Designer Organisms

The claim under test here is narrower than “we can edit genomes” and much stronger: that an organism can be built to a specification. Editing capability is owned by Genetic Engineering and the engineering-discipline claim by Synthetic Biology. This slot asks the specification question, and it is testable against an unusually good evidence base, because designed organisms have been built, field-trialled, petitioned, approved, released, abandoned and liquidated, all in public and mostly on federal dockets.

Read across the record, the failures cluster in one place and it is not where the popular argument expects. The engineering is usually competent. The specification is the thing that breaks. A tree missed its blight-tolerance target and was misidentified for eight years; a de-extinction project delivered an animal its own chief scientist calls a grey wolf with twenty edits; a mammoth programme has watched its edit count grow two and a half times while its delivery date moved out by four to six years. Those are not broken constructs. They are cases where nobody knew what was being asked for.

Established The cleanest case in the subject is the American chestnut, because the full cycle is documented and it closed in 2026. SUNY-ESF engineered Castanea dentata with an oxalate oxidase gene for tolerance to the blight fungus Cryphonectria parasitica, and petitioned APHIS on 21 January 2020 for deregulation of the line Darling 58. The petition drew 4,321 comments. It was the flagship case for engineered organisms in conservation. Established Then the design turned out to be a different organism than everyone believed. In late October 2023 researchers at the Universities of New England and Maine found a probable pollen mix-up early in the breeding programme, and The American Chestnut Foundation verified in November 2023 that the oxalate oxidase gene in all trees thought to be Darling 58 sat on chromosome 4 instead of chromosome 7. Every tree studied since 2016 was a Darling 54 descendant.

Established The performance record the foundation then published is worse than the mislabelling. Growth: stunted growth, leaf browning and curling, with leaf-scorch injury 8.4 times more intense on gene-positive than gene-negative seedlings. Fertility: the homozygous state is “largely lethal”, with a great majority of homozygous offspring dying in the embryonic stage. Blight tolerance, which is the entire point: “striking variability” between trees and severe blight on some trees derived from the line. Collateral damage: the insertion deleted 1,069 base pairs of SAL1, a salinity-tolerance gene. On 8 December 2023 the foundation discontinued support for the project.

Established And on 28 August 2026 APHIS granted the tree nonregulated status anyway. ESF filed an amended petition in August 2024 correcting the line name; APHIS republished draft materials on 6 June 2025, drawing a further 4,115 comments, withdrew its environmental-impact-statement notice on 18 August 2025, and determined that Darling 54 “is unlikely to pose a greater plant pest risk than the nonmodified comparator”, with no conditions imposed. Frontier The regulatory system verified that the tree is not a plant pest. Nothing in the system verified that it is what it was designed to be — and that is not a failure by APHIS, whose remit is plant-pest risk and not efficacy. It is a gap between what gets checked and what “designed to specification” asserts.

Established Where designed organisms have been released at scale, the biology worked and the regulatory outcome did not follow. Carvalho, McKemey, Garziera, Lacroix, Donnelly, Alphey, Malavasi and Capurro reported in PLOS Neglected Tropical Diseases on 2 July 2015 that sustained release of Oxitec’s self-limiting male Aedes aegypti in Itaberaba, Bahia — about 185,000 males initially, scaling to 540,000 a week by early 2012 over a 5.5-hectare area — produced a 95% reduction in the adult population, with a 95% confidence interval of 92.2–97.5%. Frontier Four of eight authors were or had been Oxitec employees and patents were declared in the paper; the disclosure is exemplary and the interest is real.

Established The United States record is the story, and it is a story about a product that does not exist. Per the EPA’s own page, an Experimental Use Permit was issued in April 2020 for Monroe County, Florida (6,240 acres) and Harris County, Texas (360 acres), then amended on 7 March 2022 to extend Florida to 30 April 2024 and add four California counties for a total of 29,400 acres, dropping Texas. All testing concluded by 30 April 2024 and no further releases are permitted. No full FIFRA registration has been granted. The Florida Keys Mosquito Control District confirms the release phase ended and that it is now for the EPA to decide whether the product may be registered. Established Four years of releases across two states, a decade of prior Brazilian data, and no product. Meanwhile Wolbachia-based products — a bacterial symbiont, no transgene — hold full commercial EPA registration. The non-engineered comparator won the regulatory race.

