Assembly Theory

Speculative If the assembly index really measures how many steps it took to build a molecule, rather than how compressible its description is, then life detection gains a threshold that assumes nothing about biochemistry — and the instrument that reads it is already flying. The condition comes first because it is the disputed object: whether the index is separable from ordinary compressibility is not settled, and section 7 says what would settle it.

Established The magnitude is best stated as hardware not needed. The proposal is that a single scalar can be estimated from a tandem mass spectrum without identifying the molecule, on the strength of a reported linear relation with a correlation of 0.89 between the number of MS2 fragment peaks and the assembly index, with a threshold near MA about 15 above which the 2021 study found only living samples. Frontier Every other agnostic biosignature with comparable diagnostic ambition needs an instrument that has not flown — the chirality route in Homochirality Origins needs chiral separation with a non-destructive ionisation path, which has never operated on another body. Established A mass spectrometer has flown to Mars, Titan, a comet and Saturn’s plumes. The assembly proposal asks for an analysis, not a payload, and that is the whole of its practical advantage.

Frontier Who collects is a small number of mission payload teams and, indirectly, anyone who wants life detection to stop being a search for terrestrial biochemistry. Speculative There is no product and no market at the end of this; the beneficiary is a decision procedure, and the loser if the threshold is wrong is a flagship mission that reports a false positive it cannot retract. Established The threshold has already been contradicted in print: two abiotic mineral heteropolyanions compute to assembly indices of 18, three above the line.

Handwave The condition, stated plainly: this upside is real only if the index measures construction history rather than repetition, and the literature currently contains a published correlation of 0.87 with Lempel–Ziv–Welch compression and a published proof that computing the index exactly is NP-complete. Those are not obviously claims about the same quantity, which is section 5’s subject and the reason this brief takes neither side.

Assembly theory proposes that an object should be described by the history required to build it rather than by its instantaneous structure, and that this history is measurable. Its two claims are usually run together and have very different evidential status: that molecular complexity can be read off an ordinary laboratory instrument, and that the resulting quantity is a physical measure of selection.

Established The first claim has a definition and a procedure. The assembly index of a molecule is the minimum number of bond-forming steps needed to construct it from basic parts, where any intermediate already made may be reused for free. A molecule with a repeated motif therefore has a lower index than its atom count suggests, and a molecule assembled from many distinct pieces has a higher one. Established The proposal is that this index can be estimated experimentally from the fragmentation pattern of a tandem mass spectrum, and the 2021 paper reports a linear relationship with a correlation of 0.89 between the number of MS2 peaks a fragmented ion produces and its index.

Frontier The second claim is the assembly equation, and it is where the theory stops being a measurement and starts being a theory. Assembly is defined as a sum over an ensemble, A equal to the sum over distinct objects of e raised to the assembly index, weighted by the copy number minus one divided by the total number of objects. Speculative The argument is that high index alone is not evidence of selection, because an unlikely object can happen once; high index together with high copy number is, because repeating an improbable construction requires something that remembers how to do it. Handwave Whether that sum is a law, a definition or a bookkeeping device is exactly what the dispute is about, and no experiment in this brief tests it.

Established The critiques are published, in journals, with numbers. One line of attack says the assembly-pathway method is a member of the Lempel–Ziv family of compression algorithms and therefore adds nothing to information theory; another says abiotic minerals exceed the biological threshold; a third says the index is bounded by Shannon entropy and is a loose upper bound on Kolmogorov complexity. Frontier The theory’s authors have replied in the same journals, and the reply that matters most is a formal one: computing the assembly index exactly is NP-complete, which would put it outside the polynomial-time family it is accused of belonging to.

Speculative The framing this page tests is that assembly theory’s measurement claim and its theoretical claim should be graded separately, and it survives. The measurement claim has a demonstrated instrument path, a hedged threshold and a four-sample blind test. Handwave The theoretical claim has an equation, a dispute about whether that equation is separable from existing complexity measures, and no experiment that would decide it. This brief does not adjudicate the dispute. It states what has been measured, what has been computed, and which of the two the arguing parties are each talking about.

Established Start with the anchor this brief was commissioned on, because checking it produced the sharpest finding in the brief. The commission describes “molecular assembly index measurable by tandem MS; blind test over about 114 samples found a threshold near MA greater than or equal to 15 separating biological from abiotic”. The blind test in the 2021 paper was over four samples, not 114, and the number 114 in that manuscript counts something else entirely.

