100-nucleotide ribozyme retains it in about 0.2%. Speculative Information cannot be maintained across generations at these rates without either a shorter genome than any known ribozyme or a fidelity improvement nobody has demonstrated — the classical Eigen error-threshold trap, now with measured numbers on both sides.

Frontier The fourth is analytical, and it caps what can be claimed about any network. The most information-dense prebiotic network yet published tracked 106 distinct ions. Speculative That is an achievement and a ceiling: an untargeted mixture contains thousands of species, most below detection, and the tar — the inert polymeric fraction that sequesters carbon and is the oldest objection to the whole subject — is precisely what mass spectrometry sees worst. Handwave “The network produced X” and “the network produced X at a level our instrument can see” are different statements and the literature does not always distinguish them.

Speculative The fifth is that heredity in an undirected mixture has one under-powered experiment behind it. Ten generations, 20% transfer, weak parent–offspring correlation, high variance, called weak by its own authors. Handwave The property that converts chemistry into biology has less experimental support than any individual synthesis in section 3.

Handwave The sixth is not technical: there is no agreed failure condition, so the field cannot lose an argument. Because “prebiotically plausible” is undefined, a negative result is always attributable to the wrong conditions, and no journal in the subject publishes systematic negative results. Speculative This is solvable, it costs almost nothing, and section 10 records writing the convention as a formal requirement.

Established This subject waits on capabilities rather than on results other briefs produce, and most of the capabilities belong to analytical chemistry and automation. Continuous-flow reactors with online untargeted mass spectrometry, the instrument class behind every network result in section 3, come from process chemistry and from the automated-laboratory programmes assessed in Cloud Laboratories and Autonomous Materials Discovery. Frontier Neither of those communities is trying to originate life, and their instruments are worth more to this subject for it, because a platform built to search reaction space indifferently does not have a favourite hypothesis.

Frontier One dependency is geological and nothing on this map supplies it: a constraint on Hadean surface conditions tight enough to exclude a scenario. Cyanide concentration in surface water, ultraviolet flux at the ground, phosphate availability and atmospheric redox state all sit inside ranges wide enough to accommodate every competing programme simultaneously. Speculative Until that changes, “the early Earth could have provided this” is a statement about the width of an error bar rather than about the early Earth.

Speculative A second dependency runs through molecular phylogenetics, and it has been delivering. The LUCA date and genome reconstruction, and the 2026 claim that half of core metabolism was mineral-catalysed at that point, are inferences from sequence databases rather than from chemistry, and they now constrain the chemistry more sharply than most experiments do. Frontier The dependency is fragile in a specific way: these are model-dependent molecular-clock estimates, and the 4.09–4.33 Ga interval is a confidence interval on a model, not a measurement on a rock.

Speculative Traffic with the rest of this map runs almost entirely outward. Biosignature Standards inherits whatever this subject concludes about which molecular assemblages are achievable without biology, and inherits it directly, since a false-positive biosignature is by definition a prebiotic network. Speculative Synthetic Biology and Xenobiology consume nothing from here and would be unaffected by any result on this page. Handwave Nothing on this map blocks this brief, and that is a fact about the subject rather than an omission: it is upstream of everything and downstream of nothing.

Frontier Every position in section 4 has a stated test, and they rank by what a result would settle rather than by how hard they are.

Speculative Tier one, and the decisive experiment for this brief: a continuous-flow apparatus that runs from geochemical feedstocks to a self-replicating, mutating protocell population with no operator intervention between stages. The commission named this test and it is the right one, but the research found it is already half delivered in pieces, and naming which half is the useful contribution. Established Delivered: continuous-flow prebiotic networks whose pathway selection is set by the environment rather than by an operator; open-ended exponential RNA replication across 21 to 73 cycles with 60,000-fold amplification of emergent sequences under pH and freeze–thaw cycling; fuel-driven vesicle division with content retention; primer extension inside a coacervate stable for eight days; and Darwinian selection of self-replicators on a functional trait in a flow set-up. Speculative Not delivered: any two of those in the same vessel. Handwave The experiment is therefore not a new idea but an integration, and integrations are exactly what a field organised around individual laboratories’ signature chemistries does not fund.

Speculative Tier two is a number, not an apparatus, and it is the cheapest high-value result in the subject: publish an end-to-end carbon yield and a carbon budget for one complete route. Take the cyanosulfidic sequence, run it as published, and report what fraction of input carbon arrives as the target and what fraction arrives as tar. Frontier No such figure exists for any route. Speculative The reason it does not exist is that it would be small, and a field that published it would have to argue about what small means — which is the argument worth having.