Established Approvals do exist, and they are worth listing precisely because the list is short. The purple tomato, carrying anthocyanin-pathway genes from edible snapdragon and developed by Cathie Martin over roughly twenty years, cleared all US hurdles and went on sale to home gardeners in early February 2024 — the first GM food crop sold directly to gardeners. The FDA approved Genus and PIC’s gene-editing technology for PRRS-resistant pigs on 30 April 2025. Frontier The caveats travel with it: US only, commercialisation described as not immediate, and a veterinary warning that because it has not been tested under conditions representative of large-scale pork production, producers must still apply every control measure already proven effective.

Established The release record is also a liquidation record, and the salmon is the clearest case. AquaBounty’s AquAdvantage salmon went from FDA submission in 2003 to approval in 2015 — twelve years — with Canadian approval in 2012–13, first Canadian sale in July 2017 and US sales from May 2021. The Indiana farm sold in July 2024 for $9.2 million; the Canadian farms, with trademarks and patents, sold in March 2025 for $1.9 million; the last facility wound down in December 2024. As of 30 September 2025 the accumulated deficit was $374 million against $951,000 in cash, with management stating it would exhaust its resources without new funds. Frontier Those financial figures reach this brief through a tertiary source traceable to SEC filings rather than through the filings themselves. The engineered salmon works. The company is gone.

Established The medical version of the same shape is Synlogic, and it is the most direct clinical test of designing a cell to a specification. Engineered Escherichia coli Nissle to consume phenylalanine in the gut of phenylketonuria patients; on 9 February 2024 the phase 3 Synpheny-3 trial was stopped by its data monitoring committee as unlikely to meet the primary endpoint; the company ceased operations and cut over 90% of staff while holding $47.7 million. Safety and tolerability were acceptable. The organism did what it was designed to do, and it did not help.

Frontier The most heavily engineered organisms in clinical use are pigs, and their outcome data are honest. Nature Biotechnology reported on 12 November 2025 that a recipient lived with a gene-edited pig kidney for “a record-setting eight months” before it was removed and he returned to dialysis, crediting gene editing with preventing hyperacute rejection and modulating complement-driven inflammation, and not stating the cause of eventual failure. Ten-gene-edited pigs; best reported result eight to nine months, then dialysis. Frontier The fetched news item carries internally inconsistent dates and is a journal news piece rather than a research paper; the case report should be checked before the timeline is relied on.

Established The gene-drive laboratory record is genuinely remarkable and should not be undersold. Kyrou, Hammond, Galizi, Kranjc, Burt, Beaghton, Nolan and Crisanti reported in Nature Biotechnology 36:1062–1066 (2018) that a CRISPR-Cas9 homing drive targeting the doublesex female isoform in Anopheles gambiae reached 100% prevalence within 7 to 11 generations while collapsing egg production to total population collapse. Cas9-resistant variants arose every generation and failed to block the drive, because functional constraint on the target sequence means resistant alleles are not viable. That is the field’s central technical insight: choose an ultraconserved target and resistance cannot rescue the population.

Established The large-cage bridge to field conditions held. Hammond and colleagues reported in Nature Communications on 28 July 2021 on about 570 adults per 4.7 m³ cage, permitting swarming and mating behaviour: elimination in 245–311 days from a 12.5% starting allele frequency and 266–276 days from 25%, with no functional resistance alleles evolving and three non-functional mutations each staying below 1%. Their own caveat is carried with it: comprehensive resistance testing and environmental risk assessment are needed ahead of field trials. Established The self-limiting alternative works too — Strampelli, Willis, Gulliford, Burt, Crisanti, Bernardini and colleagues eliminated a 400-individual cage in Nature Communications on 27 October 2025 after eight consecutive releases of male-drive-female-sterile males at one to two against wild-type, at the cost of needing repeated releases where a self-sustaining drive needs one.

Established And then the field programme was shut down by a government. Target Malaria’s own notice records that Burkina Faso’s Ministry of Higher Education, Research and Innovation asked for suspension of all activities on 18 August 2025 and issued a termination communiqué on 22 August 2025 — one week after an 11 August 2025 small-scale release of non-gene-drive, male-bias mosquitoes in Souroukoudingan, which the programme states took place in accord with its permits. A prior non-drive sterile-male release had taken place in Bana in July 2019. Frontier Researchers close to the programme report the stated grounds as lack of transparency in protocols, divided expert assessments, ecological doubts about eradicating mosquito populations, insufficient national biosafety capacity and a lack of broad public support, with officials asking who benefits and which countries drive the initiative. Established The crucial detail for honest flagging: no gene drive was ever released. Both Burkina Faso releases were non-drive precursor strains, and the programme was suspended at the stage before the technology under test. No self-sustaining gene drive has been released into any wild population anywhere.