Established The blind analysis, in the authors’ own description, was “a collection of samples whose origin was unknown at the time of analysis”, prepared by a co-author and revealed only afterwards. It comprised a CM2 carbonaceous chondrite sample of the Murchison meteorite, bay sediment, and biological material from two geological epochs — Holocene material of roughly 30,000 years and Mid-Miocene material of roughly 14 million years. Established That is four samples. Established The figure 114 appears in the accepted manuscript in a different sentence: “To measure this a combination of 1H and 13C NMR experiments were performed on 114 compounds”. Established A companion sentence gives the infrared calibration set: “we generated the IR spectra for 101 different molecules”. Frontier So 114 is a calibration set for nuclear magnetic resonance, and it has migrated in secondary accounts into a claim about how many unknown samples were classified correctly. The distance between those two statements is the distance between a calibration and a test.

Established The threshold is real and the paper states it more cautiously than the citations do. The sentence is: “In the case of the analysis presented here it appears that only living samples produced MA measurement above ~15”. Established The accompanying general claim carries a caveat that is routinely dropped: “Not all molecules produced by living processes have high MA—indeed complex mammals regularly produce CO2—but all high MA molecules are produced by living (or technological) processes”. Frontier “In the case of the analysis presented here” is a claim about a sample set. “All high MA molecules” is a claim about the universe, and the two sit in the same paper.

Established What the 2021 study did measure is worth stating precisely, because it is more than the anchor implies in one respect and less in another. The samples analysed spanned prebiotic soups, E. coli lysates, yeast cultures, urinary peptides, seawater, Taxol, beer, whisky, extracts of quartz, limestone, sandstone and granite, coal, burned yeast, the Murchison meteorite, bay sediment and palaeontological material. Established Three independent techniques were used, not one: tandem mass spectrometry with data-dependent acquisition, proton and carbon-13 nuclear magnetic resonance on the 114-compound set, and infrared spectra computed ab initio for 101 molecules. Frontier The anchor’s “measurable by tandem MS” is correct and understates the paper; its blind-test claim overstates it.

Established The 2023 Nature paper is a different object and is the one the disputes are mostly about. It introduces the assembly equation, A equal to the sum over i of e raised to a_i times (n_i minus 1) divided by N_T, where a_i is the assembly index and n_i the copy number, and argues that this quantity captures the causation required to produce abundant complex objects and therefore quantifies selection. Frontier The paper is largely theoretical: it presents simulations and argument, and points at mass spectrometry as the experimental route rather than reporting new measurements. Speculative Reading the 2021 measurement paper as evidence for the 2023 theory paper is the most common error in this literature, and the two are separated by three years and a different class of claim.

Established The first serious published objection is a set of computed numbers, and it is specific. Hazen, Burns, Cleaves, Downs, Krivovichev and Wong computed molecular assembly indices for mineral heteropolyanions and reported values above the threshold: ewingite at an exact index of 18 and a split-branch index of 20, ilmajokite at 18 and 21, paddlewheelite at 13 and 17, vanarsite at 11 and 16, and fluor-uvite tourmaline at 11 and 14. Established Their conclusion is that “abiotic chemical processes have the potential to form crystal structures of great complexity—values exceeding the proposed abiotic/biotic divide of MA index = 15”. Frontier Two of those are natural minerals with no biology in their formation and an assembly index three above the line.

Frontier The reply is not a retraction and it is not evasion either; it is a claim about what the threshold was ever about. The response, titled to the effect that experimentally measured assembly indices are required to determine the threshold for life, argues that a computed index for a mineral fragment is not the quantity the 2021 experiment measured, which was an index inferred from the fragmentation of a molecule that a mass spectrometer can ionise and break. Speculative That distinction is either the central methodological point of the whole programme or a boundary drawn after the counterexample arrived, and this brief cannot tell which from the published record. Handwave Neither the reply nor the counter-reply could be retrieved this session; both are cited by name, and nothing in this brief rests on their internal argument.

Established The second objection is statistical and comes with correlation coefficients. Uthamacumaran, Abrahão, Kiani and Zenil re-analysed the available data — “99 molecules available in the MS data set” and “18 extracts for which the mass spectra was available” — and report the assembly index correlating with 1D block decomposition at R = 0.8823, with Lempel–Ziv–Welch compression at R = 0.8738, with one-dimensional run-length encoding at R = 0.9, and with Huffman coding at R = 0.896. Frontier Their conclusion is that “the assembly index does not offer substantial improvements over existing methods, including traditional statistical ones”, and that other coding algorithms “systematically outperform the MA index as a discriminant”. Speculative A correlation of 0.87 to 0.90 against four different coding schemes is a strong result and it is not the same thing as an equivalence proof, which is a distinction the surrounding rhetoric has not preserved.

