The Daily

Antigen escape is still the brutal, non-glamorous reason CAR-T patients relapse — the target gets downregulated or disappears, and the engineered T cells lose their grip. A team at the Institute of Process Engineering (Chinese Academy of Sciences) takes a very different approach: instead of re-engineering the CAR, they add a biomimetic supplement called a ferritin aggregation cell engager (FACE) during manufacturing. The core trick is using CD71 (highly expressed on leukemia cells and also present on autologous CAR-T cells) as a universal “handle” to create extra cell–cell engagement once infused. In patient-derived xenograft (PDX) models with normal antigen expression, FACE-CAR-T matched conventional CAR-T efficacy with only one-fifth of the cell dose, which matters for safety and manufacturing throughput. When antigen levels dropped to <10% of normal, FACE-CAR-T still cleared leukemia and improved survival in PDX models — exactly the scenario where standard CAR-T often stalls. They also describe a drug-loaded extension (“FACED”) using ferritin’s cage-like structure to carry chemotherapeutics, aiming to hit high-burden and antigen-negative relapse niches. Next step is the unglamorous one: reproducibility and toxicology in models that better predict cytokine release and off-tumor risks before anyone gets excited about “plug-in” CAR-T upgrades.

General relativity is built on geodesics — the “shortest paths” objects follow through curved spacetime — but quantum gravity would make spacetime itself fuzzy. A group at TU Wien reports a method to quantize the spacetime metric in a specific, physically relevant case: a time-independent, spherically symmetric gravitational field (a toy model that can approximate something like the Sun’s field). From that, they derive a modified equation of motion they call the q-desic equation, meant to be the quantum analogue of the geodesic equation. Their key point is conceptual but test-driven: you can’t always replace a metric operator with its expectation value and assume classical behavior “on average.” In the purely classical-gravity limit they estimate the deviations are absurdly tiny — on the order of 10^-35 meters — which is why quantum gravity has been so hard to probe. But they argue the cosmological constant (dark-energy term) can change the scaling at large distances, hinting at possible observable handles. The next step is whether any astrophysical or precision-lab scenario can amplify the q-desic signatures into something measurable rather than mathematically elegant.

Wenn Wien sich kulinarisch selbst bestätigt, dann meistens im März: Der jährliche Falstaff-Restaurantguide ist draußen — laut Bericht wurden 2.050 Restaurants getestet, und die Präsentation fand im Wiener Rathaus statt. In Wien gab es 2026 nur einen Betrieb mit voller Punktzahl: das Steirereck im Stadtpark erreicht erneut 100 Punkte (laut Artikel seit 2016 konstant Höchstwertung). Direkt dahinter kommt die “99-Punkte”-Riege: Amador (Döbling), Konstantin Filippou (1.) und Mraz & Sohn (20.). Mit 98 Punkten liegt Silvio Nickol (1.), danach folgen Aend und Glasswing mit 97 Punkten. Was daran spannend ist: das ist nicht nur “Liste”, das ist ein brauchbarer Stadtplan für Abende, an denen man keine Energie für Recherche hat — und ein Indikator, welche Adressen ihre Konstanz halten. Next step (für uns Leser): ein Blick auf die unteren Ränge (96–94) lohnt oft am meisten, weil dort Preis/Erlebnis-Verhältnis oft besser ist als an der Spitze.

Für ein Wochenende mit “ich will raus, aber nicht konsumieren”: Die Umweltschutzorganisation Global 2000 organisiert von Fr, 13. bis So, 15. März 2026 das Filmfestival „Eyes on Earth“ im Schikaneder (Margaretenstraße 22/24, 1040). Laut Ankündigung laufen fünf Filme plus zwei Kurzfilme zu den Themen Ressourcen, Ernährung und Wirtschaft — ergänzt durch Gespräche mit Expert*innen. Tickets werden mit 8,50 € angegeben, also eher “Spontanabend” als “Projekt”. Inhaltlich passt das gut in die Woche, in der Wien wieder spürbar Frühlingsluft hat: man kann nachmittags spazieren und abends mit einem konkreten Thema im Kopf rausgehen. Next step: Wenn du nur einen Slot schaffst, wähle einen Abend mit Talk — das ist der Teil, der hängen bleibt.

