3
[2026-09-21] · src LHCb Collaboration; arXiv:2609.21921 (2026-09-18)
Doubly heavy baryons - two charm quarks bound with a single light quark - are a direct test of QCD's static quark-quark potential in a regime lattice calculations can also reach, but only one such state, Xi_cc++, had a confirmed observation before now. LHCb searched for the strange-flavored partner Omega_cc+ (css) in the Omega_c0 pi+ channel, reconstructing the Omega_c0 in its pK-K-pi+ final state, using 6.3 fb-1 of pp collisions at 13.6 TeV collected with the upgraded detector in 2024. A peaking structure at 8.7 sigma global significance is consistent with a weakly decaying particle and identified as Omega_cc+, with mass 3725.9 +/- 1.0 (stat) +/- 0.2 (syst) MeV/c2 plus two further systematics from the assumed lifetime (+/- 0.4 MeV/c2) and the external mass calibration (+/- 0.6 MeV/c2).
This is a first observation, not a confirmation of an earlier hint - doubly charmed baryons have exactly one other member with an established mass, so Omega_cc+ roughly doubles the data available to whatever quark-model or lattice-QCD calculation predicts this spectrum. The mass sits close to several pre-registered lattice and potential-model predictions, which is itself the test: a doubly heavy baryon's mass is set almost entirely by the charm-charm interaction plus a light-quark correction, so agreement or disagreement here is a direct statement about how well those calculations handle two heavy quarks bound to one light one, independent of anything about mesons.
[2026-09-21] · src Belle II Collaboration; arXiv:2609.21947 (2026-09-18)
R(D*) and R(D) compare the rate of B-meson decays to a tau lepton against decays to a light lepton (electron or muon); the Standard Model predicts both ratios almost independent of the hadronic form factors that otherwise plague this measurement, so a deviation would point at new physics coupling preferentially to the third generation. Belle II measured both ratios with hadronic tagging - reconstructing the whole recoiling B meson rather than inferring it, which cuts background at the cost of statistics - in 387x106 Upsilon(4S) decays, obtaining R(D*) = 0.242 +/- 0.019 (stat) +/- 0.016 (syst) and R(D) = 0.439 +/- 0.055 (stat) +/- 0.046 (syst). Both are stated as consistent with the Standard Model and with previous measurements, and this is now the most precise hadronic-tag determination of either ratio.
The older world average of these ratios sat mildly above the Standard Model, which is most of why this pair of numbers has been watched for a decade as a possible lepton-flavor-universality violation. A tighter, SM-consistent result from the highest-precision tagging method does not rule out new physics outright - the anomaly was never large enough for that - but it removes weight from the hypothesis with every new measurement that lands on the SM side, which is the slow way most anomalies in this field have actually resolved historically.
[2026-09-21] · src Filipello, Beordo, Akhmetov, Crosta, Lattanzi; arXiv:2609.21606 (2026-09-18)
Pulsar timing arrays probe continuous gravitational waves down to roughly the nanohertz band; below that, toward the picohertz regime where the earliest, widest-separation inspiral of supermassive black hole binaries would show up, neither pulsar timing nor astrometry alone has enough sensitivity. Filipello et al. combine the two for the first time: apparent proper-motion displacements of ~1.5x106 quasars from Gaia's CRF3 catalog, and the orbital-period derivatives of 11 high-precision binary pulsars, in a Bayesian framework that models Galactic-potential uncertainty for the pulsars via Monte Carlo and a vector-spherical-harmonics decomposition (to octupole order) for the quasars to suppress spurious signals. No signal is found (ln B_joint = -0.42 +/- 0.03); the resulting 95% upper limit on strain is h0 <= 6.4e-11 at a reference frequency of 4e-10 Hz, 20-30% tighter than either dataset alone and with full sky coverage, breaking the spatial degeneracies that limit either method in isolation.
This opens an observational window - the picohertz band, between what pulsar timing arrays and cosmological probes each reach - that had no joint-method constraint before, by combining two datasets whose systematics are close to independent rather than by building new instruments. The bound itself is a null result, but the authors' injection-recovery forecasts say the same framework run against Gaia DR4's longer baseline should break the h0 < 1e-11 barrier, which makes this a method paper as much as a measurement: the next data release, not a new telescope, is what would turn this into a detection or a much tighter exclusion.
