3
[2026-08-13] · src Holm, Pearson, Nibauer, Sand, Price-Whelan, Starkenburg, Hendel, Fielder; Nature (2026), doi:10.1038/s41586-026-10878-w
Deep Hubble imaging of the ultra-diffuse galaxy UGC 9050-Dw1 (35.2+/-2 Mpc) turned up a faint, thin arc of stars trailing one of its globular clusters — the first globular-cluster tidal stream ever resolved outside the Milky Way. Streams like this had long been predicted around other galaxies but never directly imaged; the team modeled the host's gravitational potential from the stream's width and curvature, the same technique used on Milky Way streams like Palomar 5 to map our own halo. The fit favors a massive, extended dark halo and puts no meaningful upper bound on the progenitor's initial mass — consistent with ultra-diffuse galaxies being dark-matter-dominated rather than ordinary galaxies stripped down to unusually diffuse light.
Ultra-diffuse galaxies have been a real puzzle — some appear almost dark-matter-free, which is why it matters whether they're a distinct population or just ordinary galaxies with stretched-out light. Stellar streams are one of the few model-independent probes of a halo's mass and shape, but until now the technique was confined to the Milky Way and M31, where the deep imaging already exists. Extending it to an external UDG turns a single-galaxy method into a general one; the sharper test is whether it still works on the low-mass, apparently dark-matter-poor UDGs where the puzzle actually lives, not on one that behaves as expected.
[2026-08-13] · src arXiv:2608.11124 (Muon g-2 Collaboration)
Using the same storage-ring dataset that produced the collaboration's final anomalous-magnetic-moment result, but now tracking the average vertical decay angle of the muons' decay positrons — the signature of an electric dipole moment interacting with the ring's magnetic field — the Muon g-2 Collaboration measured d_mu = (-0.35 +/- 0.19_stat +/- 0.34_sys) x 10-19 e*cm, consistent with zero, and sets a new direct limit |d_mu| < 1.10 x 10-19 e*cm at 95% CL. That is roughly 1.6x tighter than the previous best direct limit, BNL E821's 1.8 x 10-19 e*cm, and uses only the first 25% of Fermilab's collected data.
A nonzero muon EDM would be unambiguous evidence of CP violation beyond the Standard Model, whose predicted value (~10-38 e*cm) is unmeasurably small — unlike g-2 itself, this channel isn't clouded by the ongoing lattice-QCD dispute over what the Standard Model even predicts, so any future signal here would be clean. This result is still consistent with zero and only a modest improvement on the old limit, not the roughly two-orders-of-magnitude gain the full dataset is eventually expected to deliver; the number worth watching is what the remaining 75% of the data does to this bound.
[2026-08-13] · src Wang et al.; Nature Physics (2026), doi:10.1038/s41567-026-03395-0
A Wigner crystal is the state electrons fall into when mutual Coulomb repulsion beats their kinetic energy and they freeze into a rigid lattice — predicted in 1934, and previously seen only statically in monolayer WSe2, via exciton diffraction off the crystal's fixed periodic potential (an "umklapp" signature). This group found a distinct set of resonances in cryogenic reflectance spectra of charge-tunable WSe2 — "Wigner polarons" — that form when an exciton locally distorts the electron lattice and hybridizes with the crystal's own collective vibrations, rather than just scattering off its static structure. The energy splitting between Wigner-polaron and ordinary attractive-polaron resonances scales as (electron density)^(1/4), tracking the density-dependence expected of the crystal's phonon-like excitation energy — and the whole effect shows up at zero magnetic field.
Detecting that a Wigner crystal exists is one thing; watching it move is another. Until now that meant transport measurements that need a magnetic field to stabilize the crystal and mostly see its edges, not its internal dynamics. An all-optical, zero-field probe of the crystal's own vibrational spectrum turns a hard-won detection into a tool — one that could track melting, defects, or coupling to other correlated phases in the same monolayer without a magnet in the way. Whether it generalizes to other Wigner-crystal-hosting moire and monolayer systems is the next thing to watch.
