3
[2026-09-11] · src Kumar, Reiß, van Loock, Barz; arXiv:2609.10698 (2026-09-11)
A Bell-state measurement (BSM) tells you which of the four two-photon entangled states a pair is in, and it is the basic operation behind measurement-based and fusion-based photonic quantum computing and behind entanglement swapping in quantum networks - all of which need to do many BSMs in a row. Built from linear optics alone (beamsplitters and photon counting, no nonlinear interaction), a BSM cannot distinguish all four Bell states and is fundamentally capped at 50% success probability, a hard constant that has bounded photonic key rates and graph-state generation rates ever since. Kumar, Reiß, van Loock and Barz encode each qubit of the pair in a two-photon repetition code - the simplest parity code, built only to catch bit-flip errors - and perform the BSM on the encoded pairs; the redundancy lets them identify and discard exactly the failure patterns that force a bare linear-optical BSM to give up, and they measure a mean success probability of (70.8 +/- 0.4)%, comfortably above the 50% bound. Because fault-tolerant photonic architectures need this same error-correcting encoding anyway for other reasons, the gain costs no extra photons or resources beyond what such a system would already carry.
The 50% linear-optics BSM limit has been treated as fixed background physics, something protocols are designed around rather than something that moves. This shows it moves once qubits are encoded - and, crucially, that the encoding a fault-tolerant photonic architecture needs anyway for error correction is enough to buy the improvement, so there is no separate cost-benefit tradeoff to make. The demonstration is on the smallest code that can do this (a two-qubit repetition code that only catches bit-flips, not phase-flips), so it is not yet a fault-tolerant BSM in the full sense; what it establishes is that the direction - spend error-correction resources you already need, get a better BSM as a side effect - actually works in the lab and not just on paper.
[2026-09-11] · src Zhu, Cao, Liu, He, Zhang, Zhang, Du, Wang, Jiang, Su; arXiv:2609.11004 (2026-09-11)
Squeezed light - where quantum noise is pushed below the shot-noise limit in one quadrature at the cost of raising it in the other - is what already lets gravitational-wave detectors beat the shot-noise floor at the low, audio-band frequencies (tens of Hz to kHz) where most of their sensitivity lives, but every such source to date has been a large, actively-stabilized tabletop setup: technical noise grows sharply as frequency drops, and holding the local oscillator's phase steady against a fragile vacuum-level quantum signal for long stretches is hard. Zhu et al. generate two-mode squeezed vacuum down to 60 Hz directly on a silica-microcavity photonic chip, solving the phase-stability problem with a coherent-comb trick: a weak electro-optic reference comb, in the orthogonal polarization, co-propagates with the vacuum at the quantum-signal frequencies, giving a stable phase reference without ever touching (and so degrading) the quantum modes themselves. From the resulting covariance-matrix reconstruction, the positive-partial-transposition entanglement criterion gives a minimum symplectic eigenvalue of 0.395, below the 0.5 threshold required to certify genuine two-mode entanglement - the paper reports this eigenvalue rather than a decibel squeezing figure.
This is a scale, not a discovery: audio-band squeezing itself is well established, and 0.395 is a modest result next to the multi-dB squeezing routine in tabletop setups. What is new is doing it, stably, on a chip - the specific combination (low frequency plus long-timescale phase stability) that bulk optics has needed a room-sized table to hold onto. If the phase-referencing trick scales to deeper squeezing, it opens a path toward chip-scale audio-band quantum sensors and continuous-variable quantum-information hardware that does not require a lab bench, rather than only ever tabletop demonstrations of the same physics.
