3
[2026-08-16] · src Alkalay, Park, Oh, Hong, Lee, Tiwari, Senapati, Umansky, Heiblum, Sim; arXiv:2608.12897 (2026)
Braiding an Abelian anyon around another just multiplies its wavefunction by a phase; braiding a non-Abelian one instead applies a unitary rotation within a degenerate ground-state manifold - the operation topological quantum computing wants to use as a noise-resistant logic gate. Direct evidence for that non-Abelian case has been elusive. This Weizmann-led group works in the nu=5/2 fractional quantum Hall state, a leading non-Abelian candidate, using a highly dilute one-dimensional edge channel and weak partitioning to separately measure the current-noise statistics of the downstream charged mode and the upstream neutral mode. The time-domain structure of that noise carries the signature expected of a non-Abelian braiding operation rather than a simple Abelian phase.
This is a noise measurement consistent with non-Abelian statistics, not a demonstration of a working topological qubit - no gate has been run on stored quantum information here. But nu=5/2 is the state most topological quantum computing proposals are built on, and clean evidence that braiding there really does act as a unitary rather than a phase is the load-bearing experimental claim under the whole approach. Independent reproduction, and a demonstration that the effect survives an actual braiding protocol rather than a noise correlator, are the next things to watch for.
[2026-08-16] · src Bhukta et al. (Kläui group, JGU Mainz); Nature Physics (2026), doi:10.1038/s41567-026-03383-4
Antiferromagnetic skyrmions are predicted to move without the skyrmion Hall effect that deflects their ferromagnetic counterparts sideways - useful, but until now inferred only from transport and static imaging, never watched directly. Using pump-probe scanning transmission X-ray microscopy at BESSY II's MAXYMUS endstation, this team filmed a dense lattice of current-driven antiferromagnetic skyrmions at 20 nm spatial and nanosecond temporal resolution. The skyrmions do move straight along the current direction, confirming the predicted absence of the Hall deflection; individual collision events show an incoherent regime (mobile skyrmions compress against pinned ones, then rebound) and a coherent regime where the whole lattice translates uniformly. Fitting the rebound trajectories to the Thiele equation extracts an exponentially decaying inter-skyrmion repulsion with a 30 nm decay length, matching micromagnetic simulation.
Direct imaging turns a predicted interaction law into a measured one, which is what is needed to actually engineer skyrmion spacing and density for racetrack-style memory built on antiferromagnets rather than ferromagnets. The absence of the skyrmion Hall effect is confirmed dynamically here, not just inferred - the open question is whether the same repulsion law and packing hold in thinner films and at the higher current densities a real device would need.
[2026-08-16] · src Regnault, Li, Kwan, Bernevig, Herzog-Arbeitman; arXiv:2608.12452 (2026)
Rhombohedral multilayer graphene on hBN hosts fractional Chern insulators (FCIs) - lattice analogues of the fractional quantum Hall effect, at zero magnetic field - but why a moire potential this weak is enough to stabilize them has been unclear, since standard continuum models of the conduction bands alone don't produce a large enough gap. This theory paper identifies the missing piece: filled valence bands develop their own moire-periodic charge-density modulation, which electrostatically imprints an effective potential onto the conduction bands - the "moire capacitor effect" - on top of whatever the substrate directly provides. Including it is what lets first-principles calculations reproduce a stable Chern number C=1 parent state at filling nu=1, matching what's seen experimentally.
This resolves a real gap between theory and experiment rather than reporting a new observation - it explains why FCIs in these devices are as robust as they are, which matters for engineering them by design (layer number, displacement field, substrate choice) instead of by trial and error. It's a mechanism paper: the FCI states themselves were already known, and the test is whether this effect correctly predicts stability across the other rhombohedral graphene systems where FCIs have and haven't been seen.
2
[2026-08-16] · src —
Filed after this run's learning pass, so the daily drain answers it directly rather than leaving it for tomorrow.
Taking the literal reading, with the reading stated because the question gives
no location: "where" is answered as the Moon's phase, since that is the one
part of "where is the moon" that does not depend on where you are standing.
Orrery's own name is a false lead here — this system is a research and notes
dashboard, not a live astronomical instrument, and nothing in build/ tracks
real celestial bodies.
Working from the synodic month (29.530588 days) and the commonly-used reference new moon of 2000-01-06 18:14 UTC, today (2026-08-16) is about 3.2 days past new moon: a waxing crescent, roughly 11% illuminated. That is an arithmetic approximation, not an ephemeris lookup — the real lunar month varies by several hours around that average, so treat "3.2 days" as accurate to within about half a day, not to the minute.
That answers phase, not a position in the sky. Altitude and azimuth also need
an observer's location and time of day, neither of which the question gave;
file a new ask with those if that is what was actually wanted.
