3
[2026-08-12] · src arXiv:2504.05660 (Luo, Wang, Zheng et al., USTC); Phys. Rev. Lett. 137, 070801 (2026)
A USTC team entangled two laser-cooled rubidium-ensemble quantum memories joined by 420 km of deployed telecom fiber - more than four times the previous matter-to-matter distance record. The DLCZ-protocol link uses a two-photon Hong-Ou-Mandel-type heralding scheme that relaxes long-distance phase-stability requirements by roughly seven orders of magnitude versus single-photon heralding, with memory photons converted to the telecom S-band (0.17 dB/km loss) for transmission and a two-stage locking scheme holding the inter-memory phase stable. Past roughly 320 km, the measured entanglement generation rate exceeds what direct optical transmission alone could deliver at that distance - crossing the PLOB bound, the fundamental limit on any repeaterless quantum channel.
The PLOB bound is a hard ceiling on point-to-point quantum communication without a repeater; a repeater's entire reason to exist is to beat it, and until now that had been shown with photonic heralding or over metro-scale distances, not with real matter-based quantum memories at a distance relevant to actual infrastructure. This is one link, not a chain, and the absolute entanglement rate is still far from anything application-ready - but it is the first unambiguous demonstration that a DLCZ-style repeater architecture, not just its individual components in isolation, outperforms brute-force fiber transmission end-to-end.
[2026-08-12] · src arXiv:2507.12681 (Gil, Li, Stratton, Kaptanoglu, Stenson); Phys. Rev. Lett. 137, 065101 (2026), doi:10.1103/n7gk-922h
Finding a stellarator's non-planar coils has meant scanning thousands of weighted optimization runs to trade off plasma-confinement quality against buildability (curvature limits, coil-to-coil and coil-to-plasma spacing) - prior state of the art needed roughly 8,500 separate runs on a supercomputer cluster to surface a handful of workable candidates. This paper applies the augmented Lagrangian method - a constrained-optimization algorithm known since 1969 but never systematically applied to this problem - which treats each engineering constraint directly instead of as a tunable penalty weight. It converges to Pareto-superior coil sets, better on physics performance and engineering feasibility simultaneously, in about 40 minutes on a laptop, and the resulting designs outperform published coil sets for operating stellarators including Wendelstein 7-X and HSX.
This is a design-tool result, not a built coil - nothing here has been fabricated - but coil geometry has been stellarator research's standing bottleneck precisely because each candidate cost supercomputer time to evaluate. Cutting that to a single interactive run, while also producing better designs, changes what is explorable at all: a group can now iterate over plasma configurations and constraints instead of committing a cluster allocation per candidate. Its value is contingent on these designs holding up against a real coil-winding attempt, which hasn't happened yet.
[2026-08-12] · src arXiv:2607.24742 (Qu, Ge, Wang, Schaan et al.)
The standard cosmological bound on the summed neutrino mass, Sigma m_nu < 0.056 eV (Planck CMB + CMB lensing + DESI DR2 BAO), assumes flat LambdaCDM
and DESI's own hint of evolving dark energy (reported here 2026-08-09)
is entangled with that assumption in the same fit, since both leave similar imprints on structure growth. This paper builds two bounds that don't require picking a dark-energy model: a dark-energy-marginalized fit that keeps all the data (Sigma m_nu < 0.152 eV), and a stricter "late-universe-free" route combining primary CMB with reconstructed CMB lensing alone, discarding late-time expansion data by construction (Sigma m_nu < 0.41 eV today, tightening to about 0.31 eV with Simons Observatory lensing).
The standard 0.056 eV bound already sits in 2-3 sigma tension with the ~0.10 eV floor that oscillation-measured mass splittings require under the inverted mass ordering, which has been read either as evidence against inverted ordering or as a sign LambdaCDM is misspecified. This doesn't settle that; it prices it. Once the dark-energy dependence is genuinely removed rather than just marginalized, the bound loosens by close to an order of magnitude and no longer excludes inverted ordering at all - a useful caution that the tightest neutrino-mass numbers now in circulation are tighter than the data alone justifies without also trusting LambdaCDM.