Frontier Tier three is an interlaboratory ring trial on one published synthesis, and its absence is the finding that most surprised this research. One procedure, one set of reagents, pre-registered, eight laboratories, reporting yields and failures. Speculative The commission’s anchor asserted that the flagship route had been independently repeated; this session could not find such a replication, and what the field actually has is competing alternative syntheses. Handwave A ring trial costs a rounding error against a single centre’s annual budget and would settle whether the reported yields are reproducible outside the group that reported them, which is currently unknown.

Speculative Tier four is the heredity experiment at real statistical power. The 2024 pyrite serial-transfer design is correct and under-powered: ten generations, high variance, weak signal, one laboratory. Frontier The powered version — hundreds of parallel lineages, fifty or more transfers, pre-registered distance metrics, no-transfer and no-mineral controls, and an explicit null model for drift — is a straightforward and unglamorous experiment. Speculative It would establish whether heritable variation exists in messy chemistry at all, which is the single most load-bearing unknown in the subject and currently rests on one preprint.

Frontier Tier five is one-pot ribozyme self-replication, and the target is now precisely defined by what QT45 does not do. QT45 makes its complement in one reaction and a copy of itself in another, at 94.1% fidelity and about 0.2% yield over 72 days, with a pre-formed hexamer supplied for the second. Speculative The discriminating experiment is both reactions in the same vessel, with the same substrate pool, producing catalytically active progeny that then catalyse the next round. Handwave Nobody has proposed a route to the fidelity that would need, and an error rate of 6% per base is the obstacle rather than the yield.

Speculative Tier six is geological rather than chemical: a Hadean constraint sharp enough to exclude a scenario. A measured bound on surface cyanide, on ground-level ultraviolet flux, or on dissolved phosphate at 4.3 Ga would eliminate at least one of the three programmes in section 4. Handwave Whether such a constraint is recoverable from a rock record this old is genuinely doubtful, and this tier may be permanently unavailable.

Frontier Tier seven is a negative control the field has never systematically run: an abiotic complexity baseline. The assembly-theory dispute is the visible symptom. Its 2023 Nature paper proposes that high assembly index combined with high copy number indicates selection; a 2024 critique in npj Systems Biology and Applications reports that the index correlates at R = 0.87 with Lempel–Ziv–Welch compression and R = 0.9 with run-length encoding on mass-spectrometry data, offers no classificatory advantage, and produces abiotic false positives. Speculative The experiment underneath the argument is to run large prebiotic networks for a long time and measure how complex they get without selection. Frontier That baseline is what both this subject and Biosignature Standards need, and neither has it.

Speculative Tier eight is the natural experiment, and it is partly running. Sample return from a carbonaceous asteroid has already delivered a prebiotic inventory assembled without biology and without terrestrial contamination; the equivalent from an ocean world would test whether comparable networks assemble in liquid water. Frontier Bennu’s racemic amino acids are the model for what such a result looks like when it goes against the field’s expectations.

Established And the experiments that are not tests of abiogenesis and are regularly reported as though they were. Producing amino acids from a spark, a droplet spray or an asteroid; synthesising a nucleotide by any route; building a vesicle that divides. Speculative Each of these establishes that a component is reachable. None of them addresses heredity, which is the only property whose absence distinguishes chemistry from biology.

Frontier There is no engineering payoff here and this page will not manufacture one. Every requirement below is a requirement for an instrument that would answer a question, which is why section 10 records a funding condition as a formal requirement rather than a complaint.

Speculative The instrument specification for the decisive test, assembled from what the component results actually used rather than gestured at. A multi-stage continuous-flow train: at least three coupled stirred-tank or plug-flow reactors with independently controlled residence time, temperature (spanning at minimum −7 °C eutectic ice to 65 °C), pH, and ultraviolet illumination; automated transfer between stages with no manual fractionation; periodic freeze–thaw and wet–dry cycling as a programmed environmental variable rather than as an operator action; continuous feed of one-carbon and inorganic feedstocks at geochemically defensible concentrations; and online ion-mobility mass spectrometry tracking on the order of 100 or more species simultaneously, which is the current published ceiling. Handwave Each of those exists separately. Nobody has built the train.

