Open Citizen ScienceOpen Origins

Theory scoreboard

Each origin-of-life theory, the prediction it can least afford to get wrong, and the evidence from 2023-2026.

TheoryRiskiest predictionEvidence so farStatus
RNA worldSmall polymerase ribozymes are common in sequence space and can run a closed copying cycle.For: QT45 (Gianni, Holliger et al., Science 2026), a 45-nucleotide ribozyme, made a copy of itself and of its complementary strand from triplet building blocks in mildly alkaline ice, at about 0.2% yield in 72 days. A larger polymerase ribozyme replicated short double-stranded RNAs exponentially, including a fragment of itself, under pH and freeze-thaw cycles (Attwater et al., Nature Chemistry 2025). Against: the triplets are pre-activated and supplied by the experimenters, and no ribozyme has yet replicated its whole self exponentially.Partial
Alkaline vents / metabolism firstLane, RussellNatural pH gradients across iron-sulfur barriers drive CO2 reduction well past formate.Formate depends on the pH gradient (Hudson et al., PNAS 2020); micromolar formic and acetic acid across iron-sulfide walls (Altair et al., JACS 2025). Nothing beyond two-carbon products yet.Strained
Warm ponds / wet-dry cyclingDeamer, DamerRepeated drying makes long, copyable polymers faster than water breaks them.For: wet-dry cycling joins cyclic nucleotides of all four bases into chains without added activators, up to about 70% yield for G (Caimi et al., ACS Central Science 2025). Against: the chains are short (0.1% reach 10 letters) as growth and breakdown balance, and water from one Icelandic hot spring destroyed about 90% of 10-letter RNAs in four cycles (Šimonis et al., Astrobiology 2026).Partial
Cyanosulfidic chemistrySutherlandOne UV-lit surface setting supplies nucleotide, amino acid and lipid precursors without incompatible steps.Strong step by step: precursors of nucleotides, amino acids and lipids from one chemistry (Patel et al. 2015). On our pathway map its sugars come out trapped as oxazolidinones, so the first hand-off breaks, and 8 of its 9 hand-offs are unexplained.Strained
Peptide-RNA co-evolutionPowner and othersAmino acids attach to RNA in a selective way without enzymes, a first step toward coding.Thioester-driven aminoacylation of RNA in water, favoured over random peptide bonding (Singh, Powner et al., Nature 2025). Coding itself is unexplained.Partial
Autocatalytic sets (RAF)Hordijk, Steel, XavierSelf-sustaining reaction sets in real metabolism are more than an artefact of network size.A maxRAF of 1,335 reactions, a quarter of an ancient anaerobic network, once cofactors are in the food set (Xavier et al., Proc R Soc B 2020). Never tested against randomised networks.Untested
Assembly theoryCronin, WalkerMolecules with assembly index above about 15, in many copies, occur only with life.Mineral units score up to 21 (Hazen et al., J R Soc Interface 2024); assembly index argued to be a compression measure (Uthamacumaran et al., npj Syst Biol Appl 2024; Abrahão et al., PLOS Complex Systems 2024; Ozelim et al., npj Complexity 2026). Proponents reply that Hazen's values are theoretical (Walker et al., J R Soc Interface 2024) and that the index differs from compression (Kempes et al., npj Complexity 2025).Contested
HomochiralityA physical process biases one handedness strongly enough to be amplified.Bennu's measurable chiral amino acids are racemic, like Ryugu's, which counts against delivery from space (Glavin et al., Nature Astronomy 2025). A magnetite spin route (Ozturk et al., Science Advances 2023) gives about 60% excess at best and about 35% on average after recrystallisation; above about 25%, one more crystallisation gives pure crystals.Strained
Lipid worldLancet, Segré, DeamerCompartments can grow, divide and pass on their make-up before any genetic polymer exists.Peptide sequences change how fast fatty-acid vesicles grow (Baba et al., Communications Chemistry 2026); bilayers form from scratch without a parent membrane, but with enzymes (Khanal et al., Nature Communications 2026). Inheritance of composition is not yet shown.Untested

Status comes from the published literature as of November 2026. None of it has been re-tested by us yet; each will be updated as our own benched results land.

Published claims we will re-test against chance