Handwave The strong claim — that organisms can be designed to arbitrary specification — is held by popular framing and company marketing and has no supporting case in the record. Every example examined here either missed its specification, redefined it, or watched it grow. That is not a claim about difficulty; it is a claim about what the word specification is doing.

Established The weak claim is established and should be stated plainly, because the rest of this brief is not an argument against genetic engineering. Single well-characterised traits can be engineered reliably into existing organisms: Bt crops at continental scale, AquAdvantage’s growth phenotype, PRRS resistance through a receptor knockout, purple tomato anthocyanins. One trait, well understood, into a known organism, works. Everything contested lies above that line.

Frontier Population-level specification — designing an organism to change a wild population predictably — is the field’s central open question, and it is frontier in the laboratory and speculative in the field. The lab evidence is 100% drive prevalence in 7 to 11 generations and large-cage elimination in 245 to 311 days with no functional resistance. The missing item is any open release at all. Speculative It would be settled by a monitored field release with pre-registered predictions, which is exactly what the terminated Burkina Faso programme was working toward.

Frontier Whether confinable drives can deliver suppression with a built-in off-switch is the direction the field has moved, and the cost is legible. The self-limiting cage result required eight consecutive releases. Verma, Akbari and Marshall’s Y-linked editors for invasive rodent control, in Molecular Biology and Evolution 43(7) on 23 July 2026, put nucleases on the Y chromosome to disrupt female-essential genes so that females are non-viable while males keep transmitting. Established What is demonstrated is Y-linked transgene integration and expression in mice, identification of female fertility and viability genes, and Y-linked editors recently engineered in Anopheles gambiae. Speculative What is modelled only is every population-level suppression result, and the modelled requirements are demanding: elimination within five years needs monthly releases of at least 7% of the initial male population, roughly double that if the target gene is haplosufficient, more again with fitness costs, with some stochastic runs showing rebound under functional resistance.

Frontier Whether resistance is manageable by targeting ultraconserved sequence is the field’s best idea and its least field-tested one. Resistant variants arose every generation in the original result and never blocked the drive; large cages produced only sub-1% non-functional mutants. Against that, laboratory populations lack wild genetic diversity and have no migration, and both are exactly the variables a wild release introduces.

Speculative Whether a released drive could be reversed or recalled is unresolved and the honest statement is that the burden runs both ways. Reversal and recall drives are proposals; none has been deployed against a released drive anywhere. Speculative The mirror-image claim — that drives will inevitably escape and cause irreversible ecological collapse — has no observed evidence either, because no drive has been released. Nobody has demonstrated confinement, and nobody has demonstrated escape. That symmetry is why the confinement argument persists on both sides.

Frontier The strongest new hypothesis in this slot is that social licence, not biology, is the binding constraint. Burkina Faso’s stated grounds were transparency, consent, ecological doubt and national biosafety capacity — not efficacy data and not a safety incident. The assessment from researchers close to the programme is that technical validation and laboratory milestones are insufficient without social and political alignment, and that closing an advanced insectary may deter research participation across a continent. Frontier That is a blog post by interested parties rather than a measurement, and it is the most important claim in this brief that is not a number.

Handwave De-extinction as species restoration is the most visible claim in this subject and the specialist verdict is close to unanimous against it. Colossal Biosciences announced dire wolves in April 2025: Romulus and Remus born 1 October 2024, Khaleesi on 30 January 2025, produced by 20 genetic modifications across 14 genes in grey wolf endothelial progenitor cells — 15 based on the dire wolf genome and five for light coat colour. Established Nic Rawlence and Philip Seddon call them genetically modified hybrid grey wolves and note the two lineages diverged 2.5 to 6 million years ago; Jeremy Austin says the animals are not dire wolves under any definition of a species ever used; Adam Boyko describes functional proxies rather than resurrection. The IUCN SSC Canid Specialist Group stated on 18 April 2025 that the animals are “neither dire wolves nor proxies”, that creating phenotypic proxies does not change the conservation status of an extinct species, that the project will not restore ecosystem function and does not contribute to conservation, and that it may instead threaten extant grey wolves. Established And the company conceded the framing: in May 2025 its chief scientist described the animals as “grey wolves with 20 edits” and called “dire wolf” a colloquialism rather than scientific terminology. No peer-reviewed paper on the animals is recorded.

Frontier The underlying multiplex-editing capability is real, and separating it from the de-extinction claim is the point. Twenty edits across fourteen genes in a viable mammal is a genuine technical result; 38 woolly mice carrying mammoth-associated hair genes were produced in March 2025 and reproduced. Speculative Whether that capability transfers to conservation of living species is asserted by the company and has no documented outcome yet.

