ExpressVPN holds an edge in privacy and usability, while NordVPN wins on speed and features. Let's dive into a detailed comparison.

Speed Winner: NordVPN

All VPNs slow your internet speeds, but our tests show NordVPN cuts speeds by only 3% on average, while ExpressVPN loses 18%. Both are well below the 25% threshold, meaning you won't notice major slowdowns. However, if raw speed is your priority, NordVPN is the clear winner.

Value Winner: ExpressVPN

Both are expensive, but ExpressVPN's renewal prices are more palatable. Its Basic plan starts at $13/month, with occasional promotions. NordVPN's two-year plan renews at $140/year, which is steep. ExpressVPN's Advanced plan includes a password manager and identity protection features at a comparable introductory price, offering more value for bundled services.

Privacy and Security Winner: ExpressVPN

ExpressVPN goes above and beyond with its TrustedServer technology, post-quantum encryption, and a remarkable 23 independent audits since 2018. Its no-logs policy has been verified, and its ShuffleIP feature adds extra privacy. NordVPN also offers solid privacy with post-quantum protection and Onion over VPN, but its audit reports are not fully public. For critical privacy needs, ExpressVPN is the better choice.

Usability Winner: ExpressVPN

ExpressVPN is the easiest VPN to use across all platforms, with a clean, minimalistic interface. It excels on Apple TV and offers a simple router app. NordVPN's apps are also user-friendly but can be cluttered, especially on mobile. Both unblock streaming services reliably, but ExpressVPN has a slight edge with international Netflix libraries.

Ultimately, whether you choose ExpressVPN or NordVPN, you're getting one of the best. If you need the fastest speeds or advanced privacy features like multi-hop connections, go with NordVPN. For easiest use, superior streaming, and top-tier transparency, ExpressVPN is the winner.

Source: CNET News

What Are QLED and OLED?

QLED stands for Quantum Dot Light Emitting Diode. Despite the name, QLED is a variation of traditional LCD technology. It uses a standard LED backlight combined with a layer of quantum dots — microscopic molecules that emit specific colors when illuminated. This film enhances color accuracy and brightness. Samsung popularized QLED branding, but other manufacturers like TCL also produce QLED TVs.

OLED stands for Organic Light Emitting Diode. Unlike QLED, OLED is an emissive technology where each pixel generates its own light. This means no separate backlight is required, allowing for perfect black levels and exceptional contrast. LG has been the dominant OLED panel maker for years, with Sony and Panasonic also offering OLED models. Recently, Samsung introduced QD-OLED, a hybrid that combines quantum dots with OLED panels for improved brightness and color.

Image Quality Showdown

The most significant difference lies in picture quality. OLED TVs excel in contrast because they can turn off individual pixels completely, resulting in true blacks and infinite contrast ratio. This makes them ideal for dark room viewing and high dynamic range (HDR) content. In side-by-side comparisons, OLED consistently delivers more lifelike images, especially in scenes with deep shadows or bright highlights.

QLED TVs, on the other hand, are generally brighter than OLED. The brightest QLED models can produce higher peak luminance, which is beneficial in rooms with abundant ambient light. However, even the best QLED sets struggle with black levels compared to OLED. Recent innovations like mini-LED backlighting and local dimming have narrowed the gap, but OLED remains superior for contrast.

Viewing Angles and Uniformity

OLED screens maintain color and contrast from almost any viewing angle. LCD-based QLED TVs often experience color shifting and loss of contrast when viewed from the side. This makes OLED a better choice for large living rooms where viewers may sit off-center. Additionally, OLED panels exhibit excellent screen uniformity, with no clouding or backlight bleed that can affect LCD displays.

Burn-In Risk

One commonly cited concern for OLED is burn-in, where static elements like channel logos or scoreboards leave permanent ghost images. While OLED is more susceptible than QLED, modern OLED TVs include pixel shifting and other mitigation techniques. For most users who vary content regularly, burn-in is unlikely. QLED TVs do not suffer from burn-in, making them safer for extended viewing of static content such as news channels or video games with persistent HUDs.

Size and Price Considerations

QLED TVs are available in a wider range of sizes, from as small as 32 inches up to 115 inches. They also tend to be cheaper per inch, especially in larger sizes. For example, a 75-inch QLED TV often costs significantly less than a 77-inch OLED model. This makes QLED a more budget-friendly option for buyers wanting a big screen.