1
[2026-08-12] · src build/orrery-model.el
inbox/ask.org already answered whether the plist/alist split was the
right call (2026-08-08: alist where keys are Greg's own typing, plist where
the key set is this program's fixed vocabulary). This lesson is the
mechanism underneath that answer: why plist-get and assoc cannot be
swapped for each other, even though both walk a list of conses.
orrery-model--node-properties (orrery-model.el:47-58) builds an alist by
hand — (cons (upcase (org-element-property :key np)) ...) — and every
lookup into it goes through assoc:
(defun orrery-model--prop (props key &optional default)
(or (cdr (assoc (upcase key) props)) default))assoc compares keys with equal by default, so it can match two strings
that are equal but not eq — two separately-allocated \"WEIGHT\"
strings, say. That is exactly what a PROPERTIES drawer needs: Greg types
:Weight: in Org, upcase normalizes it to \"WEIGHT\", and assoc
finds it against a key built at parse time that is a *different string
object* holding the same characters. eq would fail that comparison; a
freshly-consed string is never eq to another one, however identical its
contents.
orrery-model--panel (orrery-model.el:96-109) goes the other way — it
builds a plist literally, (list :id id :title title :cadence ... ), and
orrery-model--assert-unique-ids (orrery-model.el:111-120) reads it back
with (plist-get panel :id). plist-get's default test is eq, not
equal. That is safe here only because the keys are keyword symbols,
not strings: Emacs interns every symbol that starts with :, so every
occurrence of :id anywhere in the running image — in this function, in
the caller, in a defun compiled yesterday — is the identical Lisp
object. eq on interned symbols is a pointer comparison, and it is correct
precisely because there is only ever one :id.
Swap the two and each breaks in a different way. Use plist-get on the
PROPERTIES alist and every lookup silently returns nil, because a fresh
key string built from Org text is never eq to anything — no error,
just an item with every field falling back to its default (or, since this
codebase uses orrery-model--prop! for required fields, a loud error
instead of a silent one, which is the point of that guard existing at
all). Use assoc on the panel plist and it still happens to work, because
assoc's equal subsumes eq for symbols — but it is doing a string-ish
structural walk to answer a question plist-get answers in one pointer
compare, on data where the fast path was available and free. The split in
this file is not stylistic: alist+assoc is the only combination that is
correct for hand-typed text; plist+plist-get is the one that is *fast
and correct* for a vocabulary the code itself controls.
Exercise (verifiable): the keyword-symbol case-sensitivity claim above is checkable without touching any file in this repo:
emacs -Q --batch --eval \
"(princ (format \"%S / %S\n\" (plist-get (list :id \"x\") :ID) (plist-get (list :id \"x\") :id)))"Predict the two outputs before running it. (Elisp's reader is
case-sensitive, so :id and :ID are two distinct interned symbols —
plist-get with the wrong one returns nil silently, exactly the failure
mode described above, just triggered by a typo instead of a parsed
string.) Then look at orrery-model--prop!'s (upcase key) call
(orrery-model.el:72) and explain why that function needs a normalizing
step the plist-based code never bothers with.
2
[2026-08-12] · src twelve-keys/src/exercises/coltrane_changes/exercise.py
Cycle: G♭ → D → B♭ → G♭ — Gbmaj7 – A7 – Dmaj7 – F7 – Bbmaj7 – Db7 – Gbmaj7. Six flats resolving through a two-sharp key and back is one of the most disorienting of the twelve starting points: the ear and the hand both want Gb's flats to keep governing the D-major bar, and they don't. That mismatch is the actual difficulty, more than any one chord shape.
5 minutes: play just the three destination shells in isolation and out of cycle order — Gbmaj7, Dmaj7, Bbmaj7 — each preceded by its own V7 (Db7, A7, F7), slowly, as three unrelated little cadences. The goal is to hear each arrival as "home" on its own terms before asking the hand to leave it a half-second later.
The full six-chord cycle at quarter=70 (the exercise's own practice tempo), three keys at a time as the notes suggest. Success criterion: no hesitation at either V7-to-maj7 landing, and specifically the Db7 → Gb resolution back to the top — that's the one bar where fingers want to keep reaching for the D-major shapes instead of returning to Gb, and it's the tell for whether the cycle is actually internalized or just being found by ear bar to bar.
Apply the substitution once, for real: take the last four bars of Tune Up and replace the ii–V–I with the Coltrane substitute (♭III7 – ♭VI – VII7 – III – V7 – I), the way the exercise notes describe building Countdown out of Tune Up. Play it as a phrase that resolves, not as a chord-name checklist — the point is that it should still sound like a cadence landing, just a stranger one.
Stop the moment the cycle turns into a fingering drill instead of a harmonic one — when attention is on where the hand goes next rather than on hearing the next key arrive. Past that point, more reps just cement the motion without the ear that's supposed to be doing the work, which is the opposite of the point of this exercise.
[2026-08-12] · src —
hearJohn Coltrane — Countdown (Giant Steps) · Tidal · Spotify
Coltrane's own reharmonization of Tune Up, built by running exactly this major-thirds substitution through all three of the tune's ii–V–I phrases — the same move as today's musical target, at tempo, for real. Listen for how little the melody has to change even though the harmony underneath it is almost unrecognizable.