1
[2026-08-31] · src Vijayadharan, Bolaños, Pompermaier, Bertapelle, Santagiustina, Agnesi, Vallone, Villoresi; arXiv:2608.28190 (2026)
A fiber-based platform encoding entanglement in time bins rather than polarization or spatial mode measured a CHSH violation of S = 2.583 +/- 0.002, exceeding the local-realist bound by more than 265 standard deviations, while simultaneously closing the locality, freedom-of-choice and postselection loopholes in the same run. The photon-source-to-analyzer separation was only 49.0 +/- 0.7 m - which the authors report as substantially shorter than earlier photonic Bell tests that addressed a comparable combination of spacetime constraints, because time-bin measurement normally requires an interferometer whose need to postselect on detection outcomes has kept genuinely loophole-free time-bin tests rare even though the encoding is the one already used to distribute entanglement over installed fiber.
Closing the locality loophole means the setting choice and outcome at each station must be spacelike-separated from the other station's outcome, which is normally what forces these experiments out to kilometer-scale separations - there has to be enough light-travel time between stations to fit a fast random setting choice and a full readout before a signal could cross. Getting that timing margin at 49 m rather than several kilometers, without giving up the postselection-free measurement that time-bin encoding otherwise costs, is what makes the result useful beyond foundations: it is in the size range of a metro fiber loop, not a dedicated long-haul link, which is the regime device-independent quantum key distribution and a future quantum-repeater network would actually have to run in. Nothing here is a new violation of local realism - that question has been closed since the 2015-2018 loophole-free tests - so the genuine step is the combination and the distance, not the physics.
2
[2026-08-31] · src Arban, Complete Conservatory Method — 14 Characteristic Studies
Trumpet takes today's slot — three days since its last session, the longest gap of the three instruments right now, continuing the alto → piano → trumpet three-day cycle this rotation has held since mid-August. Two sessions into the 14 Characteristic Studies, No. 1 asked for endurance and No. 2 for dynamic contrast; No. 3 asks for something neither of those touched — carrying a slurred phrase smoothly across a register break, so the connection is made with air and voicing rather than papered over with a soft tongue.
7 minutes: a slurred descending major scale from top-line F down two octaves, legato throughout with no tongue between any two notes, at mp, quarter=58, one full descent per breath. The point is a seamless connection at every step — including the break — before the study asks for the same thing inside a real phrase.
Characteristic Study No. 3, played straight through once, honoring every slur mark as written rather than smoothing a hard connection over with a quiet re-tongue. Success criterion: the longest slurred passage in the study, the one that crosses the register break, has no audible seam at the crossing — no click, no dip in pitch, no change in tone color going across.
Art of Phrasing, Melody No. 16. Play it once for notes, then once carrying any register shift in the melody with the same unbroken legato the technical target just asked for — the same demand, at a smaller scale and a slower tempo.
One clean pass at the Characteristic Study is the assignment, same as No. 1 and No. 2 — a second pass right behind it trains fatigue, not legato. Stop there, or sooner if a slur starts needing a tongue to rescue it; an embouchure that has to cheat the connection has nothing left to teach today's target.
[2026-08-31] · src —
hearClifford Brown — Where Or When (Clifford Brown With Strings) · Tidal · Spotify
Recorded in January 1955, this is Brown at his most exposed — a ballad with strings and nowhere to hide a seam. Listen for how a register shift in the middle of a phrase never announces itself: the tone doesn't thin, brighten or catch going across the break, which is exactly the join today's technical target is asking a single slurred passage to make.
4
[2026-08-31] · src Supreme Court of the United States (No. 26A124; No. 26A139 denied as moot)
In a per curiam order on the emergency docket, the Supreme Court stayed the injunction the District of Massachusetts had entered in the states' suit over Executive Order 14399 (91 Fed. Reg. 17125), pending the Government's appeal to the First Circuit and any subsequent petition for certiorari. The district court's final judgment had barred the Secretary of Homeland Security from compiling "State Citizenship Lists" (§2(a)), the Attorney General from prioritizing prosecution of officials over the order (§2(b)), and the Postal Service from beginning a ballot-rule rulemaking (§3(b)), as to the 23 states and D.C. that sued. The Court held the Government is likely to win on jurisdiction: each provision is an internal executive directive that, by its own terms, may be implemented only "to the extent feasible and consistent with applicable law," so any injury to the states rests on a chain of contingencies the Court called too speculative to support standing or ripeness — addressing §§2(a), 2(b), and 3(b) separately rather than treating the order as one unconstitutional package. Justices Sotomayor, Kagan, and Jackson dissented, in separate opinions.
The Court says so itself: "The Court's disposition of this application does not mean that any measure taken by the Government to implement the Order will necessarily be lawful... On that score, time will tell." Nothing here reaches the constitutionality of the Order or of anything DHS, DOJ, or USPS might eventually do under it — the ruling is only that these particular plaintiffs could not show, at this stage, an injury concrete enough to sue over yet. It also says nothing about a separate case, below, where a since-finalized USPS rule — a final agency action rather than an internal directive — is being challenged directly.