Frontier Morphology can apparently be specified without touching the genome at all, and this is the most conceptually interesting material in the slot. Anthrobots, reported in Advanced Science on 30 November 2023, are multicellular motile constructs that self-assemble from adult human tracheal cells with no genetic modification. Cells form organoids; conditions are optimised so cilia orient outward; the constructs move within days. Size 30 to 500 micrometres, lifespan 45 to 60 days before biodegrading, and morphology determines behaviour — fully ciliated spheres wiggle in place while irregular patchy forms travel in straight or curved paths. Concentrated assemblies applied to a scratch in a cultured neuron monolayer produced substantial regrowth where unexposed wounds showed none. Frontier The observations come through a university press release about the paper, and the therapeutic claim is a scratch assay in a dish. Frontier The reason it matters here is that the specification operates on shape and behaviour rather than sequence, and it was reached by manipulating culture conditions — which is a search, not a derivation, and it means the word specification is doing different work than the editing cases imply.

Frontier The binding bottleneck is that the specification is discovered rather than given, and it is visible as a number. Colossal’s mammoth edit list has risen from initial estimates of about 60 genes to about 150 and still climbing, while the target date moved from 2028 to “early 2030s” — the chief executive’s own words in August 2026 being “not 2036, but not 2030 either”. If organisms could be designed to specification, the specification would be the fixed thing and the schedule the uncertain one. Here the schedule slipped because the specification is still being found. Established That is an admission against interest, from the best-funded and most publicised programme in the field, reported in a magazine interview with its own chief executive.

Established The second bottleneck is that unintended edits are not detected by the process that approves the organism. The chestnut’s 1,069-base-pair deletion in a salinity-tolerance gene was collateral to the insertion; the mislabelling of an entire line went unnoticed for eight years through field trials, a federal petition and thousands of public comments, and was found by an outside laboratory checking chromosome placement. Neither the developers nor the regulator was looking.

Frontier The third is confinement, and it is unmeasured rather than unsolved. Every confinement result in the record comes from cages: 4.7 cubic metres, about 570 mosquitoes, no migration, no wild genetic diversity. No open-field confinement data exist for any drive. The self-limiting designs trade confinement for repeated releases, which converts an ecological problem into a logistics one.

Established The fourth is release logistics, and it is larger than it sounds. Suppressing a five-and-a-half-hectare area took 540,000 males a week; the rodent modelling requires monthly releases of at least 7% of the initial male population sustained for five years. A designed organism that must be manufactured and released continuously is an industrial operation, not a deployment.

Frontier And the fifth is consent, which is not a soft constraint here but the one that actually stopped the leading programme. A self-sustaining drive is transboundary by design and cannot be recalled; there is no procedure for obtaining consent from a region, and the one national government that faced the question answered it by terminating the programme and closing the insectary.

Established This brief waits on genome-writing and editing capability, which is a result Genetic Engineering produces. Twenty edits in a wolf, ten in a pig, a homing drive in a mosquito, an oxalate oxidase insertion in a tree: every case in section 2 consumes editing capability rather than producing it. The typed edge below records it, following the convention nine other Category III briefs already use.

Frontier It also waits on something Synthetic Biology has been unable to supply, and this is the sharper dependency: the ability to predict phenotype from a designed sequence. The chestnut’s variable blight tolerance, the mammoth’s growing edit list and the dire wolf’s twenty-edit compromise are all instances of the same missing capability, in three organisms, at three scales.

Established Two dependencies are institutional rather than scientific and are recorded as typed non-brief constraints in section 13: a consent procedure for a population-scale release, and an efficacy standard for a released engineered organism. Established Both are things a government could choose to create. The absence of the second is why a tree whose own sponsor abandoned it for failing its specification could be granted unconditional nonregulated status three years later.

Frontier And one is a supply-chain constraint that the evidence supports directly: mass rearing and release at population scale, which the Brazilian and modelled rodent numbers price in units per week and per month rather than in units at all.

Established The most valuable experiment in this subject is the one that was about to happen and did not: a monitored field release with pre-registered predictions. Every gene-drive result in the literature is a cage result. The variables a release introduces — wild genetic diversity, migration, seasonality, population structure — are precisely the ones that cage work cannot supply, and the programme positioned to supply them was terminated in August 2025 before releasing anything that drives.

Frontier Second, and available immediately: test the chestnut against its own specification. Darling 54 is now an unregulated tree. Its sponsoring foundation documented striking variability in blight tolerance and severe blight on some trees. A properly powered blight-challenge trial across the surviving lines would establish whether the design works, and no part of the regulatory process required one.