NVIDIA is leaning hard into a simple story: agentic AI and coding assistants are bottlenecked by latency and long-context throughput, and the path forward is extreme hardware–software codesign. Citing SemiAnalysis InferenceX and other benchmark work, the company claims GB300 NVL72 systems can deliver up to 50× higher throughput per megawatt, translating into up to 35× lower cost per million tokens compared with the Hopper platform. The blog post is unusually specific about where the wins come from: TensorRT-LLM kernel improvements, NVLink symmetric memory, and launch scheduling that reduces kernel idle time. It also calls out that software improvements alone delivered up to 5× better low-latency performance on GB200 in just four months — a reminder that “hardware generation” is only half the story now. In long-context scenarios (their example: 128k-token input and 8k-token output), NVIDIA claims GB300 can be ~1.5× cheaper per token than GB200. The practical implication is competitive: if these economics hold in production, it changes what kinds of interactive products can be profitably served at scale. Next step is verification in the wild — third-party measurements on real stacks (not just curated kernels) and power-normalized cost models that include networking, memory pressure, and failure rates.

If you strip away the hype, photonic AI hardware is a serious attempt to decouple inference from heat and electron bottlenecks. A University of Sydney team reports a nano photonic prototype where photons (not electrons) carry computation through nanoscale structures that implement a neural network “in the physics” of the device. They describe structures only tens of micrometres wide (roughly a human-hair diameter) and computation on a picosecond timescale. As a benchmark, they trained the system to classify more than 10,000 biomedical images (breast, chest and abdomen MRI scans), reporting 90–99% accuracy in simulations and physical tests. The paper is positioned as an energy-efficiency play: “AI is increasingly constrained by energy consumption,” and photonics aims to move the constraint. The next step is the hard engineering: scaling from a demonstrator to larger photonic neural networks without losing reproducibility, calibratability, or signal-to-noise — and proving that the end-to-end system beats state-of-the-art electronics once you include data movement.

This is less a product announcement and more a snapshot of where defense AI is stuck: transitioning from prototypes to deployment. The Autonomy in Defense 2026 summit (March 19, 2026, Washington, D.C.) frames the challenge as “operationalizing autonomy at scale” across domains — tactical edge, human-machine teaming, logistics, and training/digital twins. Retired Lt. Gen. Michael Groen (former director of the DoD Joint AI Center) is quoted bluntly: “We don’t have an innovation problem. We have an execution problem.” Another useful quote comes from Eileen Vidrine (former US Air Force Chief Data and AI Officer) on modernization tempo: playbooks have to match the operational speed, or someone else leapfrogs you. The practical signal is procurement and integration pressure: autonomy is being treated as infrastructure, not a special project. Next step to watch is whether this wave translates into standard interfaces, certification pathways, and deployable “edge stacks” — or stays trapped in demos and conference agendas.

Vertex reported a prespecified Week 36 interim analysis from a phase 3 trial of povetacicept (a dual BAFF/APRIL inhibitor) in IgA nephropathy (IgAN), hitting the primary endpoint versus placebo. The headline number is a 49.8% placebo-adjusted reduction in UPCR (urine protein-to-creatinine ratio), essentially right at the “best-case” analyst bar of 50%+. Secondary signals included reductions in serum galactose-deficient IgA1 and hematuria resolution, which analysts interpreted as breadth of effect, not just a single biomarker swing. Vertex says the drug was generally well tolerated with no deaths and no serious AEs linked to the molecule; one nuance mentioned is severe hypogammaglobulinemia cases, but severe infection rates were reportedly low and comparable to placebo. The strategic move is speed: Vertex plans to submit for FDA approval by the end of March, with analysts modeling a potential accelerated-approval decision around November. The competitive frame is tight: Otsuka’s Voyxact is already approved; Vera’s atacicept has a PDUFA date of July 7. Next step is how regulators weigh surrogate endpoints (proteinuria) and whether Vertex can position povetacicept as best-in-class on subgroup consistency and real-world tolerability.

Capricor says the FDA has resumed review of its Duchenne muscular dystrophy cell therapy deramiocel after previously issuing a complete response letter (CRL) in July 2025 — and the company uses unusually direct language that the CRL was “lifted.” The agency is now targeting an approval decision by Aug. 22. The resubmission is described as a Class 2 resubmission, meaning the package went beyond minor clarifications and included more substantive data. Capricor points to topline phase 3 data showing improvements in upper limb function and left ventricle ejection fraction, and says it followed up with a clinical study report in February at FDA request. The meta-signal is regulatory: the story explicitly mentions the upcoming departure of Vinay Prasad from CBER and the broader tension around rare-disease evidentiary standards. The next step is whether FDA’s restarted review reflects a genuine data threshold being met or a recalibration of internal risk tolerance — either way, it will be a closely watched precedent for other gene/cell therapy sponsors navigating CRLs.

