---
title: "Glimmer Wire"
url: "https://starstuff.earth/glimmer-wire-2026-10-09.html"
updated: "2026-10-09"
description: "Edition seven of Glimmer Wire, 9 October 2026 — the weekly graded scan behind Glimmers. Six items filed, six held, five papers read end to end, median lag 68.5 days. A deep-sea coral given three different families by three genes that each said so confidently, a date for the origin of animals pinned to a place where nobody found any, a solution written in 1931 built out of caesium, and a hundred-species survey whose own conclusion refuses both halves of the question it was asked."
collection: "Glimmer Wire"
licence: "CC-BY-SA-4.0"
licence_url: "https://creativecommons.org/licenses/by-sa/4.0/"
fact_check: "https://github.com/Stimpunks/Star-Stuff/blob/main/FACTCHECK.md"
generated_by: "tools/build-markdown.mjs, from the page's own <main> landmark"
---

[Stimpunks](https://stimpunks.org/) × [More Realms](https://morerealms.com/) · Glimmer ★ Wire · Edition Seven

[Stimpunks Foundation](https://stimpunks.org/) × [More Realms](https://morerealms.com/) · Working Paper

# Edition of 9 October 2026

A weekly read of what is going well in science and medicine, graded before it is enjoyed.

The seventh edition of [Glimmer Wire](https://starstuff.earth/collection-glimmer-wire.html), the scan that sits behind [Glimmers](https://starstuff.earth/collection-glimmers.html). **Five papers were read end to end, all five versions of record — and every one of the five was read at PubMed Central, because four of the five publishers could not be reached from here at all.** Items are sorted by *lag*, which runs from **1 day to 216**, median **68.5**, and for the first time in four editions the points are spread rather than sitting in two clumps. **Four are graded verified, one contested, one plausible.** The section that matters most is [Held](#held-items-1009), and four of its six are held for one reason in four costumes: the thing in the headline is not the thing in the paper. The new finding this week is not a refusal but a ratio — **fourteen of the seventy-six science-news items we scanned lead with a picture a machine drew, and thirty-seven more with a stock photograph.**

---

*Edition* **Seven** *Published* **9 October 2026** *Amended* **—** *Verified* **4 of 6**

6 filed · 6 held · 4 seeds · 4 verified · median lag 68.5 d

## What was actually read: five papers of record in full, every one of them from the same archive

**Answered when tried, by name:** api.crossref.org, api.semanticscholar.org, pmc.ncbi.nlm.nih.gov (on a retry), doi.org, journals.plos.org, arxiv.org and www.sciencedaily.com. **Refused at the network layer — the gateway answered 403 to the connection itself, which is not a paywall and is recorded as what it was:** zookeys.pensoft.net, peerj.com, www.tandfonline.com, www.sciencedirect.com, www.ncbi.nlm.nih.gov, eutils.ncbi.nlm.nih.gov, europepmc.org, www.ebi.ac.uk, api.openalex.org, and — for the one fact on this page we wanted and could not get — www.nobelprize.org and api.nobelprize.org. **Refused by the publisher, with a real answer:** www.science.org (HTTP 403, a 5,657-byte body) and www.nature.com, which returns HTTP 303 redirecting to idp.nature.com/authorize — an authentication handshake, in front of articles published under CC BY. **Rate-limited:** www.biorxiv.org (HTTP 429), a third week running.

**The single fact worth taking from that list: all five papers read end to end this week were read at PubMed Central, and nowhere else was possible.** Four of the five are openly licensed — the coral paper CC BY at Pensoft, the wasp paper CC BY at PeerJ, the primate paper CC BY at Oxford, the Bethe-string paper CC BY at Nature — and not one of those four publishers could be reached from here at all. Last week this page wrote that an open licence is a permission rather than a road. *This week the road was singular.* One archive, run by one government, stood between this edition and six abstracts; if it had answered the way its own parent domain did, there would be nothing below but deposits.

**PubMed Central repeated both of its documented behaviours, and one of the numbers has now held for three consecutive weeks.** A wrong search path still returns the PMC home page at HTTP 200 — **67,236 bytes, the identical figure recorded on 25 September and 2 October**. And three of the seven articles requested were answered on the first attempt with Google’s reCAPTCHA challenge, served at 200 in bodies of 20,416 to 20,485 bytes. Two cleared on the second try with a different user agent; the Bethe-string paper took three. *The right door was open the whole time and a doorman answered it, with a success code.*

**The new thing this week is not a refusal, it is a picture, and it is measurable.** Last week this page held a story because the only summary of it available anywhere was written by a language model and correctly labelled as such. The companion arrived without being looked for. **Of the 76 releases ScienceDaily carried in the seven days to 9 October, 14 lead with an image credited to a machine** — thirteen read Credit: AI/ScienceDaily.com and one reads Credit: NASA; background by AI/ScienceDaily.com — and **37 more lead with a Shutterstock stock photograph**. That is **51 of 76** whose first picture of a finding comes from neither the research nor the institution that did it. *The credit lines are honest and that is the entire defence; the ratio is the finding.* Two of the six items held below carry one. We have not counted earlier weeks, so this is a single measurement and not a trend, and we are not claiming an illustration misleads anybody — only that a reader scrolling a feed now meets a machine’s idea of a result before they meet the result.

**The lag distribution finally has a middle.** Six filed items run **1 to 216 days, median 68.5** — and unlike the last three editions, the points do not sit in two clumps with a hole between them: 1, 6, 38, 99, 172, 216 is a spread. **Two of the six are floors rather than measurements**, marked where they appear: Crossref deposits no published-online value for either the *Science Advances* paper or the *Science* paper, so their lags are computed from an issue date and the true interval can only be longer. *No preprint was involved in any filed item*, so the axis’s standing blind spot had nothing to hide this week — an absence recorded as an absence, not a clean bill.

**A note on the wire dates, and one on this checkout.** Every lag below is computed against the timestamp inside the release’s own URL rather than the date ScienceDaily’s index prints beside it, because those two disagree by a day for several items again this week; each entry prints the wire date it used. And the shallow-clone fault logged on 2 October was present again: this checkout arrived holding **50 commits of 463**, which dates *Bone Song* to 13 September 2026 instead of 17 July. It was tested for before anything was generated and git fetch --unshallow was run first, so no derived file on this site was written against a truncated history.