Established The third objection is formal. Abrahão, Hernández-Orozco, Kiani, Tegnér and Zenil argue that the assembly-index calculation method “belongs to the LZ family of compression algorithms”, that it is “bounded by LZ schemes ... and therefore by Entropy”, and that it is “a considerably constrained version of compression algorithms, and a loose upper bound” of Kolmogorov complexity. Frontier They identify three shared elements with Lempel–Ziv: identification of repeated blocks, a dictionary of repetitions and substitutions, and lossless reconstruction. Speculative If that holds, the assembly index is a statistical compressor with a chemistry interpretation, and the claim that it measures causation rather than repetition fails.

Established The reply to that is the single most important technical development in this dispute, and it is a complexity-theoretic result rather than a rhetorical one. Kempes, Lachmann, Iannaccone, Fricke, Chowdhury, Walker and Cronin establish that computing the assembly index is NP-complete, and argue the non-equivalence by scaling: for n steps of Lempel–Ziv–Welch compression the reconstructible string length grows as n² in the large-string limit, whereas an assembly path of length a reaches 2^a. Established They state directly that “approximating K is not, and was never, the goal of AT”, and give counterexamples intended to show the index is not a restricted version of Huffman coding. Frontier An NP-complete quantity cannot be identical to a polynomial-time compressor unless P equals NP, which is the form the separability question has now taken.

Speculative The honest reading of those two results together is that they are about different objects, and neither side says so. The NP-completeness result concerns the exact minimum assembly index. The Lempel–Ziv-family argument and the correlation coefficients concern the pathway heuristic and the peak-counting estimator that are used in practice. Handwave It is entirely possible for an exact quantity to be NP-hard while the estimator anyone actually runs is LZ-like and correlates at 0.87 with compression — and if that is the situation, then the theory is defended by a quantity nobody measures and attacked by a quantity nobody claims is fundamental. Speculative No paper in this dispute has stated the relationship between the exact index and the deployed estimator as a bound, which is the missing result section 7 asks for.

Established The measurement side has meanwhile broadened. Jirasek, Sharma, Bame, Mehr, Bell, Marshall, Mathis, Macleod, Cooper, Swart, Mollfulleda and Cronin showed the assembly index can be inferred from three independent spectroscopies — nuclear magnetic resonance, tandem mass spectrometry and infrared — by counting carbon resonances, molecular fragments or absorbance bands. Frontier Three independent instruments giving concordant estimates is genuine evidence that the quantity being estimated is a property of the molecule rather than of the technique, and it is the strongest thing the measurement claim has.

Established The most recent work is a machine-learning study and it is not the blind test. Rutter, Sharma, Seet, Alobo, Goto and Cronin trained a gradient-boosted model to predict assembly index from mass spectra, reporting a threefold reduction in relative mean squared error against baselines — from 0.12 to 0.04 on the National Institute of Standards and Technology reference data and 0.07 on MassBank — over 32,603 molecules with MS1 data, with comparison sets of 526,968 natural products and 99 million PubChem entries, and thresholds examined between 15 and 20. Established No blind or unknown-provenance samples were tested, and no meteorite, mineral or abiotic mixture underwent direct experimental testing in that work. Frontier The study also reports something awkward for the programme’s central selling point: models built on features that require structural elucidation predict complexity better than models built on features that do not. Speculative The advertised advantage of assembly theory is that it needs no structure determination, and the group’s own 2026 result is that not needing structure costs accuracy.

Frontier The status of the commission’s decisive experiment, five years after the original paper, is that nobody has run it. Established There is no published pre-registered blind run on unknown-provenance samples by a laboratory that did not build the index, and no published proof or disproof relating the exact assembly index to the estimator used on spectra. Speculative That is the finding, and it is more informative than any of the individual results above: the theory’s supporters and its critics have both spent five years on calculations.

Frontier Six questions carry this subject, and the first is the one the field is actually arguing about.

Speculative One. Is the assembly index separable from algorithmic complexity? The published positions are that it belongs to the Lempel–Ziv family and is bounded by entropy, and that computing it exactly is NP-complete and therefore cannot be a polynomial-time compressor. Handwave Both can be true of different objects, and the result nobody has published is the bound relating the exact index to the estimator that is actually run on a spectrum. Speculative Until that bound exists, the dispute is two groups measuring different things and disagreeing about one.

Frontier Two. Does the threshold survive inorganic chemistry? Two abiotic minerals compute to 18. Speculative The defence is that a computed index for a mineral fragment is not the measured index of a mass-spectrometry-accessible molecule, which is a defensible methodological line and also a narrowing of the original universal claim. Frontier The question has an experimental answer that nobody has produced: what index does a mass spectrometer return when it is pointed at an ewingite or ilmajokite extract?