OLED TVs come in sizes from 42 to 97 inches, though the most common are 55, 65, and 77 inches. Prices have dropped over the years but remain higher than comparable QLED sets. The premium for OLED is justified by its superior picture quality, particularly for cinephiles and gamers who prioritize contrast and response times.

Future Developments

TV technology continues to evolve. Samsung is researching direct-view emissive quantum dot displays that could combine OLED's perfect blacks with even higher brightness and efficiency, without the risk of burn-in. MicroLED is another emerging technology offering similar benefits but at a much higher cost, currently targeting the ultra-luxury market. For now, OLED remains the champion of picture quality, while QLED offers excellent value and brightness for most consumers.

Both QLED and OLED are smart TVs with built-in streaming apps and support for modern standards like HDR, 4K, and 120Hz refresh rates. The choice ultimately depends on your viewing environment, budget, and prioritization of image quality versus brightness. For the best possible picture, OLED is the winner; for a bright, affordable large screen, QLED is often the better option.

Source: CNET News

Why the Urgency?

Google is ending support for the POP3 protocol later this year. New users lost access to POP3 in the first quarter of 2026, but existing users can still use it until the shutdown. Once POP3 is gone, transferring emails between accounts will require more complex methods or third-party tools. That's why now is the time to act.

How Much Data Can You Store?

The 15GB free storage is shared across Gmail, Google Drive, and Google Photos. Large attachments, photos, and documents consume space quickly. When the storage fills up, you cannot send or receive emails until you free up space. You could upgrade to Google One (100GB for $20/year), but why pay for old emails? Deleting messages manually is tedious. The nuclear option is to move everything to a new account.

Step-by-Step Transfer Process

Before starting, back up your emails using Google Takeout. On a test account with 75,000 messages, the download took about two hours. After backup, follow these steps:

From your old Gmail account:

• Go to Settings → See all settings → Forwarding POP/IMAP tab.
• Select "Enable POP for all mail".
• Under "When messages are accessed with POP", choose "delete Gmail's copy" to automatically remove them after transfer.
• Save changes.

Create a new Gmail account (your archive):

• Log into the new account → Settings → See all settings → Accounts and Import tab.
• Next to "Check mail from other accounts", click "Add a mail account".
• Enter your old Gmail address, select "Import emails from my other account (POP3)".
• Enter the password. If standard password fails, create a Google app password (requires 2-Step Verification) at myaccount.google.com/apppasswords.
• Use port 995, check boxes for SSL, label incoming messages, and archive them (skip Inbox).
• Click Add Account. Transfer may take hours or days depending on volume.

What to Expect After Transfer

Once completed, your old Gmail will place all transferred messages in Trash. Empty Trash manually (75,000 messages took about an hour). In testing, storage usage dropped from 12GB to 0.66GB. Not all messages transfer—Drafts and Spam are not included. Handle those separately. After transfer, stop the automatic process by deleting the linked account in your new Gmail's settings. Also delete the app password if you created one.

Important Notes

Transferred messages are labeled and archived in the new account. You can keep the archive account active by logging in at least once every two years; otherwise Google may delete it. This method works now, but once POP3 is gone, you'll need alternatives like manual forwarding or third-party migration tools. Don't procrastinate—reclaim your 15GB of free storage today.

Source: CNET News

Das Wichtigste im Überblick

• Das Bero Padel Classic fand am 30. April 2026 in Sherman Oaks, Kalifornien, statt.
• Tom Holland, Initiator des Events, präsentierte dort seine alkoholfreie Biermarke Bero.
• Tom Daley überraschte Holland mit einem selbstgehäkelten Getränkehalter im Spinnennetz-Look.
• Auch Hollywoodstar Zendaya war vor Ort und unterstützte Holland auf und neben dem Platz.
• Die Aktion sorgte auf Instagram für einen Hype und zahlreiche witzige Kommentare.