[2026-09-11] · src IceCube Collaboration (Abbasi et al.); arXiv:2609.10931 (2026-09-11)
On 2023-02-13 KM3NeT's ARCA detector recorded KM3-230213A, a ~220 PeV neutrino, the highest-energy neutrino ever seen, and where it came from - a genuine astrophysical point source, a diffuse cosmogenic background, or something rarer - is still unsettled (this panel has already passed on several theory papers proposing an origin for it). Here the IceCube Collaboration searches 15 years of its own archival data for neutrinos from that same direction under three timing hypotheses - a steady source, a flare coincident with the KM3NeT detection, and a flare at an unconstrained time anywhere in the archive - and finds no significant excess under any of them. The best hotspot found in the unconstrained-flare search has a pre-trial p-value of 0.0078, but once the look-elsewhere effect across the full time-and-position search is accounted for, its post-trial significance drops to 1 sigma: fully consistent with background alone. Compared against KM3NeT's own estimate of the flux a point source would need to produce the single observed event, IceCube's upper limits for a short-timescale transient lie outside that estimate's 3-sigma range, assuming the same spectrum extends down into IceCube's more sensitive energy band.
This does not say what produced the event, but it narrows the field: a persistent point source, and the kind of coincident flare IceCube's own exposure would have caught, are both now in tension with the data, not just theoretically disfavored. What survives is a genuinely rare one-off transient with a spectrum too hard to show up in IceCube's band, a diffuse or cosmogenic origin rather than a single source, or a source sitting in part of the sky or energy range where IceCube's sensitivity and KM3NeT's simply do not overlap well. For an event this singular, ruling explanations out is most of the available progress right now.
2
[2026-09-11] · src twelve-keys/src/exercises/enclosures/exercise.py
Piano gets today's block alone: alto went yesterday, trumpet the day before, and piano carries both the longest gap of the three and rule 3's preference for a longer single block. It's line building's second lap, and rather than repeat lap one's guide tones — the notes a line is built to touch — today moves to the approach around them: the enclosure, which surrounds a target chord tone from both sides chromatically before landing on it. Key is A♭ major, a fourth from the one line-building key already used and, at four flats, one where the half-step neighbor tones the pattern needs don't fall under the fingers by habit.
5 minutes: play the A♭ major scale two octaves, hands separately, quarter = 60, naming each scale degree aloud as you play it — this is specifically so "the 3rd" and "the 7th" are addressed by ear during the warmup, not looked up mid-exercise. Then, right hand only, play C (the 3rd) approached from a half step above and a half step below in isolation, slowly, six times: D♭–B–C, D♭–B–C, six times through, before the tempo exercise below asks for it inside a line.
The full enclosures exercise as written, right hand, quarter = 80 in eighth notes: D♭ (half step above C) → B (half step below C) → C on the downbeat, cycling through the printed pattern that extends this to the 7th (A♭ half step above G, G♭ half step below G, landing on G) and then to four-note variations. Success criterion: three consecutive passes with the pickup landing exactly on the "and" of beat 4 each time, not late — a rushed or dragged pickup collapses the enclosure into a passing tone and the tension the exercise is built around disappears.
Over a slow ii–V–I in A♭ (B♭m7–E♭7–A♭maj7, one chord per bar, looped), improvise a 4-bar right-hand line that uses an enclosure to arrive at the 3rd or 7th of at least two of the three chords — not every beat, just two clean, prepared arrivals, so the ear has room to notice the approach rather than being fed one on every downbeat.
Stop after the three clean technical passes, or sooner if the enclosure starts arriving early or late more often than not — that's the pattern degrading into three notes played near a downbeat rather than a prepared approach to one, and more reps at that point rehearse the mistiming rather than the technique.
[2026-09-11] · src —
hearBud Powell — Tempus Fugit · Tidal · Spotify
Powell's right-hand lines here are enclosures end to end — barely a target note arrives without a chromatic neighbor on either side of it first, taken at a tempo fast enough that the approach has to be automatic rather than planned in the moment. Listen for how often a phrase's most emphasized note is the one that got surrounded rather than the one that got the most notes around it; that's the same arrival today's musical target is asking a 4-bar line to produce twice, not sixteen times.