Established The half of the specification that is already routine deserves saying, because it is genuinely impressive. Continuous stirred-tank operation of the formose network with systematic variation of formaldehyde, dihydroxyacetone, calcium chloride, sodium hydroxide, temperature and residence time; eight-hour runs with 106 ions resolved; eutectic-ice reactions held for 72 days; fatty-acid vesicle division controlled by fuel concentration across two orders of magnitude. Frontier This is a mature experimental technology attached to an immature question.

Speculative The requirement nobody can specify is the analytical one. Such a run would generate a time series of thousands of unidentified species over weeks, and there is no accepted method for deciding whether it shows selection, drift or nothing. Handwave The ecological distance metrics borrowed for the 2024 heritability experiment are the state of the art, and their authors call their own signal weak; the analysis is as unsolved as the chemistry.

Established The nearest brief on this map is Synthetic Biology, and the two subjects are inverse rather than overlapping. That page assesses whether biology has become an engineering discipline — standardised parts, decoupled design and fabrication, measured first-pass design failure rates — and works forward from known biological components. This page asks how there came to be components at all. Frontier The seam is sharp: everything that brief treats as a given part, this brief treats as an unexplained arrival, and neither result would change the other’s assessment.

Speculative The adjacency that actually inherits something is Biosignature Standards. A false-positive biosignature is, by definition, a prebiotic chemical network doing something that looks biological, so every result about how complex abiotic networks get is directly a result about detection thresholds. Frontier The assembly-index dispute is the live case: a proposed universal complexity threshold for life detection, and a published critique reporting that the measure correlates at R = 0.87 with a 1977 compression algorithm and produces abiotic false positives. Speculative Planetary Protection and Mega-Telescopes sit downstream of the same question, one for contamination control and one for what an atmospheric spectrum can be claimed to show.

Speculative Three further neighbours share instruments rather than subject matter: Cloud Laboratories and Autonomous Materials Discovery build the automated reactor-plus-analytics platforms the decisive experiment needs, and Artificial Scientists is the closest thing on this map to the closed-loop search a large prebiotic network would require. Handwave An automated platform that searches reaction space indifferently is worth more here than a dedicated facility with a favourite hypothesis, and the field has neither.

Frontier Two adjacencies are argumentative rather than technical. Philosophy of Science owns the demarcation machinery this subject needs and does not use: the undefined predicate “prebiotically plausible” is a textbook case of a criterion that cannot be failed. Speculative And Objective Collapse Models, the sibling brief in this category, shares the exact structural problem in a different domain — a theory family whose free parameters can be extended faster than experiment removes them. Handwave Both fields need a pre-registered convention for what counts as defeat, and neither has one. Directed Panspermia and Xenobiology are adjacent in ordinary speech and not otherwise: one is a seeding proposal, the other changes the alphabet rather than explaining it.

Established The institutional fact that explains this field’s shape is that it has no agency. There is no origin-of-life directorate, no line item, no roadmap with milestones and no facility. What exists instead is a small number of endowed centres and a handful of exploratory national grants. Established The Leverhulme Centre for Life in the Universe at Cambridge launched on 10 January 2022 with £10 million over ten years from the Leverhulme Trust, directed by Didier Queloz, spanning nine Cambridge departments plus the MRC Laboratory of Molecular Biology. Established It is a founding member of the Origins Federation together with Harvard’s Origins of Life Initiative, the University of Chicago’s Center for the Origins of Life and ETH Zurich’s Centre for Origin and Prevalence of Life, which met at Chicago on 9–12 September 2025. Frontier Harvard’s initiative reports its support as coming from the university president and provost, the Faculty of Arts and Sciences dean, individual philanthropists and private foundations — which is to say discretionary and personal rather than programmatic.

Speculative That arrangement has a specific consequence: the field’s time horizon is set by donors rather than by the problem. A ten-year endowment is long by grant standards and short against a question open since 1953, and the decisive experiment in section 7 is exactly the integration programme that a centre organised around individual investigators’ signature chemistries is worst at producing. Frontier Nobody is paid to combine another group’s replication chemistry with a third group’s compartment and a fourth group’s reactor.

Frontier The public-money half is exposed in a different way. NASA’s astrobiology programme, which funds much of the American work, spent 2025 and 2026 under explicit budget and workforce pressure, with community advocacy documents warning that reductions in workforce and research funding risk delaying discoveries; the figures were not published in the documents this session could retrieve, and no number is printed here. Speculative A subject whose national-agency support is defended by white papers rather than by a programme of record is a subject whose funding is an annual argument.