Speculative Three. Is copy number measurable in the sense the theory needs? The assembly equation weights each object by its copy number, and what a mass spectrometer returns is an ion abundance modulated by ionisation efficiency, matrix effects and fragmentation. Handwave Treating peak intensity as copy number is a step the literature takes without argument, and it is the step on which the difference between the measurement claim and the selection claim rests.

Frontier Four. Is the assembly equation a physical law or a definition? It is an exponentially weighted sum over an ensemble. Speculative A law makes predictions that could fail; a definition cannot. Handwave No published experiment reports a measured value of A for a real system together with a prediction the equation would have forbidden, and until one does the equation’s status is a matter of preference.

Speculative Five. What does the index do on the cases designed to break it? The original paper’s own caveat admits technology as a high-index producer alongside life, and the 2026 machine-learning study noticed artificial molecules with unusually high complexity values. Frontier An abiotic catalytic system that produces high-index molecules in abundance would be a direct falsification, and it is a synthesis problem rather than an instrument problem.

Speculative Six. Does agnosticism survive contact with a real mission? The selling point is that no assumption about biochemistry is required, but the estimator is trained on terrestrial spectral libraries — reference data, natural-product databases and PubChem — which are catalogues of terrestrial chemistry. Handwave A model that learns what complexity looks like from a database of Earth molecules is agnostic in its formalism and terracentric in its training set, and nobody has quantified how much of the second leaks into the first.

Established The first bottleneck is that the assembly index is not measured; it is inferred. The experimental route is a regression from the number of MS2 fragment peaks, reported at a correlation of 0.89, onto an index that was computed for known molecules. Frontier A correlation of 0.89 is a good calibration and a poor guarantee: it leaves a residual, and the whole classification hangs on whether a sample sits above or below a single integer threshold. Speculative Nobody has published the error distribution of the inferred index near 15, which is the only place the residual matters.

Frontier The second is computational and has just been quantified by the theory’s own authors. Computing the exact assembly index is NP-complete, so every deployed number is an approximation from a heuristic search. Speculative That makes the identity of the deployed estimator a first-class scientific question rather than an implementation detail, because the critics’ correlations were computed against whatever the reference implementation returns. Frontier Recent work from the group on faster assembly-space algorithms is an admission that the quantity is expensive; a 2026 preprint on formal definitions and fast approximation methods could not be retrieved for this brief.

Speculative The third is the missing abiotic control library. There is no agreed, curated set of abiotic samples — mineral extracts, meteoritic organics, laboratory prebiotic mixtures, industrial residues — against which any proposed complexity threshold must be run before it is believed. Frontier The mineral dispute is the visible symptom: a counterexample arrived from mineralogy rather than from a control set, because there is no control set. Handwave This is the same structural gap that Prebiotic Chemical Networks identifies from the origins side, and neither community owns it.

Frontier The fourth is copy number, which is the half of the assembly equation nobody has operationalised. Mass spectrometry returns abundances that depend on ionisation efficiency; converting those into the n_i the equation requires is an unsolved quantitative problem. Speculative Every published assembly-theory life-detection result therefore uses the index alone, which is the half the theory says is insufficient on its own.

Handwave The fifth is sociological and it is doing real damage. The dispute has been conducted partly through preprint revisions and public argument, with at least one contemporaneous news account noting that some of the reaction targeted the senior author rather than the content. Speculative A field where the central technical question is whether one quantity reduces to another is a field that could settle its dispute with a theorem, and instead it has produced five years of position papers.

Frontier This subject depends on four things it does not control and one it does.

Established Flight mass spectrometry is the dependency that makes the whole proposition attractive. The claim that assembly index is measurable rests on tandem mass spectrometry with data-dependent acquisition, and the instrument class already exists on planetary missions. Frontier What does not exist is a demonstration that a flight-grade instrument, with its resolution and duty-cycle limits, returns the same inferred index as a laboratory instrument on the same sample.

Established Spectral reference libraries are the second, and they are the training set. The 2026 work uses the National Institute of Standards and Technology reference data, MassBank, a 526,968-molecule natural-products collection and 99 million PubChem entries. Speculative Those libraries are the accumulated record of terrestrial chemistry, and an agnostic biosignature trained on them inherits whatever they over-represent.

Frontier Computational complexity theory is the third, and it arrived recently. The separability question is now a question about complexity classes, and the answer that would settle it is a theorem rather than an experiment. Speculative That is unusual for a life-detection proposal and it is a strength: theorems are cheaper than missions.

Frontier Mineralogy is the fourth and it was not anticipated. The strongest published counterexample came from a crystal-structure database, computed by mineralogists, and the defence rests on whether a mineral heteropolyanion is the kind of object the index was defined for. Speculative Any future threshold will have to be stated over a defined class of objects, and defining that class is a mineralogy and analytical-chemistry problem rather than a theory problem.

