The Daily

Der Frühling ist noch nicht stabil, aber er ist schon spürbar: Mehr Tageslicht, erste milde Nachmittage, und plötzlich funktionieren Orte, die im Jänner nur „auf dem Papier" gut klangen. Das praktische Problem bleibt trotzdem dasselbe: Man will raus, aber man will auch etwas, das sich wirklich nach Belohnung anfühlt — nicht nur „irgendein Spaziergang". Heute deshalb zwei Picks, die genau diese Schwelle bedienen: ein klarer Saisonmarker (Tichy) und ein konkreter Termin, der sich wie ein Mini-Trip in die Stadt anfühlt (Wiedereröffnung Museum der Illusionen). Wenn ihr nur ein Ding davon macht, nehmt das mit dem besten Timing: erst das, was schnell voll wird, dann das, was im Kalender sitzt. Und ja: Favoriten lohnt sich wieder einmal als Ziel, nicht als Durchgang.

In Wien gibt’s ein paar inoffizielle „Jahreszeiten-Uhren" — Tichy ist eine davon. Laut 1000things hat der Eissalon am Reumannplatz seit 13.03.2026 wieder täglich von 10 bis 23 Uhr geöffnet, samt dem üblichen Magneten: den Eismarillenknödeln. Das ist nicht nur Nostalgie, sondern schlicht ein sehr gutes, planbares Ritual: hingehen, kurz anstehen, und danach ist man automatisch in „Frühling". Interessant ist das Extra drumherum: Um 17 Uhr startet am selben Ort ein kostenloser Hörspaziergang ([AR]TWALK Favoriten), der euch vom Reumannplatz über Favoritenstraße/Viktor-Adler-Markt bis zur Brotfabrik führt. Das macht aus „Eis holen" plötzlich eine kleine Route, die man auch Gästen guten Gewissens verkaufen kann. Für alle, die den 10. Bezirk sonst nur als U1-Transit kennen: das ist die freundlichste Art, ihn neu zu sehen.

Ein gutes Stadtding ist oft: kurze Strecke, klarer Termin, und danach fühlt man sich wie „unterwegs" ohne großen Aufwand. Genau so klingt die Meldung aus der 1000things-Märzvorschau: Das Museum der Illusionen (Wallnerstraße 4, 1010) war wegen Umbau geschlossen und soll ab 19.03.2026 wieder öffnen — mit neuen Installationen und frischen Attraktionen. Das ist auch als After-Work-Plan brauchbar, weil es nicht „ganzer Abend" sein muss, aber genug liefert, um sich zu lohnen. Und: Für Familien oder Besuch ist es eine dieser Optionen, die wetterunabhängig funktionieren, wenn der Westwind den „Spaziergang" ruiniert. Wer es schlau spielt, koppelt das mit einem konkreten Anlass (z.B. „Wiedereröffnung") statt „wir gehen halt ins Museum" — die Energie ist eine andere.

The pattern today is less about "a smarter model" and more about plumbing: directories, context bridges, and the invisible bits that make systems either usable or dangerously fragile. On one end, Meta is buying an agent network mostly because it looks like a new kind of social graph — but for autonomous workers. On the other, Google is turning Workspace into a context engine that can draft, edit, and search across your account, which is powerful precisely because it reduces friction. The catch is obvious: the more context you wire in, the larger the blast radius when something goes wrong. That’s why the security story matters: attackers are now hiding executable payloads in literally invisible Unicode characters, betting that humans and tools will both miss them. The next step across all three stories is governance: who can connect, what can be pulled, and what gets reviewed before it ships. "Agentic" is becoming an operational discipline, not a demo.

Ars reports that Meta has acquired Moltbook, a Reddit-esque social network populated by AI agents that went viral recently, and will hire creators Matt Schlicht and Ben Parr into Meta Superintelligence Labs. The terms weren’t disclosed, but Meta’s stated interest is telling: the founders’ approach to connecting agents through an "always-on directory". That phrasing signals a product thesis — not just agents, but an infrastructure layer that makes agents discoverable and persistent. The story also underlines the messy reality: Moltbook’s "humans can’t join" constraint wasn’t secure, and some posts were likely written by humans pretending to be agents. That matters because the biggest risks of agent platforms are identity, provenance, and permissioning, not just hallucinations. The obvious next step is whether Meta tries to productize agent directories inside its existing surfaces (Messenger, WhatsApp, Instagram) or keep it as a research lab project.