## Filed

Sorted by lag, shortest first. Bars are scaled to the longest item, 216 days. **Six items, and for the first time in four editions they are spread rather than clumped** — 1, 6, 38, 99, 172 and 216 days, median 68.5. **Four verified, one contested, one plausible.** Five were read end to end as versions of record; the sixth rests on the publisher’s own deposited abstract and says so on its face.

Verified Primary source read and confirmed.

Plausible Finding sound and metadata confirmed; version of record not read.

Contested Specialists disagree, or the claim outruns the method.

Held Could not be traced to a source that supports it.

Contested *Published* 2026-10-02 (issue) *Journal* Science Advances *Read* the paper, in full

Lag 1 d

### The date animals began is pinned to a place where nobody found any

A *maximum* age calibration on a molecular clock is the claim that a lineage did not yet exist at some date — and the only thing that can support such a claim is an absence. The paper states the practice in its own plain words: “Maximum calibrations are based on evidence-of-absence.” The Weng’an Biota of Guizhou, phosphatised microfossils of about 590 million years old preserved finely enough to show ridges half a micron wide, has been used exactly that way — no animal is in it, the preservation is extraordinary, therefore animals had not yet evolved. This study tests that inference against a second deposit. The Kheseen Biota of Mongolia is roughly **550 to 526 million years old**, shares ten named microfossil taxa with Weng’an, formed under directly analogous depositional conditions, and preserves polygonal ridges under one micron and processes as fine as 1.5 microns. More than **140 samples** were examined by scanning electron microscopy. It contains no definitive animals either — at a time when animals, eumetazoans and bilaterians are documented elsewhere in the world, in the Nama assemblage, with biomineralised skeletons, crown-group sponges and the trace fossils of things that moved. **If a deposit of that quality can miss animals that were certainly alive, Weng’an missing them establishes nothing.** The authors rule the Ediacaran deposits out as maximum calibrations, propose late Tonian units instead — Svanbergfjellet, Bitter Springs, Wynniatt, about 850 to 730 million years old — and report that their analyses support a late Tonian to mid-Cryogenian origin for animals, around 800 to 700 million years ago.

**The star stuff frame** **A measurement of nobody-found-it is a measurement of the finding.** It is not a fact about the thing, and the whole weight of the date rests on the assumption that the method would have noticed. The paper is not arguing that the fossil hunters were careless; it is showing, with a worked case, that a search good enough to find anything can still fail to find something that was there — and that the failure leaves no trace to distinguish it from a real absence. *This site already has the receiving end of that argument.* [No. 41](https://starstuff.earth/conquering-gaze-zine.html) and [No. 97](https://starstuff.earth/the-channel-you-cannot-mask-zine.html) both turn on evidence of absence; so does every diagnostic history in which somebody was not autistic until a clinician looked. Here it is the oldest version of the same sentence, about everything with a nervous system, and the discipline caught it on itself.

**Why this is contested, and what the headline is doing** **There are two claims here and only one of them is settled by this work.** The critique of the calibration is an argument about method, documented at length through the two biotas’ shared taxa and matching taphonomy. The 800-to-700-million-year date is a clock estimate, and clock estimates for animals have moved in both directions for twenty years on exactly this choice of calibration — the paper’s own first sentence is “Divergent models persist.” Specialists will argue this one for a decade. **And the headline’s “200 million years earlier than we thought” appears nowhere in the abstract.** The release reaches it by measuring from the unambiguous body-fossil record at about 574 million years, which gives between 126 and 226 million depending which end of “800 to 700” you take. Nor is it news that animals might predate their fossils: the paper’s own introduction says lipid biomarkers already put them at least 75 million years earlier, and calls the gap a stark dichotomy. **The lag is a floor, not a measurement** — Crossref deposits no online-first date for this paper, only the 2 October issue, so one day is the shortest the true interval can be.

**Where it goes** Checked against the site as it stood before this page existed. Ediacaran appears on [No. 27](https://starstuff.earth/every-nervous-system-zine.html); evidence of absence on No. 41 and absence of evidence on No. 97 and the [tortoise guide](https://starstuff.earth/tortoise-field-guide.html); Cryogenian returns nothing. What is unspent is the mechanism rather than the period: **a date derived from a null result, and the conditions under which a null result is worth anything.** That is a [Star Stuff](https://starstuff.earth/collection-star-stuff.html) zine — one settled fact followed honestly until the belonging claim is already inside it — and the second panel writes itself from the clinical literature without us having to reach for it.

**Lole Durbin, O. M., Anttila, E. S. C., Briggs, D. E. G., Macdonald, F. A. & Anderson, R. P.** “Re-evaluating molecular clock maximum age calibrations revives pre-Ediacaran divergence estimates for animals.” *Science Advances* **12**(40), issue of 2 October 2026. **DOI** 10.1126/sciadv.aeg6289. **Read:** the published article in full at PubMed Central (PMC13632735), including the age model, the full discussion and the calibration argument; every quotation and figure above is from it. **Not read:** the Supplementary Materials, which carry the sample inventory and the clock analyses themselves. www.science.org returns 403. **Date note:** issue date only; the 1-day lag is a floor. **Wire:** ScienceDaily, 3 October 2026.

Plausible *Published* 2026-10-01 (issue) *Journal* Science *Read* the abstract, deposited at Crossref

Lag 6 d

### The bacterium is listening for the virus building its own head

CBASS — cyclic oligonucleotide-based antiphage signalling systems — are bacterial immune pathways that the abstract describes as “evolutionarily related to cGAS-STING … in humans.” An enzyme senses that a phage has arrived and manufactures a nucleotide signal that starts the defence. The question has been what, exactly, it senses. This paper reports that the trigger is the phage’s **prohead protease**: the enzyme a phage uses to cut and assemble its own capsid also cleaves an activation loop on the bacterial enzyme, and that cut is the signal. The authors reconstitute CBASS recognition of phage T4 in a test tube, show that phage prohead proteases activate CBASS in living cells, and say this “explain\[s\] how immunity is initiated late during infection.” Structure-based phylogenetics then finds activation loops in three separate clades of these enzymes and identifies the residues that decide which phage families each one recognises.