2
[2026-09-21] · src Arban, Complete Conservatory Method — 14 Characteristic Studies
Trumpet takes today's slot: three days since its last session on the 18th is the longest real gap of the three instruments right now, with piano having gone yesterday and alto the day before, continuing the alto/piano/trumpet cycle this rotation has held since mid-September. No. 10 opens with the same page of continuous Allegro sixteenth-note passagework as No. 9, in the same key, but it is built to be played twice: a rall. and a D.C. send it back to the top, and a "Fine" sits only two bars into its Più lento episode. The first time through, that two-bar phrase is a passing contrast on the way to more passagework; the second time, after the D.C., it's where the piece actually ends. Same notes, different job — and shaping that difference on command is today's real target, not just getting through the page clean.
5 minutes: the one-octave B-flat major scale in continuous sixteenths, quarter = 96, tongued lightly and evenly, same as this key's last warmup. Then play the tonic triad, B♭–D–F–B♭, slurred in one breath ascending and again descending, at a much slower tempo, leaning into the top note of each arc before settling out of it — a preview of the arching three-note shapes the Più lento is built from.
The full page at quarter = 96, three consecutive passes. On the first two passes, play straight through the "Fine" marking to the final rall. and D.C., as the piece asks; on the third, actually take the D.C. back to the top and stop at "Fine" this time. Success criterion: a listener who only heard the two-bar phrase leading into "Fine" on pass one and again after the D.C. on pass three should be able to tell which is which — open and still going in the first case, settled and finished in the second — without being told a repeat happened.
Art of Phrasing No. 23, aimed squarely at its own final phrase: play it once straight through, then once more shaping the ending so it decays and settles rather than simply stopping on the last written note. Borrow whatever made "Fine" convincing on the study and spend it on the melody's actual close.
Stop the moment the sixteenth-note passagework starts to blur into an even smear instead of speaking as sixteen distinct notes per beat, or the Più lento's arching phrase stops bending and starts sounding metronomic — either is fatigue substituting for control, and a fourth pass built on top of it trains the wrong habit.
[2026-09-21] · src —
hearMaurice André — Trumpet Concerto in D Major, G. 28 "Estienne Roger 188": I. Allegro · Tidal · Spotify
Today's technical target is checking for sixteenth-note evenness held across three unbroken passes of one page; a Baroque trumpet concerto's opening movement asks a soloist for the same evenness across several unbroken minutes against a steady orchestral pulse, with no rest built in. André's recording of Torelli's D major concerto, with Charles Mackerras and the English Chamber Orchestra, is one of the most-played accounts of exactly that kind of writing. Listen for how little any single note in a run stands out from its neighbors, no matter how far into a phrase it falls — that evenness under length is the harder version of what today's own success criterion is listening for.
4
[2026-09-21] · src U.S. District Court for the District of Columbia, preliminary injunction (Sept. 13, 2026), Democratic Senatorial Campaign Committee, League of United Latin American Citizens, and NAACP v. USPS
Late on September 13, Judge Carl Nichols (D.D.C.) granted a preliminary injunction barring USPS from "implementing and enforcing" the mail-ballot rule in full — categorically, not provision by provision. In a 24-page order, Nichols wrote that "no statute grants the Postal Service the power to issue key parts of the Rule," finding the Democratic groups likely to succeed on an ultra vires theory: that USPS lacks statutory authority to condition ballot delivery on state enrollment in its verification portal at all. He separately found that states voluntarily complying with the rule ahead of a final ruling would "create even more confusion and misunderstanding." This is a different theory from Judge Talwani's injunction (D. Mass., covered here September 7 and 15), which rested on the APA's arbitrary-and-capricious standard and the equities of imposing new rules this close to an election — the theory Justice Kavanaugh's September 14 concurrence at the Supreme Court suggested might eventually let USPS prevail on the merits.
This is a preliminary-injunction ruling, not a final judgment; the merits of USPS's statutory authority continue to be litigated in this suit. No appellate court has yet ruled on the ultra vires theory itself — the Supreme Court's September 14 order addressed only the Talwani injunction's stay standard, not this one, and Nichols's order was not before it.
Two independent federal injunctions, on two different legal theories, now block enforcement of the mail-ballot rule for the November 3 midterms. Even if the government's pending Supreme Court effort against the Talwani injunction eventually succeeds, Nichols's broader order — untouched by that appeal — would still bar the rule on its own. And unlike the timing-based objection Kavanaugh floated as surmountable in a future cycle, an ultra vires holding, if it survives appeal, would mean USPS needs new statutory authority from Congress before this rule can ever take effect, not just more lead time.