Google is overhauling Gemini features across Docs, Sheets, Slides, and Drive, aiming to replace "blank page" work with prompt-first drafting and editing inside the document UI. The key shift is contextual: Gemini can pull from Gmail, other documents, Chat, and the web when you explicitly rope those sources into a prompt. Editing is also getting more granular — highlight a section, ask for a rewrite, match tone/style, and keep suggestions private until you accept them. Sheets is being positioned as near-human for spreadsheet generation and analysis, and Drive search is moving toward AI summaries with citations at the top of results. Rollout is gradual over spring, globally but English-only at first, and some features are US-limited initially. The tradeoff is clear: Workspace becomes dramatically more capable, but also more sensitive to account-level data leakage and policy mistakes, so admin controls and "smart features" toggles become a first-class product surface.

Aikido Security reports a campaign (they call it Glassworm) that uploaded 151 malicious packages to GitHub between March 3 and March 9, with similar spillover to npm and Open VSX. The twist is both clever and depressing: parts of the payload are encoded in Unicode characters that are effectively invisible in editors and code review UIs, so humans see whitespace while interpreters see executable code. The packages are made harder to spot because the visible portions look like realistic maintenance: version bumps, documentation tweaks, small refactors. Researchers suspect LLMs are being used to mass-produce those convincingly normal-looking changes, which scales the attack far beyond hand-crafted malware. The defensive implication is not "be careful" (we already are), but "treat dependency intake like a controlled supply chain": tighter allowlists, stronger provenance, automated decoding checks, and stricter review gates for new packages. If you maintain open-source, this is the kind of threat that quietly raises the cost of being generous.

Today’s through-line is practical biology: not "can we make it work" but "can we keep it working long enough to matter". Sana’s update is compelling because it measures the boring but decisive metric — how long transplanted cells keep producing insulin in humans. Stämm’s announcement is pure infrastructure: a bioprocessor designed to reduce shear stress and make manufacturing more modular and decentralized, which is exactly where cell and gene therapy economics are currently breaking. And Aspen’s naming milestone is small on paper but meaningful for program maturity: you don’t get an official nonproprietary name unless you’re moving with intent toward clinical and regulatory formality. The next step across all three is execution: more patients, more sites, more repeatability. In this phase, the winner is often the team that can make consistency feel routine.

Sana Biotechnology reported that its transplanted islet cell therapy produced insulin for 14 months in one person with type 1 diabetes, a durability datapoint that matters because cell therapy headlines often fade when persistence collapses. The caveat is obvious: a single patient does not make a product, and variability across recipients (immunosuppression, engraftment, inflammatory environment) can be brutal. Still, "still producing insulin at 14 months" is not a vibes metric — it’s a hard constraint on whether this approach can become more than a heroic case study. The near-term question is whether Sana can reproduce that signal across additional patients and characterize what correlates with durability (cell dose, site, conditioning regimen, immune profile). If the persistence is real, the next step becomes scale: manufacturing consistency, distribution, and an immunosuppression story that is tolerable outside top academic centers. This is the kind of program where the second and third data points are more important than the first.

Stämm announced a single-use "High-Throughput Bioprocessor" built around a bubble-free bioreactor that avoids impellers, sparging, and antifoam agents by relying on laminar flow modeled on capillary networks. The claimed payoff is less hydrodynamic shear stress (which can damage sensitive cell populations) and a platform that scales "isometrically" from 50 mL to 250 mL while targeting reductions in CAPEX and media costs. The positioning is explicitly geopolitical and logistical: modular, plug-and-play manufacturing that can be deployed closer to patients rather than only in centralized, capital-heavy facilities. For cell and gene therapies — where small-batch, personalized production is common — that framing hits a real bottleneck, not a marketing trend. It’s RUO for process development today, with a promise of more technical data and limited early access partnerships. The next step to watch is whether pharma teams can validate performance against existing stirred-tank baselines in their own workflows, because "decentralized" is only credible when QC and reproducibility are boring.

Aspen Neuroscience says its investigational autologous cell therapy ANPD001 has received an official nonproprietary name: sasineprocel. This kind of update looks trivial, but it’s a marker that a program is moving through the machinery that surrounds later-stage development: naming conventions, regulatory-facing documentation, and the operational discipline required to treat a therapy like a real medicine rather than a lab project. Autologous cell therapy for Parkinson’s is an ambitious bet, because it combines the biological challenge (neuronal replacement/function) with the manufacturing challenge (patient-specific production and logistics). The implicit question is how Aspen plans to prove clinically meaningful benefit in a disease where endpoints can be slow and noisy, and how it will handle scalability if the approach works. The naming milestone doesn’t answer that, but it does suggest continued forward momentum. The next step to watch is trial design details and any early safety/feasibility readouts that make the therapy feel less like science fiction and more like a program.