**The star stuff frame** **The sense organ is tuned to the one thing the attacker cannot stop doing.** A phage can disguise a protein, change a receptor, suppress a gene; it cannot decline to build a head and still be a phage. So the detector is pointed not at the intruder’s identity but at an act it is committed to — which is also why the alarm goes off late, and why late is not a failure. *And the kinship is the abstract’s own word.* The architecture running in a bacterium is evolutionarily related to the one running in you, which is the sort of sentence this site exists to repeat: not a metaphor about interdependence, a shared piece of machinery.

**This is an abstract and nothing more** **Every sentence above comes from the publisher’s structured abstract deposited at Crossref.** Nothing from the results, the methods, the controls or the limitations is on this page, because www.science.org returns HTTP 403 to us — a 5,657-byte refusal, the same answer it gave on 2 October — and there is no PubMed Central copy. The deposit carries no numbers at all: its strongest quantitative word is *widespread*, and we have not converted that into a figure. The release’s framing, that bacteria turn a virus’s own weapon against it, is the institution’s and not the abstract’s. **The lag is a floor** — issue date only, no online-first value deposited.

**Where it goes** CBASS returns nothing here. phage appears on [No. 105](https://starstuff.earth/the-same-four-letters-zine.html), but there it is cyanophage S-2L and the *alphabet* of the genome rather than immunity. Nothing here is ready to be built on: filing a Science abstract is filing a lead with a journal name attached, and the honest next step is the paper, not a zine.

**Hobbs, S. J. & Kranzusch, P. J.** “Phage proteases activate CBASS antiphage immunity.” *Science* **394**(6819): 102–106, issue of 1 October 2026. **DOI** 10.1126/science.aeg3949. **Read:** the complete structured abstract as deposited by the publisher at Crossref; both quotations above are verbatim from it. **Not read:** the paper. www.science.org returns 403; no PMCID exists; Semantic Scholar records it as closed access with no open PDF. **Date note:** issue date only; the 6-day lag is a floor. **Wire:** ScienceDaily, 7 October 2026.

Verified *Published online* 2026-08-29 *Journal* Nature Communications *Read* the paper, in full

Lag 38 d

### A solution written in 1931, finally built out of caesium

In 1931 Hans Bethe solved a one-dimensional chain of interacting particles exactly, and among the solutions were states that have been called *Bethe strings* ever since: bundles of particles bound to one another by nothing but their mutual interaction, with no chemical bond anywhere, and **capable of existing only in one dimension**. Innsbruck, with theory groups in Amsterdam and Munich, has now made a crowd of them. A caesium Bose–Einstein condensate is loaded into two crossed lattice beams, producing thousands of isolated one-dimensional tubes — transverse trap frequency 10.5(1) kHz, longitudinal 29.3(1) Hz, lattice depth 25 recoil energies — and the interaction between the atoms is then swept from repulsive to attractive. The gas does not collapse into one lump. It forms a hierarchy of bound states of different sizes, and the measured average binding energy indicates strings of **more than six particles**. **The identification is the cleverest part and it is a comparison of two expansions.** Let the gas expand along the tubes and the strings survive, colliding with one another without breaking. Then release the same gas into three dimensions, where a Bethe string cannot exist at all: the states come apart, and the energy that had been holding them together turns into motion, so the cloud flies apart faster. The difference between the two expansion energies is the measurement. The results agree quantitatively with generalized hydrodynamics.

**The star stuff frame** **The thing that holds them together is the same thing that would otherwise scatter them, and it only works inside the constraint.** Take the tube away and the string does not break so much as stop being possible — one dimension was not a cage it was kept in, it was a condition it was made of. That is a precise statement in physics and a dangerous metaphor in everything else, so here is the honest version: *it cuts both ways.* A structure that exists only under a condition is as real as any other while the condition holds, and claiming it as proof that rooms make people is reading a Bose gas as a parable. What this site can take from it is narrower and better — **that “it only exists in there” is not the same sentence as “it is not real.”**

**What is new, and one thing on this page we could not confirm** Bethe strings have been observed before — in quantum magnets and in superconducting qubits — and both the paper and the release say so. What is new is the dense intermixture of many strings at many sizes, including large ones, which is what the paper claims and all it claims. Its own standing hedge is the word **“nearly integrable”**: the gas approximates the solvable model rather than being it. **And one sentence we wanted is not here.** The Innsbruck release calls Bethe a Nobel laureate, which he was; we wanted to add that his 1967 prize was awarded for work on how stars produce energy — the nucleosynthesis this entire site is named out of — and *we could not confirm the citation at its source*, because www.nobelprize.org and api.nobelprize.org are both refused at the network layer here. It is recorded as unconfirmed and should not be quoted off this page.

**Where it goes** integrable, bound state and soliton return nothing across the site. Bethe returns exactly one file, and it is a false positive of precisely the kind this house keeps a note about: the only match is the eyebrow *Link 3 · Bethesda, 1949–1952* on [No. 85](https://starstuff.earth/the-lizard-that-was-never-there-zine.html). Unspent, then — though the piece worth building is not “quantum physics is like us.” It is the measurement: **two expansions, one of which is only there so the other has something to be compared against.**

**Horvath, M., Bastianello, A., Dhar, S., Koch, R., Guo, Y., Caux, J.-S., Landini, M. & Nägerl, H.-C.** “Probing Bethe strings in an attractive one-dimensional Bose gas.” *Nature Communications* **17**, 10330, online 29 August 2026. **DOI** 10.1038/s41467-026-76018-0. CC BY 4.0. **Read:** the paper in full at PubMed Central (PMC13623881), including the trap parameters, the binding-energy expression and the expansion protocol; every figure above is from it. It took three attempts — reCAPTCHA at HTTP 200 twice, the article on the third request with a third user agent. The paper’s own citation for the 1931 result is Bethe, H., *Zeitschrift für Physik* **71**, 205 (1931), which we have as a reference line and have not read. **Not read:** the Supplementary Information. www.nature.com answers 303 to an authentication endpoint. **Wire:** ScienceDaily, 6 October 2026.