[2026-09-21] · src U.S. Court of Appeals for the Tenth Circuit, Nos. 25-1137, 25-1231 & 25-1299, Save the Colorado v. Graham (decided Sept. 15, 2026)
A three-judge panel (Tymkovich writing, joined by Chief Judge Holmes and Judge Moritz) reversed a 2024 district-court ruling that the Army Corps of Engineers acted arbitrarily and capriciously in permitting Denver Water's 131-foot raise of Gross Dam, which would roughly triple Gross Reservoir's capacity. The panel held the Corps' definition of the project's purpose, its method for comparing project alternatives on cost, and its treatment of climate-change projections for future streamflow all "fall easily within that zone of discretion" courts owe an agency — rejecting Save the Colorado's argument that the Corps ignored clear evidence on warming and precipitation trends. The case is remanded with instructions to enter judgment against the environmental groups, vacating the district court's 2025 injunction that had halted construction.
The panel did not rule on whether the reservoir expansion is good water policy, on the separate water-rights questions underlying Denver Water's claim to the additional storage, or on Colorado River Basin depletion more broadly — it reviewed only whether the Corps' permitting process was lawful under the Clean Water Act and NEPA's procedural requirements, and expressly declined to "substitute our judgment for the Corps's" on the science.
Denver Water can resume filling the reservoir immediately; major construction was already complete and had been waiting on the injunction. Plaintiff Save the Colorado has said it will not seek Supreme Court review, which — absent a rehearing petition — ends eight years of litigation and leaves the permit as issued.
[2026-09-21] · src U.S. Court of Appeals for the Seventh Circuit, No. 25-1354, United States v. Anderegg (decided Aug. 25, 2026)
Anderegg was charged with producing, distributing and transferring AI-generated child sexual abuse images (made with Stable Diffusion, depicting no real child) and with private possession of the same material, the last charged under 18 U.S.C. § 1466A(b)(1). This appeal reached only the possession count. Judge Lee, writing for a panel that also included Judges Pryor and Kolar, held that count unconstitutional as applied, affirming the district court: Stanley v. Georgia (1969) protects private in-home possession of even unprotected material from prosecution, and Ashcroft v. Free Speech Coalition (2002) already held that sexually explicit images of fictional minors, produced without using a real child, fall outside the narrower category of speech Ferber and Osborne permit the government to suppress. The panel distinguished those two precedents, which both involved actual child victims, as not controlling here. Lee and Kolar wrote separately that image-generation technology has advanced far past what the Court anticipated in 2002 and said they had "concerns about the lines these cases draw, but we are not free to redraw them ourselves," calling for Supreme Court reconsideration of virtual-CSAM doctrine in light of AI.
The production, distribution and transfer-to-a-minor charges against Anderegg were not before this panel and remain pending on remand. The ruling binds only the Seventh Circuit and does not require any other circuit to reach the same result — nor does it hold that all AI-generated material of this kind is protected; the panel's holding runs only to material that, "like Ashcroft's," depicts no actual child.
Prosecutors within the Seventh Circuit cannot rely on § 1466A(b)(1)'s private in-home-possession provision alone against AI-generated images of fictional minors while Stanley and Free Speech Coalition stand as written; production and distribution charges are unaffected and remain a live route to prosecution in cases like this one. Two of three panel judges have now asked the Supreme Court to take up the question, which is the clearest sign yet of where a future circuit split, or a grant of certiorari, could come from.
[2026-09-21] · src Federal Reserve, FOMC statement, September 16, 2026
At the conclusion of its September 15-16 meeting, the FOMC voted 12-0 to raise the target federal funds rate a quarter point, to 3-3/4 to 4 percent, effective immediately. The statement said economic activity "is expanding at a solid pace," with strong productivity growth and resilient spending, but that "inflation remains elevated," and that the increase would "support a timelier return to the Committee's 2 percent goal." The statement's language on financial conditions was unchanged from prior meetings, describing the Committee's ongoing practice of "maintaining ample reserves in the banking system."
The new target range takes effect immediately and feeds directly into consumer and business borrowing costs — credit cards, adjustable-rate loans, and anything else priced off the federal funds rate — seven weeks before the midterms. This was the meeting the panel's August PPI item (reported here September 15) flagged as the one where the Fed would have to weigh a large, energy-driven producer-price jump against its inflation target; a unanimous vote to raise rather than hold is the Committee's answer, though the statement itself does not commit to any further move at the next meeting.