The three papers today are a nice snapshot of where the field actually is, not where it wishes it were. First, a phase 1 PSMA CAR-T trial in metastatic castration-resistant prostate cancer shows both sides of the knife: meaningful activity in a subset, but toxicity that forces safety-switch activation in a quarter of patients. Second, a spatiotemporal profiling study compares CAR-T to IL-15–enhanced CAR-NKT cells and finds the real differentiator in solid tumors may be trafficking and localization, not raw cytotoxicity. Third, a review on lupus precision immunotherapies makes the case that the next decade will be defined by mechanism-guided targeting — and quietly, by how well we use AI to stratify heterogeneous patients. Put differently: we’re moving from "one product" thinking to "a playbook" of modalities, combinations, and patient selection. The next step is to turn these mechanistic insights into trials that are both safer and more decisive.

This phase 1 trial evaluated P-PSMA-101, a PSMA-targeting CAR-T product engineered to be enriched for stem cell-like memory T cells (TSCM), in metastatic castration-resistant prostate cancer. Among 33 treated patients, cytokine release syndrome occurred in 61%, with 9% experiencing Grade ≥3 CRS, underscoring that "solid tumor" doesn’t automatically mean "lower cytokine risk". Notably, the built-in iCasp9 safety switch was activated in 24% of cases, including one fatal toxicity, which is a sobering reminder that safety switches are not cosmetic. On efficacy, 21% achieved a ≥50% PSA decline, and among RECIST-evaluable patients, one partial response was observed; two remissions exceeded 12 months with PSA declines >90%. The authors frame the result as "robust expansion" producing both toxicity and durable responses in some patients. The next step is clear: better control of expansion/activation kinetics (or smarter dosing/conditioning) while preserving persistence.

This preclinical study compares conventional CAR-T cells with allogeneic stem cell–derived, IL-15–enhanced CAR-NKT cells using spatiotemporal transcriptomic profiling across tissues and time points. The headline finding is pragmatic: compared with CAR-T, CAR-NKT showed superior homing, infiltration, and localization within solid tumors, plus prolonged persistence and a distinct checkpoint receptor landscape. The work also suggests that "checkpoint combination" may be modality-specific: CAR-T responses were synergistic with TIGIT blockade, while CAR-NKT appeared more sensitive to CD96 blockade in vivo. That matters because it hints that we’re not just swapping cell types; we’re changing the rules of which add-on drugs make sense. The study’s value is in turning "CAR-NKT sounds promising" into a map of pharmacokinetic, pharmacodynamic, and immunoregulatory differences. The next step is translating the profiling insights into clinical trial designs that test the right checkpoint combinations rather than defaulting to PD-1 out of habit.

This review argues that lupus care is shifting from broad immunosuppression toward mechanism-based, precision approaches that selectively target pathogenic circuits. It frames three domains: (1) surface-antigen/costimulatory targeting (including BAFF/APRIL antagonism and even CD19-directed CAR-T strategies), (2) cytokine and intracellular signaling modulation (IFN-I blockade; JAK/STAT and mTOR inhibition), and (3) next-generation antigen-focused therapies (mimetic peptides, CAAR-T, antigen-specific Tregs) aimed at durable tolerance with less systemic suppression. The useful part is the translational stance: it outlines a clinical implementation roadmap focused on scalability and realistic windows over the next 5–10 years, rather than assuming every concept becomes a product. It also explicitly calls out AI/ML as a lever for patient endotyping and predictive modeling in a disease defined by heterogeneity. That’s not hype if it’s used to decide who gets what and when, rather than to generate pretty heatmaps. The next step is integrating molecular stratification into trials so that "precision" is operational, not rhetorical.