Established Third: routinely sequence and publish the whole genome of every designed organism before deregulation. The chestnut’s chromosome-placement error and its 1,069-base-pair collateral deletion were both findable this way, and both were found late and by outsiders. This is a disclosure standard rather than an experiment, which is why nothing in the current process produces it.

Frontier Fourth: open-field confinement measurement using a non-drive proxy. The confinement question does not require releasing a drive to make progress on it: a marked, self-limiting strain released with pre-registered dispersal predictions would measure migration, gene flow and population structure at the site where a drive would later go, and it is the kind of study that builds the social licence the Burkina Faso experience showed to be binding.

Speculative Fifth: take Anthrobots out of the dish. The neuron-bridging result is a scratch assay reported through a press release. An in-vivo model with a controlled comparator would establish whether morphology-level specification does anything therapeutic, and it is the cheapest of the experiments listed here.

Frontier And sixth, for de-extinction: publish. Three edited wolves have existed since 2024 and 38 edited mice since March 2025, with no peer-reviewed paper on the wolves recorded. The thylacine genome was announced in October 2024 as more than 99.9% accurate and the most complete ancient genome of any species, with papers expected in early 2025; the specialist response was that it would be lovely to release the claims following the data, and that a genome is not a genetic manual.

Established Multiplex editing is the engineering that genuinely works and its scale is now documented. Twenty edits across fourteen genes in grey wolf progenitor cells, cloned to term; ten edits in pigs whose kidneys function in humans for months; mice carrying multiple mammoth-associated hair variants that breed true. The number of simultaneous edits is not the constraint. Knowing which edits to make is.

Established Drive architecture has converged on one design principle and it is a good one. Target a sequence under strong functional constraint — the female isoform of doublesex — so that resistance alleles, which arise every generation, are themselves non-viable. That single choice is what turned gene drives from a resistance-limited idea into a laboratory technology that reaches fixation in seven to eleven generations.

Frontier The self-limiting architectures are an engineering answer to a governance problem and they price it honestly. Male-drive-female-sterile constructs eliminated a 400-mosquito cage after eight releases; Y-linked editors require sustained monthly release of at least 7% of the male population, roughly doubled for a haplosufficient target. Confinement is bought with throughput, and the throughput is the deployment cost.

Established Mass rearing is a real industrial capability that exists and is under-discussed. Scaling from 185,000 to 540,000 males a week over a 5.5-hectare treatment area is a factory operation with sex-sorting, quality control and release logistics attached, and it is one of the few parts of this subject that has been demonstrated at operational scale for years at a time.

Frontier And biocontainment by design remains the most interesting unfinished engineering. Recoded chassis dependent on synthetic nutrients, daisy-chain and split drives, and reversal constructs all aim to build the off-switch into the organism. Speculative None has been tested outside containment, and a containment technology that has never been tested outside containment is an argument rather than a result.

Within this map the three closest briefs divide the subject along lines worth stating exactly. Genetic Engineering owns editing and genome writing as a general capability — the tools, the delivery, the clinical and agricultural approvals for modifying existing organisms. Synthetic Biology owns the engineering-discipline claim: standardised parts, decoupled design and fabrication, hierarchical abstraction, and whether those foundations were delivered. This brief is the specification claim — whether an organism can be built to spec — and it is tested against de-extinction, gene drives and organisms actually released, because those are the cases where a specification was written down and an outcome can be compared against it.

Also adjacent: De-Extinction Technologies, which owns the restoration question this brief touches only as a specification failure; Synthetic Ecosystems, which asks the same question one level up, about assemblages rather than organisms; Biodiversity Restoration, where the IUCN objection to phenotypic proxies belongs in its general form; Xenobiology, which owns genetic firewalls and the containment-by-alphabet approach; Future Agriculture, where the EU new-genomic-techniques settlement lands; Microbiome Engineering, which inherits Synlogic directly; and Lab-Grown Organs, the alternative route to the problem gene-edited pigs are solving.

Outside the map: the origins programme’s treatment remains the entry point for readers arriving along the programme path, and is deliberately shorter. This brief supersedes it as the slot’s coverage — it adds the chestnut, the Target Malaria suspension, the Oxitec regulatory outcome, de-extinction, Anthrobots, the commercial failure record, xenotransplantation and the EU settlement, and it corrects the programme page’s implication that gene-drive field trials are near. Further out: conservation biology, veterinary and food regulation, ecological risk assessment, and the literature on free, prior and informed consent.