Frontier And the one it controls is the reference implementation. The index is defined by an algorithm, the algorithm is expensive, and the numbers everyone argues about come from code. Handwave A discipline whose central quantity is defined by a program has a dependency on that program’s version history, and no paper in this dispute reports which version produced its numbers.

Frontier Every position in section 4 has a test, and two of them would settle most of the argument.

Speculative Tier one, and the decisive experiment for this brief: a pre-registered blind run on unknown-provenance samples by a laboratory that did not build the index, paired with a proof or disproof that the assembly index is separable from compressibility. The two halves are different in kind and the brief treats them as one test because either alone leaves the other side’s objection standing. Established The blind half has a specification the 2021 paper implicitly wrote: samples of unknown origin prepared by a third party, revealed only after analysis. What it did not have was scale or independence — four samples, analysed by the group that defined the measure. Frontier The replacement needs an abiotic control set that includes mineral extracts, meteoritic organics and laboratory prebiotic mixtures, a threshold registered before the run, and a stated false-positive rate. Handwave Nobody has run it in five years, and the reason is not cost: the instrument is a tandem mass spectrometer and the samples are in existing collections.

Speculative Tier one, second half: state the bound between the exact index and the deployed estimator. The exact assembly index is NP-complete; the practical route is a heuristic search plus a peak-count regression. Frontier The missing result is a theorem or a systematic empirical study relating the two — how far the estimator can depart from the exact index, and whether that departure is what the reported 0.87 correlation with Lempel–Ziv–Welch compression is actually measuring. Speculative This is a paper, not a programme, and it would let both sides stop arguing past each other.

Frontier Tier two is the mineral experiment, and it is unusually well posed because both sides have already agreed what it would mean. The objection computed indices of 18 for ewingite and ilmajokite; the reply says computed indices are not measured indices. Speculative Dissolve or extract those minerals, run them through the same tandem mass spectrometry protocol the 2021 paper used, and report the inferred index. A measured value above 15 falsifies the threshold on the proponents’ own terms; a value below it vindicates the measured-versus-computed distinction. Frontier The experiment is a week of instrument time and it has not been done.

Speculative Tier three is the adversarial synthesis, and it is the strongest available attack. Build an abiotic catalytic system — mineral-surface, electrochemical or photochemical — and drive it toward high-index products in high copy number, then measure the assembly. Frontier A demonstration that abiotic chemistry can reach high index and high copy number together would falsify the assembly equation’s selection claim rather than merely the threshold, which is the deeper target. Handwave Nobody has proposed it as a programme, and it is the experiment a critic with a laboratory would run.

Speculative Tier four is the copy-number calibration. Take mixtures of known composition, measure ion abundances, and establish the mapping from peak intensity to molecular copy number with a stated uncertainty across matrices. Frontier Without it the assembly equation cannot be evaluated on any real sample, and every published life-detection result uses only the index. Established This is ordinary analytical chemistry with an unusual customer.

Frontier Tier five is the flight-realism test. Run an analogue sample through a flight-representative mass spectrometer under mission duty cycle and compare the inferred index with a laboratory instrument’s. Speculative The 2026 study simulated Titan and Europa conditions computationally; the physical version of that comparison is the one a payload review would ask for, and it has not been published.

Established And the studies that are not tests of this and are read as though they were. Database regression work — predicting assembly index from spectra across tens of thousands of known molecules — measures how well a model recovers a computed quantity from a spectrum. Speculative It says nothing about whether the quantity discriminates life, because every molecule in those libraries has a known provenance and none of them is a blind sample. Frontier The same applies in the other direction: the critics’ correlation analyses measure how well the index tracks other coding schemes on a fixed dataset, which is evidence about the estimator and not about what an instrument would return on Enceladus.

Frontier The engineering here is unusually modest, which is the proposition’s main attraction and also the reason it gets less scrutiny than it should.

Established The instrument is a tandem mass spectrometer with data-dependent acquisition. The measurement needs the capacity to isolate a precursor ion, fragment it, and count resolved product peaks, because the peak count is the observable that the index is regressed onto. Frontier The flight-relevant requirements follow directly and none has been stated as a specification: mass resolution sufficient to resolve the fragment peaks that carry the index, a duty cycle that permits MS2 on enough precursors, and a fragmentation energy regime matched to the calibration set.

Speculative The second requirement is on-board or downlinked computation of an NP-complete quantity. In practice this means an approximation algorithm with a characterised error, and the group has published faster assembly-space search methods for exactly this reason. Frontier For a life-detection claim the approximation’s error bound matters more than its speed, because the classification is a comparison with a single integer.