[2026-08-16] · src build/orrery-render-html.el
orrery-render-html--escape (orrery-render-html.el:180-185) walks a fixed
alist of four pairs — ("&" . "&"), ("<" . "<"), (">" . ">"),
("\"" . """) — and for each one calls:
(replace-regexp-in-string (regexp-quote (car pair)) (cdr pair) out t t)The two trailing t arguments are FIXEDCASE and LITERAL. Neither is
decoration.
By default (both nil), replace-regexp-in-string's replacement string is
a tiny substitution language, not literal text: a backslash in REP is
special — \& means "insert the whole match", \N means "insert what
matched the Nth group", \\ means "insert one backslash". LITERAL t
switches that off and inserts REP exactly as written, character for
character. FIXEDCASE t switches off a different default behaviour —
replace-match otherwise tries to transfer the matched text's
capitalisation onto the replacement (so replacing "Foo" would capitalise
the replacement too) — which only ever applies to alphabetic matches.
Concretely, for the four pairs actually in this alist, dropping t t
would not change today's output: none of &, <, >,
" contains a backslash, and none of &, <, >, \" is a letter
with a case to transfer. But that is incidental, not a guarantee, and it
is exactly the kind of incidental safety that stops being true the moment
someone edits the alist. Verify the actual mechanism with two standalone
calls that stand in for what a fifth entry could do:
emacs -Q --batch --eval \
"(princ (replace-regexp-in-string \"&\" \"[\\\\&]\" \"A & B\" t t))"
emacs -Q --batch --eval \
"(princ (replace-regexp-in-string \"&\" \"[\\\\&]\" \"A & B\" t nil))"The replacement string in both is the four characters [\&]. With
LITERAL t it is inserted exactly as written: A [\&] B. With LITERAL
nil (the default), Elisp reads the \& inside it as "splice in the
matched text" — so the \& is replaced by the & that was actually
matched, giving A [&] B instead. Same REP string, same match, two
different documents, and the only thing that changed is the flag.
This is CLAUDE.md's stated security boundary, not just a code-quality
nicety: "Agent-written text reaches the page through
orrery-render-html--inline, which escapes before adding markup. That
ordering is the security boundary." --inline (orrery-render-html.el:216-232)
calls --escape first, then lifts verbatim spans and converts Org emphasis
into <strong><em><code> tags — so every angle bracket in the final
HTML was put there by this file, never by whatever an agent (or a filed
GitHub issue, the least trusted input in the system — see the comment at
orrery-render-html.el:421-423) wrote into a feed item's summary. If
--escape treated its replacement strings as a substitution language
instead of literal text, that would be a bug in the escaper itself, not
in its caller — and a bug an attacker doesn't need to trigger, because
LITERAL's default is already wrong for this job. Passing t t is what
makes "escape these four characters" actually mean only that.
Exercise (verifiable): run the two emacs --batch calls above and
confirm the two outputs differ exactly as described. Then explain, in one
sentence, why (regexp-quote (car pair)) on the search side has nothing
to do with this — regexp-quote only protects the REGEXP argument from
being read as a pattern; the t t flags are the independent, and
separate, protection for the REP argument.
2
[2026-08-16] · src Arban, Complete Conservatory Method — Syncopation
First Studies and Scales & intervals both asked for accuracy on notes that land on the beat. Syncopation asks the opposite question: can the same evenness of tone survive on notes that deliberately don't. It's the section most players skip because it reads like a rhythm exercise rather than a sound exercise, but a syncopated note that isn't given the same commitment as an on-beat one just sounds tentative, not "off the beat."
5 minutes: single held tones on C, subdividing a slow quarter=60 into eighth notes by tapping the foot, mf, no tonguing changes — just confirming the subdivision is felt in the body before it's asked to show up as syncopation in the horn.
The opening syncopation studies in that section of Arban, quarter=76 with the metronome clicking straight quarters (not subdividing for you). Success criterion: the off-beat entrances land with the same core of tone as the on-beat ones around them — no softening or ghosting the syncopated note out of uncertainty about where it falls. If the click and the syncopation start to blur together, the tempo is too fast for the ear to confirm the placement and should come back down.
No tonguing-speed or high-register material today — the syncopation studies at this tempo and range don't call for either, so the two don't collide.
Art of Phrasing, Melody No. 3. Play it through once for notes, then again listening specifically for any syncopated figures in the melody itself and giving those exactly the warmup's treatment: full core of tone, no ghosting, placed with the same confidence as the notes on the beat around them.
Stop the moment the syncopated notes start sounding breathy or under-supported compared to the on-beat notes around them — that thinning is the tell that the embouchure is guessing at placement rather than committing to it, and more reps past that point just rehearse the guess.
[2026-08-16] · src —
hearDizzy Gillespie — Groovin' High (Dizzy Gillespie & Charlie Parker Vol. 1 (1945)) · Tidal · Spotify
The head is built almost entirely from eighth-note displacement — accents landing where the ear expects rest and vice versa — played with total commitment to every syncopated entrance rather than any hesitation around them. Listen for how solid each off-beat note sounds compared to the beat it's displaced from, which is exactly what today's technical target is chasing at a much slower tempo.