Britische Starpower auf Kaliforniens Courts

Das Bero Padel Classic, das am 30. April 2026 stattfand, war weit mehr als nur ein lockeres Sportevent. Es war eine Bühne für zwei der bekanntesten britischen Stars der Gegenwart. Tom Holland, der nicht nur als Spider-Man weltberühmt ist, sondern auch als Unternehmer seine alkoholfreie Biermarke Bero promotet, nutzte die Veranstaltung, um sein Produkt einem prominenten Publikum vorzustellen. Dass ausgerechnet Tom Daley, der olympische Turmspringer und bekennende Strick-Enthusiast, mit einem gehäkelten Geschenk auftauchte, verlieh dem Event eine unerwartete, aber charmante Wendung.

Die Stimmung war ausgelassen. Während die Gäste auf den Padel-Courts ihre Schläger schwangen, zog Daley mit seinem selbstgebastelten Accessoire die Aufmerksamkeit auf sich. Der Getränkehalter, der an ein Spinnennetz erinnerte, war nicht nur ein praktisches Utensil, sondern auch ein Statement. Daley, der seit Jahren für seine Strick- und Häkelkunst bekannt ist, hatte damit eine perfekte Brücke zwischen seiner Leidenschaft und Hollands Filmkarriere geschlagen. Holland, der sich selbst als „Strick-Anfänger“ bezeichnet, war sichtlich begeistert und posierte stolz mit dem Geschenk für die Kameras.

Häkelkunst trifft Hollywood-Charme

Die Chemie zwischen den beiden Toms war unübersehbar. Daley, der normalerweise eher zurückhaltend auftritt, zeigte sich in Gesellschaft von Podcaster Jay Shetty und TikTok-Star Ollie Muhl ungewohnt entspannt. Doch der wahre Aufmerksamkeitstrudel entstand durch Daleys Instagram-Posts mit Holland. In einem Video ist zu sehen, wie Holland den gehäkelten Getränkehalter in die Höhe hält und dazu eine spöttische Spider-Man-Pose einnimmt. Die Kommentare unter dem Beitrag spiegelten die Begeisterung der Fans wider: „British twunk crossover of the millennium“ war nur einer der vielen humorvollen Kommentare. Andere Nutzer forderten scherzhaft, die beiden sollten sich küssen. Der Hashtag TomDaleyTomHolland trendete kurzzeitig auf der Plattform.

„Ich liebe Häkeln, weil ich dabei an nichts anderes denken muss und richtig runterkomme“, verriet Tom Holland ehrlich im Gespräch mit Daley. Diese Aussage überraschte viele, die Holland nur aus actiongeladenen Filmen kennen. Doch der Schauspieler betonte, dass er und seine Partnerin Zendaya zu Hause regelmäßig zur Häkelnadel greifen. Diese gemeinsame Leidenschaft verband die beiden Stars noch mehr und verlieh dem Event eine persönliche Note.

Zendaya als Massageprofi im Hintergrund

Wer dachte, dass sich Zendaya bei diesem Promi-Event zurückhalten würde, lag falsch. Die Schauspielerin und Sängerin wurde dabei erwischt, wie sie Tom Holland mit einer Massagepistole behandelt. Ein weiterer Schnappschuss, der auf Social Media für ordentlich Wirbel sorgte. Fans spekulierten sofort über den Gesundheitszustand des Schauspielers, doch es stellte sich heraus, dass es sich um eine routinemäßige Erholungsmaßnahme nach einem intensiven Padel-Match handelte. Zendaya zeigte sich als fürsorgliche Partnerin, die Holland sowohl auf als auch neben dem Platz unterstützt.

Die Gerüchteküche um eine mögliche Hochzeit des Paares brodelt indes weiter. Offizielle Bestätigungen gibt es nicht, doch die beiden lassen sich immer wieder gemeinsam bei öffentlichen Anlässen blicken. Das Bero Padel Classic war keine Ausnahme, und die Bilder von Holland und Zendaya gingen viral. Die Frage, ob und wann sie sich das Jawort gegeben haben, bleibt offen – das Netz spekuliert munter weiter.