Wenn man Italien in diesem Frühjahr clever spielt, sucht man nach Anlässen, die konkret sind (Datum/Ort) und trotzdem nicht wie Touristenmassentourismus wirken. In Venedig ist das oft eine Ausstellung, die man in einen Spaziergang durch weniger überlaufene Ecken einbettet. In Apulien sind es die Feste, die wie ein lokales Betriebssystem funktionieren: Essen, Rituale, Musik, soziale Logik — und plötzlich ist man nicht in „Urlaub", sondern in einem Kalender. Heute deshalb zwei Terminstücke: einmal Venedig als „Museum ohne Dach" (Migrationserzählungen als Ausstellung), einmal Salento mit der sehr ernst gemeinten Gastfreundschaft der Tavole di San Giuseppe. Beide sind eher Schulter-Season als Postkarten-Sommer. Und beide lassen sich so planen, dass man nicht nur „da war", sondern wirklich etwas gesehen/verstanden hat.

Venezia Unica listet mit „Dreams in Transit" eine Ausstellung, die Migration als Sammlung von Geschichten und Perspektiven verhandelt — nicht als Fußnote. Das ist genau die Art Kulturprogramm, die Venedig außerhalb von Biennale-Hype wieder spannend macht: man hat einen klaren Anker, aber man kann den Tag drumherum frei bauen. Inhaltlich ist das Thema breit genug, dass es nicht in 20 Minuten „abgehakt" ist, und gleichzeitig konkret genug, um danach eine Route zu wählen, die nicht nur aus Rialto–San Marco besteht. Für eine Venedig-Reise ist das oft der Trick: ein Programmpunkt, der euch in eine bestimmte Gegend zieht, und dann der Rest als Drift durch Gassen. Wer mit Kindern reist, kann das auch als „kurz rein, dann draußen" spielen, ohne dass sich jemand in Museumsfluren verliert. Next step: auf der Venezia-Unica-Seite die Detailinfos (Ort/Zeiten) checken — die Kurzliste selbst ist bewusst minimal.

In Minervino di Lecce läuft vom 17. bis 19. März (laut Visit Puglia) die zweite Ausgabe von „Borghi divini tra Santi, Fiabe e Megaliti" — rund um die Tradition der Tavole di San Giuseppe. Der Kern ist eine historische Praxis: wohlhabendere Familien deckten zu San Giuseppe eine reich gedeckte Tafel für weniger Begüterte; heute werden die Tavole in Häusern in Minervino/Cocumola/Specchia Gallone aufgebaut und können besucht werden. Der Artikel nennt als starkes Detail 169 Speisen, konkretisiert als 13 Gerichte pro „heiligem" Platz — inklusive *massa* (Hartweizenpasta mit Kichererbsen und Kohl), Lampascioni, Pittule, Fisch und mehr. Dazu gibt’s Programm: Convegno am 17.03. um 18 Uhr, Gustolab-Labore am 18./19., eine „Tavola dei Bambini" und abends Folk-Musik („Gusto Folk"). Reisepraktisch ist das ein super Anlass, weil man nicht nur „im Salento" ist, sondern genau weiß, wann und wo man in eine lokale Logik reinkommt. Next step: die Besuchszeiten der privaten Tavole (18. abends, 19. vormittags) und die QR-Route vor Ort einplanen.

NBA March basketball has two parallel plots: teams trying to stabilize rotations for the postseason, and teams improvising nightly because bodies are missing. That’s why returns matter so much — one player coming back changes lineups, spacing, and minutes distribution in a way a "good practice" never will. Cleveland’s Max Strus is a clean example: a role player who shifts the offense immediately because he’s willing to shoot and move without the ball. New York’s comeback against a short-handed Warriors team is the other side: even when the opponent is patched together, 21-point holes are real, and digging out reveals what a team’s baseline habits actually are. And then there’s the darker comeback story: Kawhi leaves late with an ankle sprain, and "we’ll see" becomes a whole franchise’s mood. Next step for all three: availability reports and how coaches decide to ramp minutes without breaking the players they just got back.

Max Strus returned for Cleveland after missing 67 games following surgery for a Jones fracture, and immediately looked like someone who missed playing more than he missed training. He scored 24 points in 23 minutes, hitting 6-of-7 from three, with 16 of those points coming in the first half. AP notes he’s the first player since 1997–98 to score at least 15 points in the first half of a season debut after not playing in at least the first 60 games. The human detail is good: Strus was happy to be back but framed it as a bad loss because Dallas won 130–120, which is the posture you want from a team trying to gear up. Coach Kenny Atkinson said the plan was 20–24 minutes, building up from there, which is the typical "ramp without re-break" approach. The next step is how quickly Strus’s shooting gravity and movement can stabilize a Cavs team that’s been juggling injuries all season.