Established The approval record is short and each item carries a qualification that matters. AquAdvantage salmon: FDA 2015, twelve years after submission, producer liquidated by 2025. PRRS-resistant pigs: FDA 30 April 2025, US only, commercialisation not immediate. Purple tomato: cleared, on sale February 2024. Darling 54 chestnut: APHIS nonregulated status 28 August 2026, no conditions, sponsor withdrawn. Wolbachia products: full EPA commercial registration. Established What has not been approved is as informative: Oxitec’s OX5034 holds no FIFRA registration after four years of permitted releases, and no gene drive has been authorised for open release by any regulator anywhere.

Established The largest regulatory shift in the subject happened in Europe and is not yet in force. A provisional trilogue agreement on 3–4 December 2025 under the Danish Presidency splits new genomic techniques into Category 1, treated as conventional with reproductive material labelled but consumer products not, and Category 2, remaining under GMO law with mandatory labelling. Herbicide tolerance and known insecticidal substances are automatically excluded from Category 1; member states keep opt-outs for Category 2 cultivation; and applicants must disclose patents to an EU database. Frontier It requires formal Council and Parliament endorsement, and this brief’s source is an industry trade magazine — an interested party reporting a provisional agreement.

Established The international instruments are weaker than the governance discussion assumes. CBD Decision 16/21 of 1 November 2024 adopted no moratorium on gene drives and does not mention gene drives at all; it created a technical expert group, a capacity-building and technology-transfer action plan, and reaffirmed a precautionary approach to environmental release. The WHO’s guidance framework for testing genetically modified mosquitoes, second edition, dates from 19 May 2021 and sets best practice for quality and comparability so that countries can decide on deployment. Frontier Neither instrument decided anything about drives, and the page this brief supersedes described the governance framework as maturing. The record since is one national termination and no new binding instrument.

Frontier The institutional finding that generalises is that approval and deployment come apart, in both directions. An approved salmon whose producer liquidated; a deregulated tree its own foundation abandoned; approved pigs not immediately commercial; released mosquitoes with no registration; a laboratory-validated drive with no jurisdiction willing to host it. Regulatory status has predicted almost nothing about whether a designed organism reaches use.

Established The consent problem is the live one and Burkina Faso made it concrete. A self-sustaining drive does not respect property lines or national borders and cannot straightforwardly be recalled, so the question of who consents on behalf of an affected region is not a philosophical framing device but the thing that ended the leading programme. Frontier The stated grounds were transparency, divided expert assessment, ecological doubt, insufficient national biosafety capacity and absent public support — and the question of who benefits and which countries drive the initiative. None of those is answerable with better efficacy data.

Established The de-extinction case raises a distinct and more specific objection than “playing God”, and the IUCN made it precisely. Creating phenotypic proxies does not change the conservation status of an extinct species; the project will not restore ecosystem function; it does not contribute to conservation; and it may threaten extant species. Frontier The ethical risk is misdirection rather than harm to the animals: a highly visible claim that extinction is reversible operating in the same attention economy as the conservation of species that still exist.

Frontier There is an animal-welfare cost in the record that is rarely stated alongside the achievements. The chestnut’s homozygous state is largely lethal at the embryonic stage. Xenotransplant recipients have carried edited organs for months before returning to dialysis. Cloning-and-editing pipelines are attrition-heavy by construction. Speculative No source assembled for this brief quantifies that cost across programmes, and its absence from the discussion is itself notable.

Speculative The dual-use hazard is structural rather than demonstrated. A suppression drive is species-agnostic by design, so the same mechanism aimed at a malaria vector could be aimed at a beneficial species. No incident has occurred, and the argument rests on the mechanism rather than on evidence.

Established And there is an equity dimension the technical literature under-weights. The organisms are designed in wealthy countries and released in poorer ones; the terminated programme was hosted by a country whose officials asked who benefits; and the CBD decision that declined to restrict anything spent its substance on capacity-building and technology transfer. Those are the same question asked from two directions.

Frontier The terminal position is that the failure mode nobody plans for is a wrong specification, and three of the strongest cases share it. The chestnut accumulated eight years of field data, a federal petition and thousands of comments around a tree that was not the line everyone believed it was. The dire wolf’s specification was “dire wolf” and the deliverable was a grey wolf with twenty edits, as its own chief scientist eventually said — and the IUCN’s objection is not that the edits failed but that the specification was never a species. The mammoth’s edit list has grown two and a half times and is still climbing. Frontier In each case the engineering was competent and the specification was the failure point, and that is a synthesis across three cases rather than a claim any single source makes.