As of August 24, the Government may resume §§2(a), 2(b), and 3(b) as to these 23 states and D.C. while the appeal proceeds; the stay lifts automatically if certiorari is denied, and runs until this Court rules if certiorari is granted. It changes nothing for the separate, broader injunction Judge Talwani entered in League of Women Voters of Massachusetts v. Trump (No. 1:26-cv-11549, D. Mass.), which blocks the Postal Service nationwide and predates this stay — that injunction, and the new one below, are what actually govern USPS's conduct going into the midterms.
[2026-08-31] · src D. Mass. (No. 1:26-cv-13917)
On August 21 USPS completed rulemaking on "Ballot Mail for Federal Elections" and sent it to the Federal Register for publication; the Government had argued the rule fell outside the existing injunctions because it "applies only to future elections" after November 3. Twenty-three states, D.C., and Pennsylvania's governor sued the day the rule was scheduled to publish, in California v. United States Postal Service (No. 1:26-cv-13917, D. Mass.) — before Judge Indira Talwani, who also presides over League of Women Voters of Massachusetts v. Trump (No. 1:26-cv-11549), the suit whose nationwide injunction the new rule was meant to route around. On August 27 Talwani granted a 14-day temporary restraining order blocking the rule's mandatory ballot-envelope design standards, its requirement that states register in a USPS electronic portal and upload voter data, and USPS's claimed authority to refuse to mail noncompliant ballots — for elections on or before November 3,
She wrote that "the record continues to lack any evidence regarding
fraudulent absentee or mail-in voting," and that the government's interest in addressing an "unsubstantiated problem through likely unconstitutional means is dwarfed by the overwhelming risk of pervasive disenfranchisement." A hearing on a longer preliminary injunction is set for September 3.
The order is expressly temporary and time-limited; it is not a ruling that the rule is unlawful, only that the plaintiffs are likely enough to succeed and the harm of waiting is great enough to justify a 14-day pause. Nothing here is affected by the Supreme Court's stay two days earlier, above — that stay ran to a different docket (No. 1:26-cv-11581) brought over the Order's internal directives, not to USPS's now-final rule.
Until at least September 10, or sooner if Talwani rules first at the September 3 hearing, USPS may not enforce the envelope-design, portal-registration, data-upload, or ballot-refusal provisions of its rule against any state for an election held on or before November 3 — which covers the midterms. Nothing stops USPS from enforcing the rule for elections after that date, which is exactly the gap the Government argued let it proceed in the first place.
[2026-08-31] · src Ninth Circuit (No. 25-7516)
A unanimous three-judge panel (Nelson writing, Bade joining, Lee concurring) affirmed in part the District of Nevada's dissolution of KalshiEX's preliminary injunction against the Nevada Gaming Control Board's cease-and-desist order, which had told Kalshi to stop offering its sports-event contracts in the state. Kalshi's theory was that its contracts are federally regulated "swaps," traded on a CFTC-designated contract market, and that the Commodity Exchange Act's express preemption of state regulation of swaps "traded or executed" on a DCM therefore displaces Nevada's gaming law. The panel disagreed: under the CEA's own definition (7 U.S.C. §1a(47)(A)(ii)), a swap must turn on a "financial, economic, or commercial consequence," and the outcome of a sports contest is none of those — so Kalshi's sports contracts are not swaps at all, and neither express, conflict nor field preemption reaches them. The panel also rejected Kalshi's argument that a CFTC "Special Rule" provision (7 U.S.C. §7a-2(c)(5)(C)) gives the agency exclusive authority to decide whether gaming-linked contracts may be listed, noting that an existing CFTC regulation, 17 C.F.R. §40.11(a), already categorically bars them.
The panel did not rule on Kalshi's election-market contracts at all — it remanded Nevada's challenge to those specifically for the district court to consider "consistent with this opinion." Judge Lee's concurrence separately flags, without resolving, whether the Special Rule actually gives the CFTC discretion to permit gaming contracts it currently bars by regulation — a question the panel says it need not reach while 17 C.F.R. §40.11 stays in effect. And because this is review of a preliminary-injunction dissolution for abuse of discretion, the underlying suit continues in the district court; nothing here is a final judgment on the merits.
Nevada, and any other state within the Ninth Circuit, may now enforce its gaming law against Kalshi's sports-event contracts without a federal injunction standing in the way — but only sports contracts, and only in the nine states the Ninth Circuit covers; Kalshi's election contracts in Nevada remain unresolved, and the Third Circuit's contrary ruling for New Jersey means the same kind of contracts are federally protected there. That circuit split, not this opinion alone, is what points toward Supreme Court review.