Speculative There are three institutional requirements this field lacks, and none of them is money. The first is a definition: a written statement of what “prebiotically plausible” requires, with the criteria separated — feedstock provenance, concentration regime, operator intervention, atmospheric assumption — so a result can satisfy some and fail others explicitly. Speculative The second is a venue for negative results, because a route that does not work is not published and every group rediscovers the same dead ends privately. Handwave The third is a ring-trial convention, and its absence is the most surprising thing this research found: the flagship syntheses have alternatives but no documented independent replication, and nobody appears to consider this a problem.

Frontier The direct hazards here are close to nil and saying so precisely matters, because this subject is routinely assigned risks that belong to its neighbour. Nothing in section 3 produces a self-sustaining organism, a pathogen or a release candidate; the biosecurity apparatus that Synthetic Biology needs — synthesis screening, benchtop-synthesiser oversight, institutional biosafety capacity — has no purchase on a flow reactor making sugars. Speculative The one live containment question is planetary rather than terrestrial: returned samples and the forward-contamination controls covered in Planetary Protection.

Frontier The largest hazard is epistemic, and this field is more exposed to it than almost any other. Origin-of-life results are reported to the public through headlines about creating life in a laboratory, and the underlying papers are usually about a coupling yield. Established The QT45 result is the clean recent example: a 45-nucleotide ribozyme at 0.2% yield over 72 days in two separate reactions, whose own first author said plainly that this is not yet self-replication, reported in the general press as RNA copying itself and solving the field’s decades-old bottleneck. Speculative The authors and the journal were careful; the amplification was not; and the corrective is for the field to state its own numbers in its own abstracts, including the ones that are small.

Speculative Second, and more uncomfortable: the field’s strongest internal critique is quoted most often by people arguing that abiogenesis did not happen. Orgel’s 2008 objection to self-organising metabolic cycles — specificity, not efficiency, is the barrier; a non-specific catalyst disrupts the cycle it is supposed to run — is a technical argument from inside the field that appears verbatim in anti-evolution material. Frontier The correct response is not to soften the critique but to answer it, and seventeen years on the metabolism-first programme has answered it with mineral catalysis rather than with demonstrated specificity. Handwave A field that avoids its own best objection because of who else quotes it has conceded the argument to the wrong people.

Speculative Third, the allocation question, which deserves better than a slogan. This work competes for instrument time and early-career researchers against applied biology with clear endpoints, and its defensible claim is not that the question matters more but that the marginal cost is genuinely small. Handwave That answer stops working the moment someone proposes a dedicated facility for the integration experiment in section 7, and nobody has yet had to make the real argument.

Frontier Fourth, a distributional point that belongs here rather than in section 12. The one policy lever this subject touches is exoplanet observing time: if abiogenesis is inferred to be easy, the case for atmospheric biosignature surveys strengthens; if hard, the same money buys something else. Speculative A field whose inferences move a multi-billion-dollar observational programme should state its uncertainty in units, and it states it in adjectives.

Speculative The concrete consequence of a result here is a number in somebody else’s equation, and it is worth being exact about which number. Estimates of how many inhabited planets exist multiply reasonably well-measured astronomical terms by a probability that chemistry becomes biology, and that term is currently unconstrained across many orders of magnitude. Frontier An end-to-end demonstration would not measure the probability, but it would replace a free parameter with a bounded one, which is the difference between a calculation and an opinion.

Speculative The second consequence is about what we should expect to find, and it cuts both ways with equal force. If the transition requires a rare sequence of environments, the expected number of neighbours falls and the case for treating Earth’s biosphere as irreplaceable strengthens in a way no ethical argument has managed. Speculative If it is generic — and the 4.2 Ga LUCA date, plus the possibility of two independent cellular origins on one planet, is the strongest available hint that it might be — then life is common and the interesting question moves to whatever keeps technological civilisations rare. Handwave Neither inference is licensed by present evidence, and this page declines to prefer one.

Speculative A third consequence is capability rather than knowledge, and it is small but real. The engineering that would come out of a solved origin is not synthetic organisms — Synthetic Biology already builds those from known parts and would gain nothing — but chemical systems that improve at a task under selection without a genome, of which the 2025 flow-reactor selection on photocatalytic function is the existing proof. Handwave Any claim beyond that, including de novo life as an industrial platform, is assertion, and this page does not make one.

Speculative And the outcome that is more likely than either: nothing decisive is found, and the value is the map. A subject that ran for a century and ended with a well-characterised inventory of what abiotic chemistry can and cannot do — which molecules assemble freely, which networks self-organise, how complex a mixture gets without selection — would have produced the baseline that life detection needs, whether or not it explained anything. Frontier That is a real result and an unglamorous one, and it is the most defensible thing to expect.