1
[2026-08-13] · src build/orrery-feed.el
orrery-feed-collect (orrery-feed.el:108-124) is where "the earliest sighting
of an id wins" — the rule README.org states in prose — actually gets enforced,
and it does it without ever comparing two timestamps:
(dolist (file (orrery-feed--fragment-files feed-dir))
(pcase-let ((`(,panel . ,items) (orrery-feed-read-fragment file)))
(dolist (item items)
(let ((id (plist-get item :id)))
(unless (gethash id seen)
(puthash id t seen)
(puthash panel (cons item (gethash panel by-panel)) by-panel))))))seen is a hash table used purely as a set: the value stored under each id
is always t, never read for its content, only checked with gethash for
presence. "First wins" isn't a rule this code evaluates — it falls straight
out of the fact that dolist visits files in a fixed order and the
unless (gethash id seen) guard only ever lets the first visit to a given
id through. That order is set two functions away, in
orrery-feed--fragment-files (orrery-feed.el:94-99): (sort
(directory-files-recursively feed-dir "\\.org\\'") #'string<). Sorting file
paths lexicographically is what makes this correct at all, and only because
data/feed/<YYYY-MM-DD>/panel.org is zero-padded ISO-8601 — plain string
comparison of \"2026-08-08/...\" against \"2026-08-12/...\" happens to
equal date comparison. orrery-feed--newest-first (101-106) makes the same
move for display order, sorting the deduplicated items within a panel by
:captured_at with string>, again relying on ISO-8601 being fixed-width.
The (pcase-let ((`(,panel . ,items) ...)) ...) line is the other idea worth
naming: orrery-feed-read-fragment returns a single cons, (PANEL-ID .
ITEMS), and pcase-let destructures it in one step with a backquote
pattern instead of (let* ((result ...) (panel (car result)) (items (cdr
result))) ...). The backtick pattern mirrors the shape of the value being
matched — a dotted pair — the same way it would mirror a longer list.
Every dedup guarantee this repo makes rests on file ordering, not on any
field inside an item. Reverse the sort in
orrery-feed--fragment-files and nothing errors: orrery-ir-validate
still passes, because a reversed feed is still a well-formed one, just a
wrong one — the newest fragment's copy of a re-reported id would win instead
of the oldest. That failure mode is silent by construction, which is exactly
why the ordering is asserted by test rather than left as an implicit
property of directory-files-recursively.
Exercise (verifiable): flip the comparator to (sort ... #'string>) in a
scratch copy of orrery-feed--fragment-files, then run make golden. It
will fail — read the diff make golden prints rather than guessing. Name
which item's :captured_at changed in the failing panel and confirm it's
the item whose id appears in more than one fragment across the five days of
feed history now in data/feed/. Restore the comparator before committing
anything; this is an exercise, not a change to ship.
2
[2026-08-13] · src Arban, Complete Conservatory Method — Scales & Intervals
First Studies on 2026-08-10 was about making every note match; today asks a harder question of the same tone — can it hold steady across a leap instead of a step. Intervals are the part of the method most players skip because scales feel more like "practising" and intervals feel like guessing at pitch.
5 minutes: long tones on C and G a fifth apart, alternating — whole note on C, whole note on G, no note in between — at mp, quarter=60. The only thing being checked is whether the embouchure resets cleanly on the jump or grabs for the new pitch with extra lip pressure.
The opening interval studies in Scales & Intervals: thirds first, ascending and descending through the C major scale, quarter=72. Success criterion: each interval lands on pitch on the first attack, with no scoop or slide up to the note — if the ear can hear the pitch arriving late, slow the tempo back down rather than pushing through it.
No tonguing-speed or high-register material today — interval work at this tempo and range doesn't touch either, so the two don't collide.
Art of Phrasing, Melody No. 2. Play it once straight for notes, then again picking out the largest leap in the melody and treating it the way the warmup treated C-to-G: same embouchure reset, same clean attack, just now inside a phrase instead of a drill.
Stop the moment an interval starts getting found by ear instead of landing directly — a slide, scoop, or brief wrong note before the right one arrives. That's the tell that fatigue is making the lip search for pitch rather than commit to it, and no amount of extra reps at that point trains the interval; it just trains the search.
[2026-08-13] · src —
hearWoody Shaw — Rosewood (Rosewood) · Tidal · Spotify
Shaw built entire solos out of wide, deliberate leaps — fourths and beyond — rather than the stepwise bebop vocabulary most trumpet players default to. Listen for how committed each leap sounds, with no audible search for the landing note, which is exactly what today's interval work is chasing at a much slower tempo.