2
[2026-09-21] · src LaTeX News Issue 44, draft dated 2026-09-15 (target release 2026-11-01); latex-base-dev pre-release 2, CTAN, 2026-09-16
LaTeX News 44 ships alongside a real pre-release, testable today via
pdflatex-dev: latex-base-dev pre-release 2. Its largest single piece is a
general order key for templates — a way to tell a template "typeset these
named parts, in this order, with these separators" instead of hard-coding the
sequence in the template's own code. Theorem-like environments and
\begin{proof} have already been rewritten onto it: the default order for a
theorem head is order{title,separator-1,number,separator-2,note,punct}=,
and turning "Lemma 3.2 (note)" into "3.2 Lemma (note)" is now a matter of
writing number before title in that list on the template instance,
nothing else in the document touched.
That reordering is not new by itself — thmtools' headformat key already
lets you do it today, in stable LaTeX, with \NAME, \NUMBER and \NOTE
placeholders standing in for the pieces. What the kernel version closes is
the gap thmtools cannot: a head built by reordering \NAME\NUMBER\NOTE
is boxes with no memory of which piece was which, so it produces a tagged PDF
with the wrong structure, or none. The order key lives inside the same
template machinery that emits the tag structure, so — the News' own words —
"tagging support is automatically provided" for any order you choose. A
document that needs both a house style's theorem order and an accessible,
correctly tagged PDF currently cannot have the first without breaking the
second.
Nobody sets a paper in latex-base-dev, so this changes nothing in a
document compiled today, and the release the draft targets is still six
weeks out. What it changes, once it ships, is narrower than "theorems are
now configurable" — they already were, via thmtools. It is the one case
where configurable and correctly tagged currently pull apart: a journal that
wants the number before the title and a document that also needs to pass as
accessible cannot have both today without hand-patching amsthm in a way the
tagging project has no way to understand. The order key removes that
specific trade-off at the source, for theorem heads and nothing wider than
that.
[2026-09-21] · src thmtools user guide (the =headformat= key); Typst documentation, =state= and =counter=
Set the same lemma two ways — "Lemma 3.2 (note)." and "3.2 Lemma (note)." — in LaTeX and in Typst, then flip the order by editing exactly one line.
LaTeX needs nothing beyond amsthm and thmtools, which ship with any
standard TeX Live or Overleaf install:
\usepackage{amsthm}
\usepackage{thmtools}
\declaretheoremstyle[headformat=\NAME\ \NUMBER\NOTE]{namefirst}
\declaretheorem[style=namefirst,name=Lemma]{lemma}
\begin{lemma}[note={see appendix}]
Every bounded sequence has a convergent subsequence.
\end{lemma}To flip to number-first, change one key: headformat\NUMBER\ \NAME\NOTE=.
The \begin{lemma} line is untouched.
Typst has no built-in theorem, so the environment is a four-line function written once:
#let lemma-count = counter("lemma")
#let lemma(note: none, body) = {
lemma-count.step()
block[*Lemma #lemma-count.display()*#if note != none [ (#note)]. #body]
}called as #lemma(note: "see appendix")[Every bounded sequence has a
convergent subsequence.] at each use site. To flip to number-first, swap the
two interpolations inside that one block[...] line —
#lemma-count.display() before *Lemma* — and every call site stays as
written.
Success criterion: after the one-line edit, recompile with no other change in
either engine, and every lemma in the document — however many there are —
changes order together. If one instance still reads the old way, the order
lived in the call site rather than the style, and the exercise didn't do what
it was supposed to. Under an hour: both halves use machinery already present
— amsthm and thmtools ship with any standard LaTeX install, the Typst
function needs no import — so there is no toolchain to install first.
The technique that transfers: put the thing a house style controls in exactly
one place, so changing venues is a one-line diff rather than a document-wide
search-and-replace that a script can't safely do, because "Lemma 3.2" and
"3.2 Lemma" aren't related by a regex — they're related by which template
rendered them. That is the same principle LaTeX News 44 is formalizing at the
kernel level for this exact case (see above); an hour with thmtools'
headformat today previews, in the tool already installed, what the order
key is set to make the default way to do it in November. A paper resubmitted
to a second journal with a different numbering convention is the case this
actually pays for.