Established The second durable finding is that regulatory approval has almost no predictive value here. Twelve years to approve a salmon whose producer then liquidated; unconditional deregulation of a tree three years after its sponsor abandoned it; four years of permitted mosquito releases followed by no registration while a non-engineered competitor holds one. A society that treats approval as the decision point is watching the wrong variable.

Frontier The third is that the binding constraint on the most consequential technology in this slot is social rather than technical, and that is unusual on this map. Gene drives work in cages. They are not deployed because a national government withdrew permission over transparency, consent and capacity — a week after a release that involved no drive at all. If that generalises, the technology tree for population-scale intervention runs through consent procedures rather than through constructs.

Speculative And the most interesting long-run possibility is that specification stops meaning sequence. Anthrobots reach a reproducible morphology and behaviour from unmodified human cells by manipulating culture conditions. If organisms can be specified at the level of shape and function without touching the genome, then the whole editing-centred framing of this subject describes one route rather than the route — and it is a route that reached its result by search over conditions, which is the same finding Synthetic Biology reaches from the other direction.

These horizons track regulatory decisions, release permissions and disclosure practice rather than laboratory capability, because laboratory capability is not what has been binding:

• 10 yr: Established Wider deployment of self-limiting released insects continues, and Wolbachia replacement rather than transgenesis is the likelier vehicle on current registration status. Frontier The EPA either registers OX5034 or does not; the decision has been outstanding since releases ended on 30 April 2024. Speculative A first carefully-bounded gene-drive field evaluation is possible in this window and is now less likely than it looked in 2023: the leading programme was terminated, its insectary closed, and no successor host country has been announced. Frontier The EU new-genomic-techniques regulation either receives formal endorsement or lapses.
• 25 yr: Frontier Confinable and reversible drive designs are validated in cages and possibly in bounded field settings; the open question is whether any jurisdiction hosts the test. Speculative Recoded chassis with robust biocontainment in applied use depends on results this brief's neighbour shows are not delivered. Speculative De-extinction produces a woolly-mammoth-like elephant or does not; the company's own edit count is still rising, which is the variable to watch rather than the announced date.
• 50 yr: Speculative Managed, monitored population interventions in the wild become plausible if and only if the confinement and consent problems are genuinely solved, and the second has no research programme attached to it at all. Speculative Morphology-level specification — designing constructs from unmodified cells — may turn out to be a larger field than genome-level specification, on the strength of one line of work and a scratch assay. Handwave Designing organisms and their ecological roles together as an engineering discipline assumes the specification problem was solved somewhere along the way, and nothing in the record shows the first step.
• 100 / 250+ yr: Handwave Arbitrary designer organisms to order remain a cultural image rather than a specification, and the reason is stated throughout this brief: the constraint is not edit count but knowing what to ask for. Handwave Beyond useful forecasting. The one thing the record supports at this horizon is that anything released and self-sustaining outlives every institution that authorised it, which is an argument about governance rather than a prediction about biology.

• Depends on Genetic Engineering, which owns editing and genome writing as a general capability. Every designed organism in this brief — twenty edits in a wolf, ten in a pig, a homing drive in a mosquito, an oxalate oxidase insertion in a chestnut — consumes that capability rather than producing it. That is the one typed depends-on edge claimed here, following the convention nine other Category III briefs record. The sharper dependency is on a result Synthetic Biology has not delivered, phenotype prediction from designed sequence, and it is recorded there rather than duplicated as an edge.
• Requires (not on this map) Three constraints that are not research results and that no brief on this map produces. The first is the one that actually stopped the leading programme: there is no procedure for obtaining consent from a region for a release that crosses property lines and borders and cannot be recalled. Burkina Faso's ministry suspended Target Malaria on 18 August 2025 and terminated it on 22 August 2025, one week after a release of non-gene-drive male-bias mosquitoes conducted under valid permits, citing transparency, divided expert assessment, ecological doubt, insufficient national biosafety capacity and absent public support — not efficacy and not a safety incident. The second is an efficacy standard, and its absence has a worked example: APHIS granted Darling 54 unconditional nonregulated status on 28 August 2026 on the ground that it is unlikely to pose a greater plant pest risk than the unmodified comparator, three years after the project's own foundation had documented striking variability in blight tolerance, severe blight on some trees, a largely lethal homozygous state and a 1,069 base-pair collateral deletion in a salinity-tolerance gene, and had withdrawn support. The regulator did its job; nothing in the system asks whether the organism does what it was designed to do. The third is a supply chain the evidence prices directly: suppression required scaling to 540,000 males a week over 5.5 hectares in Brazil, and the Y-linked rodent modelling requires sustained monthly releases of at least 7% of the initial male population for five years, roughly doubled for a haplosufficient target. Mass rearing, sex-sorting and release logistics at that scale are an industrial capability, not a deployment step.
• Enables Population-scale disease-vector control, engineered livestock disease resistance, xenotransplantation supply, and the organism-level inputs to Synthetic Ecosystems and Biodiversity Restoration. No typed enabling edge is claimed, because in every case the enabling step is a release that no regulator has authorised.
• Adjacent Synthetic Biology, De-Extinction Technologies, Synthetic Ecosystems, Biodiversity Restoration, Xenobiology, Future Agriculture, Microbiome Engineering and Lab-Grown Organs; and outside the map, conservation biology, ecological risk assessment, veterinary and food regulation, and the consent literature.