Frontier The third is sample handling, and it is inherited rather than new. Extraction chemistry, blanks and contamination control for organics on a planetary surface are the same problems every organic instrument faces. Speculative Assembly theory is more exposed to contamination than most, because a single terrestrial high-index molecule in a blank is by construction a positive detection. Planetary Protection carries the forward-contamination regime this depends on.

Handwave And the cheapest engineering requirement is a versioned, publicly pinned reference implementation of the index. The numbers in this dispute come from code, the exact quantity is NP-complete so every number is an approximation, and no paper reports which implementation produced its values. Speculative A tagged release with a test suite would let the correlation studies and the complexity proofs refer to the same object, which is currently not established.

Established The nearest page on this map is Biosignature Standards, and the division is clean. That brief owns the evidential framework: the seven-level confidence scale, the four well-documented failures of past life-detection claims, and the argument that validation is a problem in evidential standards rather than detection physics. It carries the 2021 assembly-theory paper as the leading agnostic proposal with an actual measurable threshold. Frontier This page owns the object that brief cites: what the threshold was measured on, what it has been contradicted by, and whether the quantity behind it is well defined. Speculative A confidence scale cannot grade a claim whose underlying measure is disputed, which is the joint problem neither page can solve alone.

Frontier Prebiotic Chemical Networks reaches the same missing number from the origins side. That brief names the abiotic complexity baseline as the negative control its field has never systematically run, and treats the assembly-theory dispute as the visible symptom of that absence. Speculative This page agrees and adds the specific form the control would have to take: a curated abiotic sample library, run blind, with a threshold registered in advance.

Speculative The sibling case on this map is Homochirality Origins, and the two are instructive opposites. Chirality is a well-defined physical property with a contested abiotic ceiling and an instrument that has never flown; assembly index is a contested quantity with a clear abiotic counterexample and an instrument already flying. Frontier Both fail at the same point, which is that no one has measured what abiotic chemistry can do.

Speculative Two further neighbours are methodological. Integrated Information Theory is the other place in this corpus where a proposed scalar measure of an ill-defined property is defended and attacked on formal grounds by people who agree on no experiment; the parallel is worth noticing because the failure modes rhyme. Scientific Revolutions carries the general question of how a pre-paradigmatic framework gets adjudicated, which is the situation this brief describes from inside.

Frontier One adjacency is institutional. The work sits inside an agnostic-biosignature research programme that also produced the blind samples for the original test, which is a reasonable arrangement and is also why independence is the thing tier one has to supply.

Frontier The institutional facts here are about who adjudicates, and the answer so far is journals, slowly.

Established The dispute has been conducted where disputes should be conducted, which is more than is usually true of pre-paradigmatic arguments. The mineral counterexample appeared in the Journal of the Royal Society Interface with a reply and a counter-reply; the compression critique appeared in npj Systems Biology and Applications and in PLOS Complex Systems; the complexity-theoretic defence appeared in npj Complexity. Frontier Four journals, an exchange in each, and no resolution, which is what it looks like when a field lacks a decision procedure rather than a venue.

Speculative The missing institution is preregistration, and it is cheap. The decisive experiment is defined by its independence and its advance commitment to a threshold, both of which are administrative rather than technical. Handwave A registered report — threshold, sample classes and analysis pipeline fixed before the blind samples are opened — would convert a five-year argument into a result, and no funder or journal in this area requires one.

Frontier The second missing institution is custody of the reference implementation. The index is defined by an algorithm whose exact evaluation is NP-complete, so the community is arguing about the output of approximations. Speculative A versioned implementation with a public test suite is the kind of infrastructure that standards bodies provide in other measurement fields and that nobody provides here.

Speculative The third is the abiotic control library, which needs a curator rather than a discovery. It would be assembled from existing mineral, meteoritic and laboratory collections, and its value is entirely negative — it exists to make claims harder. Handwave That is the class of research product agencies fund least and life detection needs most.

Frontier And there is a governance question about mission adoption. A complexity threshold attractive to payload teams because it needs no new hardware can be adopted into an analysis pipeline long before its status is settled. Speculative The cost of adopting a contested measure is not the measure; it is that a mission result reported against it inherits the dispute, and Biosignature Standards documents what happens to a claim that fails low on the confidence scale.

Frontier The ethical exposure here is almost entirely about a false positive with a large audience.

Established A life-detection claim that fails in public damages more than the claim. The documented pattern is that such claims fail at instrumentation, reduction or abiotic-source exclusion rather than at the interesting step, and the failures are cited for decades. Speculative An assembly-index detection reported from a mission before the threshold dispute is settled would fail at abiotic-source exclusion, which is the failure mode with a published counterexample already waiting.