These horizons track experimental capability and parameter closure rather than any engineering programme, because there is no engineering programme here to track:

• 10 yr: Frontier Expect continuous-flow prebiotic networks to be chained two or three stages deep with automated transfer, and the first published end-to-end carbon yield for some complete route — a number that will be small and whose publication will be the more important event. Speculative Expect ribozyme replication to move from 94% toward 98% fidelity and from 72-day to week-scale timescales, without reaching the error threshold a functional genome needs. Speculative Expect the heritability experiment of section 7 to be run properly by someone, and expect it to be the most cited result of the decade in this field whichever way it comes out.
• 25 yr: Speculative This is the window in which the decisive integration plausibly runs: feedstock to replicating, mutating, compartmentalised population in one apparatus with no intervention. Speculative A positive result would be the first measured statement about how hard abiogenesis is and would immediately reopen every position in section 4. Handwave A null — the apparatus runs and nothing heritable emerges — would be far harder to interpret, because the failure is always attributable to the conditions chosen, which is precisely why the definitional work in section 10 has to happen first.
• 50 yr: Speculative On the observational side this is the horizon on which biosignature surveys either return a population of candidate inhabited worlds or return a clean null, and either outcome constrains this subject more sharply than any laboratory result. Handwave On the theory side the plausible development is consolidation rather than discovery: a settled account of what abiotic networks can reach, functioning as the baseline for life detection.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The defensible statement is about shape: a subject that cannot say what would refute it either acquires that criterion or stops being a science, and this one has had the criterion available and unwritten for seventy years. Handwave Any specific claim about what a solved origin of life would be used for is an assertion, and this page does not make one.

• Depends on No typed dependency edge to another brief on this map, and the absence is informative rather than an omission: this subject is upstream of everything in biology and downstream of nothing on this map. Cloud Laboratories and Autonomous Materials Discovery build the automated reactor-and-analytics platforms the decisive experiment needs, but neither is blocking, because those platforms are being built for materials and pharmaceutical search whether or not anyone points them at the Hadean. Synthetic Biology is the nearest brief by subject and supplies nothing: it builds forward from known biological parts, where this page asks where parts came from.
• Requires (not on this map) Six conditions, in the order of the tokens above. The yield. No prebiotic route has a published end-to-end carbon yield, so the composite probability of any complete pathway is unknown to its own authors; step yields of 40 to 77%, 23% and 0.2% are reported and never multiplied. The reactor. Three or more coupled flow stages with independently controlled residence time, temperature from eutectic ice to 65 °C, pH, ultraviolet illumination and programmed freeze–thaw cycling, with automated inter-stage transfer and online ion-mobility mass spectrometry resolving 100 or more species; every component exists and the train does not. The ring trial. An interlaboratory replication convention for the flagship syntheses, whose absence this research found unexpectedly and which costs a rounding error. The heredity result. Hundreds of parallel lineages over fifty or more serial transfers with pre-registered metrics and a drift null, against which the existing evidence is one ten-generation preprint its own authors call weak. The geological bound. A constraint on Hadean surface cyanide, ultraviolet flux or dissolved phosphate tight enough to exclude one of the three programmes in section 4, which may not be recoverable at all. The money. Decade-scale funding for integrating other groups’ chemistry rather than extending one’s own, which is what the endowment-plus-philanthropy structure is worst at buying.
• Enables A bounded estimate of how readily chemistry becomes biology, which is the free parameter in every forecast of inhabited worlds, and a measured baseline of abiotic complexity, which is what Biosignature Standards needs to set a detection threshold and does not have. Downstream of that, Mega-Telescopes and Planetary Protection inherit a sharper prior. No typed enabling edge is claimed, because a subject that has never demonstrated its central phenomenon enables nothing on a schedule.
• Adjacent Within this map: Synthetic Biology and Xenobiology for the forward-engineering inverse of this question, Biosignature Standards for the detection consequence, Artificial Scientists and Cloud Laboratories for the search apparatus, Philosophy of Science for the demarcation problem this field has and does not use, Directed Panspermia for the seeding literature, and Objective Collapse Models as the sibling case of a theory family that outruns its own falsification. Outside the map: analytical mass spectrometry, systems chemistry, molecular phylogenetics, Hadean geochemistry and asteroid sample curation supply every number on this page.