Verified *Published online* 2026-06-30 *Journal* ZooKeys *Read* the paper, in full

Lag 99 d

### Three genes, three families, and every one of them sure

Off Isla del Coco, Costa Rica, in October 2023, the remotely operated vehicle *SuBastian* found gardens of a bright yellow, fan-shaped soft coral growing out of a mat of brittle stars so dense that the colonies’ holdfasts are buried under it. Depths of 362 to 529 metres; water at 8.12 to 10.8 degrees Celsius; and oxygen between 0.5 and 3.0 micromoles per kilogram, which is an extremely low oxygen zone. The animal turned out to be new, and so did its genus, and so did its family. **What makes it a Glimmer Wire item is what the three standard barcoding markers did with it.** The mitochondrial gene *mtMutS* placed it inside Incrustatidae — a family of encrusting corals that have no axis at all, which this one plainly does — with **ultrafast bootstrap support of 100% and SH-aLRT of 100**. The other mitochondrial gene, *COI*, placed it in a different family, Eunicellidae, at 91% and 91. The nuclear *28S rDNA* tree was poorly resolved and suggested a third position again. Phylogenomics settled it: ultraconserved elements and exons, **153 loci and 37,654 base pairs** at 75% taxon occupancy, confirmed on a second matrix of 1,385 loci and 263,175 base pairs, support the *COI* answer. The morphology matches none of the three outright, which is why there is a new family at all.

**The star stuff frame** **Strong support is a statement about the instrument.** Two genes out of the same mitochondrion, aligned the same way, analysed the same way, returned incompatible families — and both reported near-maximal confidence while doing it. The paper writes the consequence out rather than leaving it implied: “Use of this gene alone to infer the phylogenetic placement of *Laurinque* gen. nov. would have resulted in a spurious assignment to a family with which it shares few if any diagnostic morphological characters.” *A confident number from a single marker is the marker telling you about itself.* This site keeps arriving at that sentence from the other end — one score, one scale, one cut-off, pointed at a person — and here it is a coral, with the error visible because somebody ran three instruments instead of one and let them disagree in print.

**The name, and the hedges** **We are not going to pretend the name is the finding.** But it is worth the line: *Laurinquë* is Quenya for *golden tree*, after the colour of the colonies, and the species epithet *elenya* is Quenya for *stellar* — chosen, in the paper’s own etymology, “in allusion to the field of stars where the species was found,” meaning the brittle stars it is growing out of. A taxonomic paper that names an animal for the company it keeps rather than for a collector is doing something small and deliberate, and the title is *A coral among stars*. The hedges that belong beside it: the species has been seen in Panamá and the Galápagos, and the authors say that occurrence “needs to be corroborated”; the deepest record, at Seamount 6, found scattered colonies and no gardens; and the family rests on two specimens for the phylogenomics, not on a population.

**Where it goes** brittle star, seamount, Costa Rica, new family, barcod, incongru, Tolkien and Quenya all return zero across the site; octocoral appears once, in [this page’s own fourth edition](https://starstuff.earth/glimmer-wire-2026-09-18.html). Entirely unspent, and it belongs in [Kin](https://starstuff.earth/collection-kin.html) rather than in [More Than Human](https://starstuff.earth/collection-more-than-human.html) — the moral is not about the coral’s world, it is about ours, and it is **what a confident answer is worth when you only asked once**.

**Breedy, O., McFadden, C. S., Murillo-Cruz, C., Yánez-Suárez, A.-B., Robert, K. & Quattrini, A. M.** “A coral among stars: A new octocoral family (Anthozoa, Octocorallia, Malacalcyonacea) from seamounts in the tropical eastern Pacific.” *ZooKeys* **1283**: 253–272, online 30 June 2026. **DOI** 10.3897/zookeys.1283.191899. CC BY 4.0. **Read:** the paper in full at PubMed Central (PMC13342932) — the diagnosis, both etymologies, the ecology note, the molecular results and the discussion; every figure and quotation above is from it. **Not read:** Supplementary material 1, the taxon list underlying the gene trees, and the scanning electron micrographs at resolution. zookeys.pensoft.net is refused at the network layer here, so an openly licensed paper was readable only because a government archive carries it. **Wire:** ScienceDaily, 7 October 2026.

Verified *Published online* 2026-04-15 *Journal* PeerJ *Read* the paper, in full

Lag 172 d

### A third animal moves into a two-party arrangement, and the plant is measurably fine

*Macaranga* trees hollow their own stems and grow food bodies for *Crematogaster* ants; the ants live inside and defend the tree from anything that eats leaves. It is one of the most studied mutualisms there is. This paper reports a third party: *Dasyproctus agilis*, a cavity-nesting wasp, hollowing out *Macaranga* stems itself and provisioning the chamber with adult flies for its larvae — the first record of this wasp nesting in any member of the plant family Euphorbiaceae, and the first record of it in Sabah at all. **Where it turns up tracks the habitat, hard.** Wasps occupied **12 of 43** *M. pearsonii* trees in oil palm plantations against **1 of 41** in heavily logged forest — two-tailed Fisher exact test, *p* = 0.002. Between plants, larger ant colonies went with fewer wasps once stem diameter was accounted for, and a null model found ants and wasps segregated across plants more than chance would give (*p* = 0.035).

**The star stuff frame** **The partnership is not failing because a partner betrayed it. The room changed.** The authors’ own reasoning is about homogeneity: an oil palm plantation is biotically and abiotically uniform, fertilised, herbicided and pesticided, and that uniformity suits *a widespread generalist that is pre-adapted to anywhere* better than it suits a lineage co-evolved with one genus of tree. The wasp is not an invader with a motive; it is the species that was already shaped for the conditions somebody made. *Which is the environment rung, stated in insects* — and it is the reason this page prefers habitat findings to villain findings, because only one of the two tells you where the lever is.

**The harm is not demonstrated, and the paper says so twice** **The best model for leaf area loss contained no predictors at all** — an intercept and nothing else, across 113 plants. The authors’ summary sentence is flat: “There was no evidence for reduced ant protection of plants from herbivores in conjunction with wasp colonisation.” On the correlation between wasp numbers and ant colony size, their own words are “although causal direction remained unclear” — the wasps may be crowding out small ant colonies, or may simply be moving into trees whose ant colonies were already small. At the within-plant scale, wasp abundance did not enter either of the two best models. The abstract’s verb is *potentially disrupts*. **“Wasps are taking them over” is a tense and a certainty this paper does not have**, and the authors also allow that the whole phenomenon may be a reporting gap rather than a new colonisation — earlier studies of this system may simply never have sampled degraded habitat.