Frontier The second issue is the conduct of the dispute itself. Contemporaneous reporting noted that part of the reaction to the 2023 paper targeted its senior author rather than its argument, and that a co-author was largely ignored in the criticism. Speculative A technical disagreement about whether one quantity reduces to another is resolvable by a theorem; conducting it as a reputational argument makes the theorem less likely to be written, which is a cost borne by everyone downstream.

Speculative The third is attribution. A recurring charge in the critiques is that the results were available from existing complexity measures and were presented as novel. Handwave This brief cannot adjudicate a priority claim from the published record, and notes only that the charge is specific, is made in peer-reviewed venues, and has been answered in kind.

Frontier The fourth is what a contested measure does to the people who have to use it. Payload and mission scientists are being offered an analysis that costs no mass and no power, which is an unusual thing to refuse. Speculative The honest advice is that adopting it as a reported quantity is cheap and adopting it as a detection criterion is not, and the distinction should be written into an analysis plan before launch rather than argued after a spectrum arrives.

Speculative The civilizational stake is whether life detection can be made independent of terrestrial biochemistry, and assembly theory is the most concrete attempt at that in circulation.

Frontier If a substrate-independent complexity threshold works, the search for life stops being a search for molecules Earth life makes. That changes the target list rather than the instruments: any sample with enough organic content becomes testable, including samples already returned and already archived. Speculative The first consequence would be retrospective — existing mass-spectrometry archives from Mars, Titan and cometary encounters would be re-analysed against a threshold nobody had when the data were taken.

Speculative If the threshold fails, the loss is larger than one measure. Agnostic biosignatures are a small family, the assembly index is its most developed member, and a public failure would be read as evidence that agnosticism itself does not work. Handwave That inference would be wrong and would be drawn anyway, which is an argument for settling the dispute before a mission rather than after one.

Frontier There is a third outcome that is more likely than either and gets almost no attention: partial adoption. The index becomes a routinely reported quantity, like a spectral index, without ever being a detection criterion. Speculative That would be a reasonable end state and it is not what either side in the dispute is arguing for.

Handwave Claims that assembly theory reorganises physics, explains evolution, or supplies a new account of time are made in and around this literature and are not assessed here, because no experiment in this brief bears on them. Speculative The measurable content of the programme is a number you can get from a mass spectrum, and that is what this page has graded.

These horizons track the two things that could change the assessment — a blind test and a separability result — because there is no hardware programme here to track:

• 10 yr: Frontier Expect the mineral question to be settled experimentally, because it costs a week of instrument time and both sides have committed to what the answer would mean. Speculative Expect continued machine-learning work on spectral databases, which will improve the estimator and will not address the blind test. Speculative Expect at least one further formal paper on the relationship between the exact index and its approximations, and expect it to come from the complexity-theory side rather than the chemistry side.
• 25 yr: Speculative This is the window in which a flight instrument plausibly returns assembly-index data from an ocean world or Titan, on a mission scoped for organic analysis rather than for this measure. Handwave Whether the number is reported as a detection or as a descriptor will have been decided years earlier by an analysis plan, and that decision is being made now by default. Speculative A pre-registered independent blind test, if it happens, most likely happens in this window rather than the last one, because nothing currently funds it.
• 50 yr: Speculative The plausible shape is consolidation: either the index is absorbed into the standard toolkit of complexity measures with its relationship to compression understood, or it is retained as a chemistry-specific estimator whose theoretical claims are quietly dropped. Handwave The third possibility — that the assembly equation turns out to be a genuine physical law of selection — requires an experiment nobody has designed.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The defensible statement is structural: a proposed measure whose central question is a mathematical relationship either gets a theorem or gets abandoned, and this one has been waiting five years for a theorem that two groups both think they have written.