Established “Miller and Urey made the building blocks of life, so the hard part is done.” The building blocks are the easy part and the strongest proof arrived from space: a sample of asteroid Bennu, curated to exclude terrestrial contamination, contains 33 amino acids, all five canonical nucleobases, and ammonia at twelve times the Murchison level, assembled with no biology and no chemist. Frontier Monomer availability has never been the bottleneck; heredity has.

Established “Meteorites delivered left-handed amino acids, which is why life is homochiral.” The chiral non-protein amino acids in the Bennu samples are racemic or nearly so. Speculative The best-preserved extraterrestrial sample ever returned has no enantiomeric excess to deliver, which removes the most-cited external explanation for homochirality and puts the problem back on Earth.

Frontier “Sutherland synthesised RNA from simple chemicals.” Two different papers with two different feedstock sets are routinely merged into one claim. The 2009 work makes activated pyrimidine ribonucleotides from cyanamide, cyanoacetylene, glycolaldehyde, glyceraldehyde and phosphate; the 2015 cyanosulfidic work makes precursors of ribonucleotides, amino acids and lipids from hydrogen cyanide and hydrogen sulfide under ultraviolet light. Speculative Neither makes RNA, no published sequence runs end to end without a chemist between stages, and no end-to-end yield exists. Frontier This brief’s own commission made the merge, which is how routine it is.

Frontier “The RNA world has been experimentally demonstrated.” As of February 2026 the closest result is QT45: a 45-nucleotide ribozyme that makes its complement at 94.1% per-nucleotide fidelity and a copy of itself from defined substrates, both at about 0.2% yield in 72 days, in two separate reactions rather than one pot. Established Its own first author states that this is not yet self-replication. Speculative The RNA world remains the best-supported hypothesis in the subject and it has not been demonstrated, and those two statements are compatible.

Speculative “It was a warm little pond” / “it was a hydrothermal vent.” The evidence does not discriminate. The vent programme has the better continuous reactor, a physiology that matches the inferred LUCA, and an unanswered specificity objection; the surface programme has the better synthetic chemistry, requires ultraviolet light that a vent cannot supply, and has an unanswered concentration problem. Handwave People who state the answer confidently are stating a preference, and the preference usually tracks their own laboratory’s technique.

Handwave “The probability of life arising by chance is so small it cannot have happened.” Every calculation of this kind computes the probability of assembling a specific modern macromolecule at random, which is not what any origin hypothesis proposes and not what any experiment tests. Speculative The honest statement is the opposite and less comfortable: the probability is not small, it is unknown, because no one has measured the size of the space or the number of acceptable outcomes in it. Frontier That is the missing number section 1 is about.

Frontier “Assembly theory can tell life from non-life.” The 2023 proposal is that high assembly index plus high copy number indicates selection; a 2024 critique reports that on mass-spectrometry data the measure correlates at R = 0.87 with Lempel–Ziv–Welch compression and R = 0.9 with run-length encoding, offers no classificatory advantage, and produces abiotic false positives. Speculative This page takes no side on the theory, but the practical point stands: no complexity threshold for life detection can be validated until somebody measures how complex abiotic networks get, and nobody has.

Speculative “Panspermia solves it” and “LUCA is the first living thing” are the two cheapest errors and they fail the same way. Delivery of monomers is established and unhelpful, because monomers were never scarce, and delivery of life itself relocates the question somewhere it cannot be worked on. Established LUCA is the last common ancestor of surviving life: an anaerobic acetogen with a 2.5 to 3.0 megabase genome and about 2,600 proteins, dated to 4.2 Ga. Speculative An organism with 2,600 proteins is a late product, and the entire subject of this brief lies in the unknown interval before it.

Frontier And one correction this page makes to the summary of this subject still in circulation, including in its own commission. The received statement is that the cyanosulfidic route produces ribonucleotides, amino-acid and lipid precursors from HCN, H₂S and ultraviolet light with measured yields and independent repetition, and that the decisive next step is a continuous flow reactor from geochemical feedstocks to a self-replicating mutating protocell. Established The first half is two papers merged and the replication claim did not verify; what the field has is competing syntheses, not repetitions. Speculative The second half is right and is already half delivered: exponential RNA replication over 21 to 73 cycles, Darwinian selection of replicators on a functional trait in flow, primer extension inside stable coacervates and fuel-driven vesicle division all exist, in four different laboratories and four different vessels. Handwave What nobody has done is put any two of them in the same apparatus, which makes the decisive experiment an integration problem rather than a discovery problem — and integration is the thing this field’s funding structure is worst at buying.