Handwave “Colossal brought back the dire wolf.” Twenty edits across fourteen genes in a grey wolf, fifteen of them based on the dire wolf genome and five for coat colour, in a lineage that diverged 2.5 to 6 million years ago with hundreds of thousands of genetic differences between the two. The IUCN SSC Canid Specialist Group states the animals are “neither dire wolves nor proxies”; the company’s own chief scientist calls them “grey wolves with 20 edits” and “dire wolf” a colloquialism. Established No peer-reviewed paper on the animals is recorded. The multiplex editing is a real technical achievement and it is a different claim.

Established “Gene drives are being released, or are about to be.” No self-sustaining gene drive has been released into any wild population anywhere. Both Burkina Faso releases — Bana in July 2019 and Souroukoudingan in August 2025 — were non-drive precursor strains, and the programme was terminated a week after the second, at the stage before the technology under test. Frontier Coverage that treats cage results as field-adjacent is skipping the only step that has ever been refused.

Established “The Oxitec mosquito is an approved US product.” It was released under an Experimental Use Permit; all testing concluded by 30 April 2024; no FIFRA registration has been granted, and the EPA’s decision remains outstanding. Meanwhile Wolbachia products, which involve no transgene, hold full commercial registration. The engineered organism has a decade of efficacy data and no product; the bacterial-symbiont competitor has a product.

Frontier “The blight-resistant chestnut is a success story.” It is deregulated, which is not the same thing. Its sponsoring foundation documented striking variability in blight tolerance and severe blight on some trees, a largely lethal homozygous state, leaf-scorch injury 8.4 times more intense in gene-positive seedlings, and a 1,069 base-pair collateral deletion — and discontinued support in December 2023. Established APHIS assessed plant-pest risk, which it found acceptable, and did not assess whether the tree resists blight.

Handwave “Xenobots self-replicate.” Swarms of cells push loose dissociated cells into piles that become new swarms. That requires a continuing supply of externally provided cells and produces no heredity. Established The phenomenon is real and interesting; calling it self-replication in the biological sense overstates it, and the coverage that did so is the reason the claim is in circulation.

Frontier “Anthrobots are a therapy.” They are an in-vitro observation reported through an institutional press release: a scratch in a cultured neuron monolayer, bridged where the constructs were applied and not where they were absent. Frontier The self-assembly result is genuinely surprising and the therapeutic claim is a dish.

Established “Engineered gut bacteria treat disease.” Synlogic’s phase 3 failed on efficacy with safety and tolerability intact, and the company ceased operations. The mechanism worked and the patients did not benefit, which is the least convenient result for the specification framing and the most informative.

Frontier “Gene-edited pig organs provide durable function.” The best reported result is on the order of eight to nine months before removal and a return to dialysis, from ten-gene-edited animals. That is a serious achievement against a baseline of hyperacute rejection, and it is not durable function.

Established “The CBD has restricted gene drives.” Decision 16/21 adopted no moratorium and does not mention gene drives at all. It created an expert group and an action plan on capacity-building and technology transfer. Frontier The WHO framework it is usually paired with dates from 2021 and sets testing best practice rather than deciding deployment.

Speculative “A released drive will escape and collapse an ecosystem.” No drive has been released, so there is no observed evidence for this — and the symmetric point is the one usually omitted: there is no observed evidence for confinement either. Every confinement result comes from a 4.7-cubic-metre cage holding about 570 mosquitoes with no migration and no wild genetic diversity, and no reversal drive has ever been deployed against a released drive. The claim persists because the question is genuinely unresolved in both directions.

Handwave And the framing itself: “organisms can be designed to specification” assumes the specification is the known part. The mammoth’s edit list grew from about 60 genes to about 150 and is still climbing while its date slipped by four to six years; the dire wolf’s specification was never a species; the chestnut’s was recorded under the wrong line name for eight years. In the best-funded and best-documented cases in the field, the specification was the thing that was not known.