Hintergrund: Tom Daley – vom Turm zum Strickstar

Tom Daley, geboren 1994 in Plymouth, England, ist nicht nur für seine sportlichen Erfolge bekannt. Bei den Olympischen Spielen 2020 in Tokio gewann er Gold im Synchronspringen und Bronze im Einzelspringen. Doch abseits des Beckens hat er sich eine treue Fangemeinde mit seinen Strick- und Häkelprojekten aufgebaut. Auf Instagram teilt er regelmäßig Fotos und Videos von seinen Kreationen – von Pullovern über Mützen bis hin zu komplizierten Mustern. Daley hat sogar ein Buch über seine Strickleidenschaft veröffentlicht und eine eigene Kollektion mit dem Label „Made with Love by Tom Daley“ herausgebracht. Seine Fähigkeit, Sport und Handarbeit zu verbinden, hat ihn zu einem Vorbild für viele Menschen gemacht, die traditionell männlich dominierte Hobbys hinterfragen.

Sein Auftritt beim Bero Padel Classic ist ein weiteres Beispiel dafür, wie Daley seine Leidenschaften vereint. Indem er Holland einen gehäkelten Getränkehalter schenkte, zeigte er nicht nur seine handwerkliche Finesse, sondern auch seinen Humor und sein Gespür für Popkultur. Die Aktion wurde von den Medien als „Crossover der Superlative“ gefeiert und unterstreicht Daleys Status als kulturelle Ikone.

Hintergrund: Tom Holland – vom Tänzer zum Superhelden

Tom Holland, 1996 in London geboren, begann seine Karriere als Tänzer und Schauspieler im Musical „Billy Elliot“. Der Durchbruch gelang ihm 2016 mit der Rolle des Spider-Man im Marvel Cinematic Universe. Seitdem hat er sich als einer der beliebtesten Schauspieler Hollywoods etabliert. Neben seiner Filmkarriere hat Holland auch geschäftliches Talent bewiesen: Mit der Marke Bero brachte er ein alkoholfreies Bier auf den Markt, das auf Events wie dem Padel Classic beworben wird. Holland ist bekannt für seinen bodenständigen Charakter und sein Engagement für wohltätige Zwecke.

Seine Begeisterung für Handarbeiten mag überraschen, passt aber zu seinem Image als vielseitiger Künstler. Indem er sich öffentlich zum Häkeln bekennt, bricht er mit Klischees und zeigt, dass Männer ebenso kreativ sein können. Die Begegnung mit Daley war daher nicht nur ein freundschaftlicher Austausch, sondern auch eine Bestätigung dieser Haltung. Holland erklärte, dass er durch Daley inspiriert wurde, mehr zu häkeln, und dass das Spinnennetz ein willkommenes Accessoire für zukünftige Bero-Werbungen sein könnte.

Padel: Der Trendsport der Stars

Das Event fand im Zeichen des Padel-Tennis statt, einer Sportart, die in den USA und Europa immer beliebter wird. Padel ist eine Mischung aus Tennis und Squash, die auf einem kleineren Feld gespielt wird und besonders bei Prominenten und Sportlern Anklang findet. Das Bero Padel Classic, benannt nach Hollands Biermarke, zog zahlreiche Gäste an, darunter Schauspieler, Musiker und Influencer. Die Mischung aus Sport, Prominenz und social Media sorgte für eine hohe Reichweite und machte das Event zu einem der meistdiskutierten des Jahres.

Tom Daley, der selbst begeisterter Padel-Spieler ist, nutzte die Gelegenheit, um sein Können auf dem Platz zu zeigen. Er trat in mehreren Matches an und bewies, dass er nicht nur im Wasser, sondern auch auf dem Court eine gute Figur macht. Holland hingegen konzentrierte sich mehr auf die Organisation und das Networking, ließ es sich aber nicht nehmen, ein paar Bälle zu schlagen. Die Dynamik zwischen den Gästen war durchweg positiv, und die Veranstaltung wurde als großer Erfolg gewertet.

Instagram-Hype: Die Macht der sozialen Medien

Die Posts von Daley und Holland auf Instagram erzielten innerhalb weniger Stunden Millionen von Aufrufen. Die Kombination aus britischem Humor, Handarbeit und Superhelden-Kultur erwies sich als viraler Hit. Fans kommentierten und teilten die Bilder tausendfach, und Medien weltweit berichteten über das ungewöhnliche Crossover. Die Aktion zeigt, wie wichtig soziale Medien für die Vermarktung von Events und Persönlichkeiten sind. Daley und Holland haben erkannt, dass authentische, humorvolle Inhalte oft mehr bewirken als klassische Werbung.