**Where it goes** mutualism is well spent here — [No. 92](https://starstuff.earth/two-things-on-one-leaf-zine.html), [the Symbioses guide](https://starstuff.earth/symbioses-field-guide.html), [Third Nature](https://starstuff.earth/third-nature-field-guide.html), [No. 75](https://starstuff.earth/the-mushroom-that-cannot-be-farmed-zine.html). But Macaranga, oil palm, cheater, exploiter and free-rider all return **zero**. The unspent object is specific: not a parasite of one partner, but **a third party helping itself to the traded good**, and the habitat condition under which it shows up. That is an entry in the Symbioses guide before it is a zine.

**Lestina, D., Houadria, M., Gavin, J. A., Chung, A. Y. C., Yusah, K. M., Rindos, M., Schlosserová, K., Feldhaar, H. & Fayle, T. M.** “Exploitation of an ant-plant mutualism by a cavity-nesting wasp.” *PeerJ* **14**: e20984, online 15 April 2026. **DOI** 10.7717/peerj.20984. CC BY 4.0. **Read:** the paper in full at PubMed Central (PMC13091584) — abstract, field methods, every result subsection and the discussion; all figures and quotations above are from it. **Not read:** Tables S1–S4 and Fig. S1, which carry the model comparisons behind the herbivory result. peerj.com is refused at the network layer here. **Wire:** ScienceDaily, 4 October 2026.

Verified *Published online* 2026-03-03 *Journal* Evolution Letters *Read* the paper, in full

Lag 216 d

### The paper refuses both halves of the question it was asked

One hundred free-ranging, non-provisioned, group-living primate species, humans among them. Five kinds of aggression scored as separate traits rather than as one: mild aggression between groups and within groups, killing of adults between groups and within groups, and infanticide. Bayesian phylogenetic mixed models then ask how far each pair has evolved together. **The three lethal kinds hang together, moderately.** Between-group and within-group adulticide correlate at 0.52 (95% credible interval 0.01 to 0.87) in the null model and 0.51 (0.06 to 0.85) in the full one; between-group adulticide with infanticide at 0.44; and all of these are stronger for male attackers than female ones. **Mild aggression does not join them.** Every correlation involving a mild type has a credible interval straddling zero, in both models and across five alternative datasets built with different ways of scoring mildness. Occurrence, for scale: 22% of the species show between-group adulticide, 19% within-group, and 65% infanticide; the mean proportion of between-group encounters that were mildly aggressive is 0.52, with a standard deviation of 0.36.

**The star stuff frame** **The finding is not that we are gentler than somebody thought. It is that the axis does not exist.** The paper’s own conclusion carries a parenthesis the coverage drops: “humans and other animal species cannot be described as ‘indiscriminately’ violent (or peaceful); different types of aggression are not all strongly evolutionarily linked, and they can occur under distinct socio-ecological contexts.” *Or peaceful.* It declines both answers to the question it was handed, which is a much harder thing to publish than either. And the corresponding author says the consequence outright in the release: it is “biologically inaccurate to rank species on the basis of their general propensity for aggression.” **A ranking needs one dimension.** Five measures that will not correlate are five dimensions, and there is no line left to put anybody on — which is the [spiky profile](https://starstuff.earth/not-a-line-broadside.html) argument, Damian Milton building on Francesca Happé, arriving from comparative primatology, about a different trait, in a literature that has never heard of us.

**What it could not do, and where it disagrees with its own field** **The data are presence and absence, not rates.** The authors say so: “we could not consider the rates of lethal aggression due to lack of data,” so whether rarely killing and often killing are different adaptations is simply not addressed here. The adulticide counts are called “a conservative figure” — killings are rare, hard to witness, and only a handful of species have long-term study across multiple populations. Infanticide is reported about three times as often as adulticide, which the authors suggest may be a fact about what gets studied rather than about primates. **And the result conflicts with a published one**: Gómez *et al.* (2021) found adulticide and infanticide correlated in female but not male mammals, and this paper finds the reverse, which it states rather than smooths. Finally, “including humans” means one row in a table of a hundred species. *Nothing here measures a human population, and no sentence about any human society can be drawn off it.*

**Where it goes** infanticide appears on [Star Gazing’s](https://starstuff.earth/collection-star-gazing.html) collection page and in [the Being Family guide](https://starstuff.earth/being-family-field-guide.html); adulticide, hardwired and rank species return nothing. The spiky profile is thoroughly spent as an argument about people — it is drawn on the face of [a broadside](https://starstuff.earth/not-a-line-broadside.html). **What is unspent is the same argument arriving from outside**, with a hundred species and a null result where a correlation was assumed, in [Kin](https://starstuff.earth/collection-kin.html): citation-dense natural history with a neurodivergence moral, and the moral is that the ranking was the error.

**Majolo, B., Wakes, S. J. & Ruta, M.** “Origins of violence: evolutionary decoupling between mild and lethal conspecific aggression in primates.” *Evolution Letters* **10**(2): 207–216, online 3 March 2026. **DOI** 10.1093/evlett/qrag002. CC BY 4.0. **Read:** the paper in full at PubMed Central (PMC13043903), including Tables 2 and 3 and the whole discussion; every number and quotation above is from it. **Not read:** the supplementary file, which holds the four alternative mild-aggression datasets. academic.oup.com is refused at the network layer here. **Wire:** ScienceDaily, 5 October 2026.

## Held

Six this week, and four of them are held for the same reason in four different costumes: **the thing in the headline is not the thing in the paper**. One is a specimen tens of millions of years younger than its own headline number. One is a discovery from 2023 carried in the present tense. One puts a synthetic ruthenium compound inside photosynthesis. One reports a 2020 result as this week’s. The other two are held for want of a source at all — three lions with no reachable paper, and a feed item that carried no journal reference. **Three of the six are papers we would have liked to file**, and one of them is the single best-looking result of the week. Nothing here was invented to fill the section, and nothing looked at and declined has been quietly omitted.