• Depends on Two soft dependencies on this map and no hard ones. Biosignature Standards supplies the evidential framework any threshold would have to be graded within, and its finding that no journal, funder or mission requires the community confidence scale is the reason a contested measure can reach a mission analysis plan unchallenged. Prebiotic Chemical Networks supplies the abiotic complexity baseline this subject needs and that neither field has produced. Nothing on this map produces the formal result in the second token below.
• Requires (not on this map) Five conditions, in the order of the tokens above. The bound. A theorem or systematic study relating the exact assembly index, which its own authors show is NP-complete, to the heuristic search and peak-count regression actually used — the missing result that lets the NP-completeness defence and the 0.87 correlation with Lempel–Ziv–Welch compression be about the same object. The control library. A curated, shared set of abiotic samples — mineral extracts, meteoritic organics, laboratory prebiotic mixtures — against which any complexity threshold is run before it is believed; the mineral counterexample arrived from a crystal-structure database because no such library exists. The protocol. A registered-report convention fixing threshold, sample classes and pipeline before unknown samples are opened, and requiring the analysing laboratory not to be the one that defined the measure. The instrument characterisation. A flight-representative tandem mass spectrometer shown to return the same inferred index as a laboratory instrument under mission duty cycle. The implementation. A versioned, tagged reference implementation with a test suite, so that correlation studies and complexity proofs refer to one artefact.
• Enables A settled threshold would give Biosignature Standards its first agnostic biosignature with a defensible abiotic-exclusion argument, which is the level at which three of the four historical life-detection claims failed. It would give Prebiotic Chemical Networks a stopping criterion for how much complexity a prebiotic network has to reach. No typed enabling edge is claimed beyond those, because a measure under dispute enables nothing on a schedule.
• Adjacent Analytical mass spectrometry, mineralogy and crystal chemistry, algorithmic information theory and computational complexity supply every number on this page. Within this map: Homochirality Origins for the sibling agnostic biosignature with the same missing abiotic baseline, Biosignature Standards for what a claim is worth, Prebiotic Chemical Networks for the origins-side version of the control problem, Planetary Protection for the contamination regime a high-index blank would defeat, Integrated Information Theory for the nearest methodological parallel, and Scientific Revolutions for how a dispute of this shape usually ends.

Established “Assembly theory was validated by a blind test over about 114 samples.” It was not. The blind analysis in the 2021 paper covered four samples of unknown provenance — a Murchison meteorite fragment, bay sediment, and biological material from the Holocene and the Mid-Miocene. Established The number 114 appears in the same manuscript as the size of a nuclear-magnetic-resonance calibration set: “a combination of 1H and 13C NMR experiments were performed on 114 compounds”. Frontier This is the anchor error this brief was asked to look for, and it matters because the strength of the whole programme is usually asserted through that number.

Established “An assembly index of 15 or more means life.” The paper’s own sentence is conditioned on its sample set — “In the case of the analysis presented here it appears that only living samples produced MA measurement above ~15” — and its general claim admits technology alongside biology. Established Two abiotic mineral heteropolyanions, ewingite and ilmajokite, compute to exact assembly indices of 18. Frontier The proponents’ reply is that computed indices are not the measured quantity, which is a real methodological distinction and also a narrowing of a claim that was originally stated about all high-index molecules.

Frontier “Assembly theory is just Lempel–Ziv compression with extra steps.” That is one published position, supported by a formal argument that the pathway method belongs to the LZ family and by correlations of 0.87 with Lempel–Ziv–Welch, 0.9 with run-length encoding and 0.896 with Huffman coding. Established It is contradicted by a second published result from the theory’s authors: computing the assembly index exactly is NP-complete, and no polynomial-time compressor can be identical to an NP-complete quantity unless P equals NP. Speculative The reconciliation nobody has published is that these are claims about different objects — the exact index and the deployed estimator — and the brief’s decisive experiment asks for the bound between them.

Frontier “Assembly theory explains selection and evolution.” The 2023 paper claims it, and the claim rests on the assembly equation weighting index by copy number. Speculative No published experiment evaluates that equation on a real system and reports a result the equation would have forbidden, and every deployed life-detection use of assembly theory uses the index alone — which is the half the theory itself says is insufficient. Handwave A quantity defended by its authors as measuring causation, and used in practice without its causal half, is not yet an explanation of anything.

Established “The 2026 PNAS paper is the independent test.” It is a machine-learning study from the same group that reduces relative mean squared error from 0.12 to 0.04 in predicting assembly index from mass spectra across 32,603 reference molecules. Established No blind or unknown-provenance samples were tested, and no meteorite, mineral or abiotic mixture underwent direct experimental testing. Frontier It is a better estimator, which is a real contribution and a different kind of result from the one the dispute needs.

Frontier “The advantage is that you do not need to identify the molecule.” That is the programme’s central selling point for flight use and its own 2026 results qualify it: models regressed on features that require structural elucidation predict complexity better than models using features that do not. Speculative Structure-free estimation is therefore a real capability with a measured accuracy cost, and the size of that cost near the threshold is the number that has not been published.

Speculative “The critics have refuted it” and “the critics have been answered.” Neither is true on the published record. Frontier The threshold has a counterexample whose relevance is contested on methodological grounds; the compression critique has a complexity-theoretic answer that addresses a different object; the selection claim has neither a supporting experiment nor a falsifying one. Handwave A brief that declared a winner here would be inventing a result, and this one declines to.

Speculative “Nobody can run the decisive experiment.” Everybody can. Established The instrument is a tandem mass spectrometer, the abiotic samples exist in mineral and meteorite collections, and the blind protocol was demonstrated on four samples in 2021. Handwave The obstacle is that the experiment is nobody’s deliverable: the proponents have a working measure and the critics have a working argument, and neither position is improved by a result.