Roche reported that the phase 3 persevERA study in first-line ER+/HER2- locally advanced or metastatic breast cancer did not meet its primary endpoint of a statistically significant improvement in progression-free survival for giredestrant + palbociclib versus letrozole + palbociclib. The company emphasizes a “numerical improvement” and says the combo was well tolerated, with AEs consistent with known safety profiles. A key strategic detail: Roche frames persevERA as one piece in a broader program, pointing to positive readouts in evERA (advanced setting) and lidERA (early-stage) as validation of giredestrant’s activity. They also highlight that the FDA has accepted a New Drug Application based on evERA data and that lidERA data submission is planned “in the coming weeks.” In other words, this is a miss — but not necessarily a derailment — and Roche is trying to keep the narrative focused on where giredestrant can still become standard-of-care. Next step is seeing full persevERA subgroup data and how Roche positions giredestrant in endocrine-resistant populations (pionERA readout expected in 2027).

The useful travel insight here isn’t that Italy is popular — it’s that municipalities are operationalizing crowd control like infrastructure. A Yahoo Travel guide walks through six places (including Venice, Cinque Terre, Florence, Portofino, Sirmione and the Amalfi Coast) and summarizes the tactics: day-visitor fees and scheduled windows, reservation slots (example: Via dell’Amore with 30-minute booking slots and caps), “no-waiting zones” with fines in tight scenic chokepoints, and even road rules like Amalfi’s stated ban on vehicles longer than 10.36 meters from 06:30 to midnight. The message is explicitly not “don’t come” but “plan like it’s a performance” — book early, read local notices, and treat limited-capacity trails like tickets. The practical next step is personal: if you want spontaneity, you’ll increasingly need to buy it with off-peak timing (early mornings, shoulder season, one-village days). For family travel, the upside is real: less chaos when systems work, more predictable logistics; the downside is less “let’s just see what happens.”

Bari spielt diese Woche Cocktail-Stadt: Laut Bargiornale läuft die Amaro Montenegro Week 2026 von 8. bis 14. März und bindet über 35 Lokale im Quartiere Umbertino ein. In den teilnehmenden Bars gibt’s neue Signature-Drinks wie den „Monte Punch“ (Milk Punch modern interpretiert) und den „Dry March“ (ein strukturierter Sour), jeweils mit konkreten Rezepturen im Artikel. Spannend (und typisch Italien) sind die angekündigten „Gemellaggi“: Gastauftritte von Bars aus Bologna (der “Heimat” von Montenegro) und sogar aus Potenza. Parallel ist die Gruppe auch bei der Splash (Beverage/Hospitality-Messe) in der Fiera del Levante präsent (8.–11. März, Stand 42). Warum das relevant ist: Für eine Puglia-Reise im März ist Bari oft nur “Ankunft/Abflug” — solche Mikro-Events machen die Stadt abends plötzlich zum Hauptprogrammpunkt. Next step: Wenn du hinfährst, such dir zwei, drei Bars in Laufdistanz und mach’s als kleinen “Quartier-Loop”, statt dich quer durch die Stadt zu hetzen.

BariExperience highlights “Villaggio del Gusto” as one of Bari’s recurring food events, staged on the seafront at Largo Giannella. The pitch is straightforward: producers, chefs and food artisans set up stands to showcase Puglian and broader Italian specialties, with a parallel entertainment program (live music and performances). A notable detail is the claim that the event has been recognized as “Italian Excellence 2026”, positioning it as more than a casual weekend market. For travelers, the format is exactly what works in early spring: you can do a long seaside walk and then eat your way through small bites without committing to a full restaurant evening. The next step is practical: check dates and peak times before you go (these seafront events can be calm at lunch and chaotic in the early evening). If you’re traveling with kids, it’s also an easy win because the “strolling + snacks” loop keeps energy levels stable.