Held — the specimen is not the age in the headline, and the paper says eukaryotes where the coverage says animal

### 97-million-year-old fossils reveal the earliest known animal “GPS”

**We wanted this one more than anything else this week.** Giant magnetofossils are micron-scale crystals of biogenic magnetite, ten to twenty times bigger than the ones magnetotactic bacteria string into compass needles, and nobody knows what made them; the standing interpretation is armour. This paper images one in three dimensions by magnetic vector tomography and finds a **single magnetic vortex whose response is optimised to spatial variation in the *intensity* of Earth’s field** — which is the harder half of a navigational map sense, and a genuinely new kind of evidence. Three things stop it being filed. **First, the particle imaged is not 97 million years old.** The Methods place the sample in International Ocean Discovery Program North Atlantic Hole U1403B, section 23X1W, 24–25 cm, “within the Paleocene–Eocene Thermal Maximum period” — conventionally about 56 million years, a figure the paper does not state and which we are therefore not taking from it. The 97 million years is the age of the *oldest sediments in which giant magnetofossils of any kind are known*, cited to other work; the release prints it as the age of the fossils examined, and the headline prints it as the age of the finding. **Second, the paper says eukaryotes and the coverage says animal.** Its words: giant magnetofossils “are not considered to be the product of bacteria, but rather (as yet unidentified) eukaryotes.” The release says “an ancient animal.” **Third, the claim is explicitly conditional:** “if confirmed, our work would provide the first fossil evidence that navigational magnetoreception developed as a sense in eukaryotes at least 97 million years ago” — and the discussion allows that the particle may have evolved for protection “in some small worm” and been put to navigational use only in a descendant. One particle, one vortex, an unknown maker. *Harrison, R. J. et al., “Magnetic vector tomography reveals giant magnetofossils are optimised for magnetointensity reception,” Communications Earth & Environment 6, DOI 10.1038/s43247-025-02721-3, CC BY, online 20 October 2025 — read in full at PubMed Central (PMC12537488); wire ScienceDaily, 30 September 2026; lag 345 days, by some way the oldest thing on this page.*

Held — the discovery is from October 2023, and the lag is 1,098 days

### Scientists accidentally discover a genetic code that breaks the rules of life

The result is lovely and exactly our register. Testing a sequencing pipeline on whatever was in a pond in Oxford University Parks, a team assembled the genome of an unknown ciliate — *Oligohymenophorea* sp. PL0344 — in which two of the three stop codons have been reassigned, and reassigned to **different amino acids**: UAA to lysine, UAG to glutamic acid, with suppressor transfer RNAs to match. Every previously reported variant had moved those two together, so their evolution was assumed to be coupled. The most conserved feature of life turns out to be a local arrangement, and the group holding the most variants of it is a phylum of protists almost nobody looks at. **It was published on 5 October 2023.** The release says so in its fourth paragraph, which is to its credit; the headline is in the present tense, and a feed shows the headline. The three journal references ScienceDaily attaches are that 2023 paper, a December 2024 follow-up on UAG reassignments in phyllopharyngean ciliates, and a March 2026 paper about *Bodo* flagellates and their bacterial endosymbionts — a different subject. **Nothing here is from this week, or this year.** Our own [No. 105](https://starstuff.earth/the-same-four-letters-zine.html) is about the four *letters* rather than the dictionary, so this does not correct it and does not duplicate it; if it is ever built, it is a separate piece about the codon table, and it will be built from the 2023 paper with the 2023 date on its face. This is also the second of the six held items whose lead illustration is credited AI/ScienceDaily.com. *McGowan, J. et al., PLOS Genetics 19(10): e1010913, DOI 10.1371/journal.pgen.1010913, CC BY — abstract read at Crossref, paper not read; wire ScienceDaily, 7 October 2026; lag 1,098 days.*

Held — the headline says nature, the paper says a ruthenium model complex

### Scientists finally see how nature moves energy so efficiently

The measurement is a first and the abstract is honest about what it is. Femtosecond optical spectroscopy, nitrogen K-edge X-ray absorption and time-resolved X-ray solution scattering at the Linac Coherent Light Source, combined, following one sequential proton-coupled electron transfer in water: the electron redistributing on photoexcitation, a proton arriving at a ligand nitrogen at about **460 picoseconds**, and the first solvation shell switching its hydrogen bonds from N···HO to NH···O at the same time. **The object being measured is a synthetic ruthenium polypyridyl model complex**, named as such in the abstract’s own methods sentence; photosynthesis appears in the paper where applications are listed, as “artificial photosynthesis.” The release concedes the point in passing — “For the particular experimental system examined in this study, researchers already understand the basic mechanism” — and then the headline puts the result inside nature, where no measurement in this work was taken. We are not holding it because the chemistry is in doubt; we are holding it because **the sentence we would have to write to file it is the headline’s and not the paper’s**, and a page that files this one has no standing to complain about the next one. Lead image credited AI/ScienceDaily.com. *Kahraman, A. et al., Nature Communications 17(1), DOI 10.1038/s41467-026-75943-4, CC BY, online 26 August 2026 — read in full at PubMed Central (PMC13518934); wire ScienceDaily, 4 October 2026; lag 39 days.*

Held — the headline finding was published in 2020 and this paper adds one specimen to it

### Scientists found Homo erectus 200,000 years earlier than expected

What this paper actually is: a descriptive catalogue of **18 cranial and mandibular hominin fossils from Drimolen Main Quarry, excavated between 1992 and 2019 and never described** — some of them sitting undescribed for thirty years, which is the unglamorous labour the whole record runs on. One is assigned to *Homo erectus*, fourteen to *Paranthropus robustus*, three to indeterminate hominins, and the authors identify a new mandibular trait that may further separate the Drimolen *P. robustus* population from two others. Its own abstract says it sets out to test **“existing hypotheses concerning the first appearance of *H. erectus* at DMQ ~2.0 Ma”**, and its conclusion is that one specimen, DNH 127, “provides **additional** evidence for the presence of *H. erectus* in South Africa at ~2.04–1.95 Ma.” That date and that claim are Herries *et al.*, *Science*, 3 April 2020 (DOI 10.1126/science.aaw7293). **The headline is a six-year-old finding carried as this week’s news**, and the thing the paper is for — getting fossils out of a drawer and into the literature where other people can argue with them — is not mentioned. *Martin, J. M., Leece, A. B., Baker, S., Herries, A. I. R. & Strait, D. S., Annals of Human Biology 53(1), DOI 10.1080/03014460.2026.2725046, CC BY, online 25 September 2026; wire ScienceDaily, 7 October 2026; lag 12 days.* **Not read:** the paper. www.tandfonline.com is refused at the network layer and there is no PMCID; the structured abstract above came from Semantic Scholar’s abstract field — a third-party index’s copy of a publisher deposit, and *not* the machine-written tldr field this page refused to use last week, which is null for this paper.