[2026-08-31] · src S.D.N.Y. (No. 1:26-cv-00858)
Judge Jeannette Vargas ruled, in Catholic Legal Immigration Network (CLINIC) v. Rubio (No. 1:26-cv-00858, S.D.N.Y.), that the State Department's policy suspending immigrant-visa issuance to nationals of 75 countries exceeded the Secretary of State's statutory authority, and vacated it. The policy, in effect since January, directed consular officers to refuse immigrant visas to otherwise-eligible applicants from the listed countries without individual review. Vargas held that this predetermination of outcomes by nationality conflicts with the Immigration and Nationality Act's bar on nationality-based discrimination in immigrant-visa issuance and with consular officers' statutory discretion, and that the Administrative Procedure Act does not let an agency adopt a categorical rule where Congress required case-by-case adjudication.
The order does not disturb any visa refusal where a consular officer also identified a ground of ineligibility independent of the 75-country policy — only refusals based solely on that policy are set aside. It leaves the State Department's separate public-charge bond pilot program untouched, and it does not reach whether nationality-based screening could be structured lawfully in some narrower form; Vargas gave the parties until September 11 to propose how to handle the remaining questions in the case.
Consular officers processing immigrant visas for nationals of the 75 listed countries — including Afghanistan, Bangladesh, Brazil, Cuba, Haiti, Iran, Iraq, Nigeria, Syria, and Yemen — may no longer deny a visa on the sole ground that the applicant is from one of those countries; those applications go back to individualized review under the ordinary statutory criteria. No stay is in place, so the vacatur is in effect now, though the State Department has not said whether it will appeal.
2
[2026-08-31] · src LaTeX News, Issue 43 (June 2026)
LaTeX News 43, the release notes for the 2026-06-01 kernel, lists three
places "fake math" — using math mode purely for its built-in box-positioning
machinery — has been removed: \textsuperscript and \textsubscript,
\parbox and \minipage, and "around tabular structures." The reason is
the tagging project: math mode sitting inside what should be ordinary text
confuses the code that builds a PDF's accessibility tree, and it complicates
right-to-left text. TeX Live 2026's own release coverage names the casualty
directly: "With new changes to tabular structures in the LaTeX kernel, the
tabu and tabls packages are no longer supported in TL2026 — they will cause
compile errors now" — packages that patched tabular's column-width and
vertical-alignment limits by reaching into exactly the internals that just
moved. The same kernel release lands the first cut of the reciprocal change:
heading commands "re-implemented ... using the template mechanism," so a
legacy \section can be tagged without titlesec's patches — though
titlesec's own interface is not emulated yet: "Over the summer we intend to
add emulations for the titlesec package interfaces ... so stay tuned."
For anyone loading tabu or tabls for a wide or vertically-centred table — the exact "work around tabular's defaults" job tabu exists for — TL2026 does not warn, it refuses to compile: those two packages hit the moved internals directly, and both are declared unmaintained by their own repositories with no fix coming. The move is to the kernel's own facilities for that job, not to wait the package out. The heading side is the opposite shape: nothing breaks yet, because titlesec's actual macros are untouched, but its replacement was promised "over the summer," and summer is behind this date — worth checking whether that emulation shipped before reaching for titlesec on a new tagged document.
[2026-08-31] · src The Emacs Lisp Reference Manual, "Defining Macros"
Write the same two-branch conditional twice, once in TeX and once in Elisp, with a broken command sitting in the branch that never gets taken, and watch it stay silent until it does.
In plain TeX or a LaTeX preamble:
\def\empty{}
\def\myifempty#1#2#3{\def\tmp{#1}\ifx\tmp\empty #2\else #3\fi}Compile \myifempty{}{it was empty}{\undefinedcommand}: the output reads
"it was empty" and the log has no complaint, even though \undefinedcommand
is defined nowhere in the document — #3 was substituted as raw, unexpanded
tokens, and only the chosen branch, #2, was ever expanded. Compile
\myifempty{x}{it was empty}{\undefinedcommand} and watch the same document
fail with "Undefined control sequence."
In Elisp:
(defmacro my-if-empty (val then else)
`(if (string-empty-p ,val) ,then ,else))Evaluate (my-if-empty "" "'empty" (error "boom")) in M-x ielm: it returns
empty and (error "boom") never runs. Evaluate (my-if-empty "x" "'empty"
(error "boom")) and get the error. Then rewrite my-if-empty as an ordinary
defun with the same body and call it the same two ways: it errors on
both calls now, because a function evaluates every argument before its
body runs at all, where a macro's argument is still source, not yet a value,
until the expansion chooses to use it.
Success criterion: the defmacro/TeX-macro version raises the broken
branch's error only when that branch is the one taken, and swapping the
defun in for the defmacro is what turns the same broken content into an
error on every call — one run showing the boundary, not a claim about it.
This is the actual reason a LaTeX document can carry a stale
\iffalse ... \badcommand ... \fi block for years with nobody noticing it
is broken, and the same reason a package's compatibility shim wrapped in an
untaken \ifx branch can fail silently for months until the one class or
option that takes that branch shows up. It is also the one place this week
where "TeX's macro expansion and Lisp's evaluation model rhyme" is something
you ran rather than something asserted: both defer looking at a branch's
content until the branch is chosen, which is exactly the boundary that
matters when reading why a Lisp macro someone wrote is not simply a function
with different syntax.