This is one of those box scores that reads like a typo until you stare at it long enough: Bam Adebayo scored 83 in a 150–129 Heat win over the Wizards. The quarter-by-quarter pacing is even crazier: 31 in the first, 43 by halftime, and 62 after three. AP reports he went 20-for-43 from the field, 7-for-22 from three, and an absurd 36-for-43 at the line. The free-throw line is where history got rewritten: the article notes both free throws made and attempted were NBA records for a game, surpassing the previous marks (attempts: 39; makes: 28). The only scoring night above Adebayo now is Wilt’s 100 in 1962; Adebayo jumps past Kobe’s 81 into the #2 slot. Beyond the trivia, it also reframes Miami’s season momentum: the Heat have now won six straight and moved to a season-best eight games over .500. Next step is the “come down” game: historically, these explosive nights affect how defenses scheme the very next matchup (Milwaukee is next).

NBA.com’s Nightly Recap is the fastest way to compress a 10-game slate into a few minutes without pretending you watched everything. The headline items here: a “Texas showdown” where Wembanyama wows, Dončić drops 35 on the Knicks, and Miami extends its streak (which now looks even more surreal after the Adebayo 83-point follow-up). It’s also a good gut-check on team rhythm: you can see which offenses are humming versus which ones are scraping for late-clock shots. The next step is to use it like a filter: pick one or two games that looked structurally interesting (not just highlight plays) and then go to the full-game condensed versions.

If you only have time for one clip: the Top 10 is the guilt-free version of “keeping up.” It’s essentially a snapshot of where the league’s athletic edge is right now — transition defense broken by one extra gear, weakside blocks that feel like physics violations, and the occasional pure footwork clinic in the half-court. Watching these back-to-back with the Nightly Recap is also a neat trick: the recap gives context (why a run mattered), the Top 10 gives the visceral “that’s why we watch” moments.

Ein sehr konkreter Tipp für heute Abend: Laut Gebärdenwelt.tv lädt WIENXTRA am Mittwoch, 11. März zu einem Elternvortrag mit ÖGS ein. Thema: Inklusion und Wertschätzung bei Kindern — und das ist (ehrlich gesagt) genau die Art von Eltern-Input, die im Alltag wirkt, wenn man ihn nicht als “Erziehungstrick”, sondern als Haltung versteht. Zeitfenster ist 18:00 bis 20:00. Der größere Kontext des Artikels ist ein Monatsüberblick über Veranstaltungen mit ÖGS-Dolmetschung oder ÖGS-Angebot, aufgeteilt in kostenlose und kostenpflichtige Termine. Das ist praktisch, wenn man sich wieder einmal vornimmt, Kultur nicht nur “für die, die eh können” zu planen. Next step: Wenn du heute nicht kannst, markier dir zumindest den Wien-Museum-Termin (siehe unten) — der Mix aus Ausstellung + Zugänglichkeit ist selten.

Ebenfalls aus dem ÖGS-Kalender: Das Wien Museum eröffnet am Mittwoch, 25. März eine neue Ausstellung zur Donauinsel; die Eröffnung wird laut Artikel gedolmetscht und startet um 18:30. Zusätzlich ist eine Führung mit ÖGS am 12. April um 10:30 angekündigt. Für Familien ist das gleich doppelt brauchbar: Donauinsel ist ein Thema, das Kinder sofort greifen (Freizeit, Wasser, “wo wir hingehen”), und das Museum bietet einen Anlass, das Ganze als “Stadtgeschichte” zu erzählen. Next step: die Führung ist morgens — das ist genau das Zeitfenster, in dem man mit Kindern oft bessere Konzentration hat als am späten Nachmittag.

Der Prater ist nicht “Event”, er ist Infrastruktur für Familien — und am Sonntag, 15. März 2026 starten laut 1000things wieder alle Fahrgeschäfte in die Sommersaison. Im Artikel wird von 250 Attraktionen gesprochen, die nach dem Winter wieder öffnen. Das ist ein guter, niedriger Stress-Punkt im Kalender: man muss nichts “verstehen”, man muss nur hingehen, ein bisschen schauen, eine Fahrt machen, wieder gehen. Tipp: wenn du wirklich mit Kindern entspannen willst, geh am Vormittag oder am späten Nachmittag — mittags wird’s schnell “zu viel”. Next step: Wenn das Wetter so bleibt (16° als Tagesmaximum), könnte der 15. März der erste echte “Frühlings-Sonntag” werden.