Held — three lions, and no text of the paper is reachable from here

### Scientists changed how zoo lions eat. Their wild side came out

**This is the shape of finding this site exists for, and we still cannot file it.** At Monarto Safari Park, whole carcasses were added every nine days to an existing feast-and-fast routine for three adult male African lions, and eight infrared cameras recorded them continuously for three months. On carcass days the lions reportedly spent nearly three times as long eating, fed in longer bouts, rested far more afterwards, and grew steadily more active as the fast went on — a rhythm matching wild lions — while pacing tracked the routine rather than hunger. The behaviour that had been missing was never missing from the animal; it was missing from the schedule, and changing the schedule gave it back. That is the environment rung, in a species, with video. **What stops it: three lions.** And no version of the paper is reachable — www.sciencedirect.com is refused at the network layer, Crossref deposits no abstract, Semantic Scholar returns "abstract": null, and there is no PubMed Central copy. *Every sentence above is Adelaide University’s release, and none of it has been checked against the paper.* The lag cannot be stated either: Crossref deposits no online-first date, its record was created on 11 May 2026 and Semantic Scholar dates the paper 1 May, which puts the interval somewhere around 147 to 157 days, and this page is not going to pick one of them. *Hryhorenko, L. et al., Applied Animal Behaviour Science 301, 107041, DOI 10.1016/j.applanim.2026.107041 — the record lists CC BY; wire ScienceDaily, 5 October 2026.*

Held — the feed carried no journal reference at all, and the frame is the authors’ newspaper argument

### A forgotten 30-year forest experiment just delivered a huge surprise

ScienceDaily carried a piece from *The Conversation*, written by the researchers, about a silvicultural experiment established at Tanjil Bren and Cabbage Tree Creek in Victoria between 1987 and 1989 and revisited thirty years later. The reported result is good: in the heavily thinned stands the retained trees are enormous — the largest averaging well over a metre in diameter at eighty years old, 20% larger than in unharvested controls — and the thinned stands are said to hold as much carbon as the controls or more, having recovered everything the harvest removed. **The item carries no “Journal Reference” block**, the one every other story this week has, so a reader of the feed is given no citation at all. We found one anyway, by the route this page recommends — it is inside the piece, as the link under the words “our new research”: Hammond, Nitschke, Trouvé & Baker, “Alternative Silvicultural Systems Maintain Multiple Values in Tall Eucalypt Forest,” *Forest Science*, online 1 July 2026, DOI 10.1007/s44391-026-00076-6, which compares seven treatments across 53 plots and finds three distinct treatment groups at thirty years. **Three things still hold it.** The deposited abstract is truncated before it reaches carbon, and we could not read the paper. The piece’s conclusion — “win-win-win”, “the false dichotomy of conservation versus production” — is an argument the authors are making in a newspaper rather than a result, and the link under that phrase points at a *different* paper by other people about the Pacific Northwest. And a finding about logging, written by its own authors for a public policy fight, is the one kind of piece where the citation mattering most is the one the feed left out. *Wire ScienceDaily, 1 October 2026; lag 92 days.*

## Seedbed

Three seeds, all three off papers read end to end this week, and one of them arrives with two independent cases already attached. Plus the ruling this page promised last week on its lead list, which had grown for three editions running without anything coming off it.

Zine · Star Stuff, and it has two cases in one week

### An instrument that answers confidently in three directions

The coral filed above was placed in **three different families by three markers**, two of them from the same mitochondrion, with ultrafast bootstrap support of 100% behind the one that is certainly wrong — it put an animal with a proteinaceous axis into a family of corals that have no axis. The paper names the consequence: a single marker would have produced “a spurious assignment to a family with which it shares few if any diagnostic morphological characters.” **The second case is the calibration.** A date for the origin of animals is pinned to a deposit where nobody found any, and the test of that reasoning is a second deposit, just as well preserved, where nobody found any either — at a time when they certainly existed. In both, the number reported is a property of the instrument, and in both the discipline caught it by *running the instrument twice and letting the two answers disagree in print*. The piece is a Star Stuff zine: one settled fact, followed until the belonging claim is inside it, and the fact is that **confidence is produced by the method, not by the world.** It needs no human example on the page at all, which is why it would work.

From the *ZooKeys* and *Science Advances* papers filed above, both read in full at PubMed Central. **Not obtained:** the coral’s Supplementary material 1 and the *Science Advances* Supplementary Materials, which hold the analyses behind both numbers.

Zine · Kin, and the argument arrives from outside

### The ranking was the error, and a primatologist said so first

This site has argued for a year that a spiky profile cannot be collapsed onto a line, and it has argued it about people, from Damian Milton building on Francesca Happé, and drawn it on the face of [a broadside](https://starstuff.earth/not-a-line-broadside.html). **Here is the same argument, built out of a hundred primate species and a null result, by people who have never heard of us.** Five kinds of aggression; the lethal ones correlate moderately with each other; the mild ones correlate with nothing; and therefore, in the corresponding author’s own words, it is “biologically inaccurate to rank species on the basis of their general propensity for aggression.” A ranking needs one dimension. The paper found five that will not fold into one — and it refuses both answers it was handed, including the flattering one. **That refusal is the piece**, in [Kin](https://starstuff.earth/collection-kin.html): citation-dense natural history with a neurodivergence moral, where the moral is not *we are better than they said* but *the scale was never measuring anything.*

From the *Evolution Letters* paper filed above, read in full at PubMed Central, plus the University of Lincoln release for the quoted sentence. **Not obtained:** the supplementary file with the four alternative scoring datasets; Gómez *et al.* (2021), which this paper contradicts and which would have to be read before the contradiction is put on a page.