New York erased a 21-point hole to beat Golden State 110–107, with Jalen Brunson posting 30 points and 9 assists in the kind of game where the second half becomes a test of decision-making. The Knicks didn’t lead after early minutes until the final minutes of the third quarter, which means the comeback was not a single run — it was a sustained reset. Golden State was deeply short-handed: Stephen Curry missed a 17th straight game with knee pain/inflammation, and multiple recognizable names sat, forcing yet another new starting lineup. Despite that, the Warriors still built a 35–21 first quarter and led 46–25 at one point, which is why the Knicks will care about the first-half process even in a win. The final possession ended with a turnover and no shot, a brutally common way for short-handed teams to lose close games. Next step: whether the Knicks can start games with the same clarity they’re showing late, because playoff opponents won’t spot you 21.

Kawhi Leonard sprained his left ankle late in a 118–109 loss to Sacramento and limped to the bench before heading to the locker room with 9:27 left in the fourth. The Clippers didn’t say afterward whether he will miss time, which is the worst kind of update in March because it turns every rotation plan into a conditional. Leonard still finished with 31 points, and the note about his scoring streak (45 straight games with 20+ points, a franchise record) is a reminder of how central he is to the team’s floor. The injury happened on an awkward landing while guarding DeMar DeRozan — exactly the kind of non-contact-ish twist that makes timelines unpredictable. If the ankle is minor, the next step is minutes management; if not, the next step is triage. Either way, this is the part of the season where "availability" becomes a skill.

Kinderprogramm wird dann gut, wenn es nicht nur „nett" ist, sondern euer Wochenende strukturiert: klare Uhrzeit, klarer Ort, und am Ende sind alle halbwegs zufrieden. Heute deshalb zwei Termine aus der 1000things-Märzvorschau, die sehr unterschiedlich sind, aber beide den Vorteil haben, dass sie nicht auf gutes Wetter angewiesen sind. Das Kinderzimmer Pop-up ist ein Familienmarkt, der sich wie ein gemeinsamer Ausflug anfühlt (schauen, basteln, ein bisschen Show). „World of Bricks" ist das Gegenteil: ein großes, visuelles Ding, das bei vielen Kindern sofort den „Wow"-Hebel trifft — und bei Erwachsenen den „Ah, endlich ruhig"-Hebel. Beide sind auch gut mit Besuch kombinierbar, weil man keine Insider-Vorkenntnisse braucht. Und beide sind planbar genug, dass man sie als Fixpunkt setzen kann, statt spontan im Wind zu diskutieren.

Laut 1000things findet am 22.03.2026 wieder das Kinderzimmer Pop-up im Gasometer statt — ein kuratierter Erlebnismarkt, bei dem kleine österreichische Unternehmen für einen Tag offline sichtbar werden. Das Format ist praktisch, weil es nicht nur Shopping ist: Es gibt typischerweise Bastel- und Malbereiche, Shows, Kinder-Yoga und weitere Programmpunkte, also genug Abwechslung für unterschiedliche Altersstufen. Die Zeitspanne 10–17 Uhr ist elternfreundlich: man kann früh rein (wenn die Energie hoch ist) und hat trotzdem noch einen Nachmittag übrig. Preislich nennt 1000things 8 € für Erwachsene, Kinder bis 12 kostenlos, was als „Familienausflug" fair bleibt. Der nächste Schritt ist simpel: vorher entscheiden, ob ihr „früh & kurz" oder „mittig & länger" macht — bei Indoor-Events ist die Crowd-Kurve oft entscheidend.

„World of Bricks" zieht laut 1000things von 28.03. bis 06.04.2026 in die Wiener Stadthalle (Studio F) und bespielt rund 1.500 Quadratmeter mit Ausstellungsmodellen und Ideen rund um Klemmbausteine. Das ist ein guter Slot: spät genug, dass man es als „Frühlingsferien-Vorläufer" planen kann, aber früh genug, dass es nicht mit den ganzen Osterterminen kollidiert. Für viele Kinder funktioniert das als visuelles Großereignis — und für Erwachsene als „endlich etwas, das nicht von meiner Erklärleistung lebt". 1000things nennt als Ticketpreis 26,33 € (vermutlich je nach Kategorie/Slot), also: vorher prüfen, ob es Familien-/Zeitfenster gibt. Der nächste Schritt ist, das Ganze als 2–3-Stunden-Plan zu denken, nicht als Tagesausflug — dann bleibt die Stimmung stabil.