Field guide entry · Symbioses, off a paper read in full

### The third party at a two-party arrangement

A *Macaranga* tree grows hollow stems and food bodies; *Crematogaster* ants live in the stems and defend the tree. A wasp with no part in any of it hollows out the same stems and raises its larvae there on flies. Not a parasite of either partner — **a third party helping itself to the traded good.** The Symbioses guide has mutualisms and it has parasitisms; it does not have this, and cheater, exploiter and free-rider return zero across the whole site. The entry writes itself with two columns the guide already uses, and it carries a condition rather than a villain: the wasp is twelve times out of forty-three in oil palm and once out of forty-one in logged forest, because uniformity suits a generalist. **And the entry has to carry the null** — the plants lost no measurable leaf area, and the paper says the causal direction is unclear. An entry that leaves that out is a better story and a worse guide.

From the *PeerJ* paper filed above, read in full at PubMed Central. **Not obtained:** Tables S1–S4 and Fig. S1; Houadria *et al.* (2020), the earlier study of the same plots that the discussion leans on for ant occupancy.

Leads · the list is closed, as promised

### One worked, two closed unworked, and no new list

On 2 October this page wrote that its lead list, which had grown for three consecutive weeks with nothing coming off it, would this week be *either worked or closed*. **One was worked.** The *Geology* paper on the Zaonega Formation has a publisher-deposited abstract at Crossref and we read it: it reports the first compound-specific carbon and hydrogen isotope compositions of hydrocarbons and carbon dioxide trapped in fluid inclusions through a 500-metre core, finds a **thermogenic** origin for the C2–C4 gases, and derives an apparent temperature gradient running from 72 to 350 degrees Celsius — hottest low in the core beside a gabbro sill and declining upward. That is a local, syndepositional explanation for a carbon signature “which many have interpreted as reflecting global-scale processes.” It is now a lead with a source rather than a title, and it is a good candidate for the chain collection. **The other two are closed unworked and named as such.** Zhang & Dumberry in *Nature* still has *no abstract deposited anywhere we can reach*, which is the identical finding as last week and is the entire reason it never moved; the *Astronomical Journal* item was never opened. **And there is no new list.** Several things went past this week unopened; writing them down would start the same pile again, and a list that only grows is a way of looking thorough while deferring the work. *If an item is worth a line here next week, it will be worth finding again.*

Thiagarajan, N., Lepland, A., Eichinger, F., Prave, A. & Eiler, J., “Paleoproterozoic thermogenic hydrocarbon generation, Zaonega Formation, Russia,” *Geology* **54**(10), 1015–1020, DOI 10.1130/G54338.1, online 13 July 2026 — publisher-deposited abstract read at Crossref; paper not read. Zhang, H. & Dumberry, M., *Nature* **658**(8134), 119–124, DOI 10.1038/s41586-026-10999-2 — no abstract deposited at Crossref, second week confirmed.

## Not

**Not a citation.** Five papers were read end to end this week and the standing rule does not move. Nothing filed here may be promoted to a zine, a [Glimmers](https://starstuff.earth/collection-glimmers.html) entry or a card without the primary source in hand, a [changelog](https://starstuff.earth/changelog.html) entry and its row in the fact-check ledger. The *Science* item rests on a publisher’s deposited abstract and nothing else, and lifting a figure out of it would be laundering a deposit into a fact — which is easy to do, because a structured abstract looks exactly like a result.

**Not a picture of a finding.** Fourteen of this week’s seventy-six science-news items lead with an image a machine made, and thirty-seven more with a stock photograph. Every one of those credit lines is honest, and we are not accusing anybody of deception. *But a reader scrolling a feed meets the picture before the paper*, and this page is not going to reproduce one or describe one as though it showed something. Last week the thing in the abstract slot was written by a model; this week the thing in the photograph slot was drawn by one. Neither is a source, and neither announces itself as anything but ordinary.

**Not “science says you are valid.”** A hundred-species study finds that mild aggression and killing are not the same trait, and it would be very easy to walk off with the comfortable half. *We are not going to* — because the paper’s own sentence refuses “indiscriminately violent” *and* “peaceful” in the same breath, and a page that accepts a comparative method only when it flatters has no method. The belonging claim was never waiting on a correlation coefficient and does not move when the next analysis finds one.

**Not a deficit, and not a superpower either.** A coral that three genes could not agree on, a wasp that does well in a plantation, a thousand caesium atoms bound in tubes: these are facts about *architecture, history and the conditions things are kept in*, never about merit. The coral is not more interesting than its relatives; the markers were less reliable than everyone assumed. Flipping a deficit into a gift keeps the ranking and changes the sign, which is the one move the primate paper above says the data will not support.

**Not anti-science.** Four corrections on this page are of positions a field held for years — that a single barcoding gene can place a family, that an exceptional fossil deposit without animals dates the absence of animals, that mild aggression escalates into killing, that a bacterium’s immune trigger must be the intruder’s identity. Every one was overturned from the inside, with instruments the discipline built on purpose: three markers run against each other, a second biota on another continent, five traits scored apart instead of summed, a protease reconstituted in a tube. *None of those corrections is an argument against evidence; each is an argument for the part that was missing.* If a draft off this digest starts reading as a case against the field, it has gone wrong.

**Not *because of*.** A coral grows out of a field of brittle stars at four hundred metres in water with almost no oxygen in it, and somebody named it for the stars rather than for themselves. That is a rhyme, not a proof. The universe declining to pathologize its own variation is why the phrase carries — *down to your star stuff* measures how far the loving reaches, never why it is owed.

[Stimpunks Foundation](https://stimpunks.org/) × [More Realms](https://morerealms.com/) · Glimmer Wire · L★S
 Edition Seven of the weekly scan · published 9 October 2026
 Every edition is kept on its own page · nothing deleted, the Held list included
 A working scan, not a finding · off the reading chain, on purpose
 [All editions](https://starstuff.earth/collection-glimmer-wire.html) · [Glimmers](https://starstuff.earth/collection-glimmers.html) · [Changelog](https://starstuff.earth/changelog.html) · [Too Good to Check](https://starstuff.earth/too-good-to-check.html)
 stimpunks.org · morerealms.com · starstuff.earth · [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)