Das Spinnennetz symbolisiert dabei mehr als nur eine Filmrolle. Es steht für die Verbindung von Kreativität und Sport, von Tradition und Moderne. Daley hat mit seiner Häkelkunst gezeigt, dass Handarbeit im digitalen Zeitalter eine neue Bedeutung bekommen kann – nicht nur als Hobby, sondern als Ausdruck von Persönlichkeit und als Marketinginstrument.

Die Frage, ob es bald Häkel-Kurse mit den beiden oder eine limitierte „Spidey-Halter“-Kollektion geben wird, bleibt offen. Aber eines ist sicher: Das Bero Padel Classic wird in Erinnerung bleiben als das Event, bei dem Spider-Man und ein Olympionike gemeinsam am Netz zauberten. Instagram wartet schon ungeduldig auf das nächste Crossover.

Source: SCHWULISSIMO.de News

While the pair may not engage in genuine arguments, Billie acknowledged that they do have heated disagreements from time to time. “We can get into the most intense argument you’ve ever heard in your life, and five minutes later we’re cracking jokes and making music,” she explained. This dynamic is a cornerstone of their relationship, blending fierce creative tension with unwavering mutual support. Finneas himself told Elle that he is not particularly fond of touring—though he loves performing—and that he misses his sister deeply while she’s on the road.

The strength of their bond was on full display in a trailer for the upcoming tour film. In it, Billie tearfully read a note from Finneas that said, “Good luck tonight, but you don’t need it—nobody does it like you.” The moment highlighted the emotional foundation of their partnership, which has produced ten Grammy Awards, including a recent Song of the Year win for “Wildflower.”

Their musical journey began when Billie was just a child. Originally aspiring to be a professional dancer, a severe hip injury forced her to abandon that dream and pivot to music. Finneas, already a budding musician, began writing and producing songs with her in their childhood home. Their first collaboration, “Ocean Eyes,” went viral and launched Billie’s career. Since then, the siblings have become one of the most successful duos in modern pop, with Finneas producing and co-writing almost all of Billie’s material.

Despite their close working relationship, the decision for Finneas not to join the “Hit Me Hard and Soft” tour was practical. “He doesn’t love being away from home for months, and I totally get it,” Billie said. “We both have our own lives and projects now. But that doesn’t mean we’re not still a team.” Finneas echoed this, adding that he remains heavily involved in Billie’s creative process from behind the scenes, often sending her ideas and feedback remotely.

Their individual pursuits have also flourished. Finneas has released his own solo music, won an Oscar for producing the James Bond theme “No Time to Die,” and has collaborated with artists like the Kid Laroi and Olivia Rodrigo. Billie, meanwhile, has expanded her acting resume, taking on roles in television and film. Yet they always return to each other. “She’s my best friend,” Finneas said. “The whole world can think what they want, but we know the truth.”

Fans have at times speculated about a rivalry given their separate career paths, but both insist their bond is unbreakable. In the interview, they recalled how their creative process often starts with a simple loop or a bedroom recording session, just like the old days. Billie added that she values Finneas’s honesty above all: “He’s the only person who will tell me when something sucks, and I need that.”

The duo’s history is filled with milestones. Their debut album, “When We All Fall Asleep, Where Do We Go?” won multiple Grammys, including Album of the Year and Record of the Year for “Bad Guy.” Their follow-up, “Happier Than Ever,” broke streaming records and earned further accolades. “Hit Me Hard and Soft,” their fourth studio album, spawned the hit “Wildflower,” which dominated charts and won Song of the Year at the 2026 Grammys.

Billie’s personal evolution has also been a central theme in her music. She has spoken openly about mental health, body image, and her relationship with fame. Finneas has been a constant anchor, helping her navigate the pressures of stardom. “I wouldn’t be here without him,” Billie said. “He’s not just my brother; he’s my everything.”

The tour separation, while fueling rumors, has allowed both to grow individually. Finneas spent the time producing for other artists and working on his own projects, while Billie connected with fans in a more intimate concert setting. She noted that the tour included stripped-down performances and smaller venues, a departure from her previous arena shows. “It was scary at first, but it ended up being really special,” she said.

Looking ahead, the siblings are already planning their next collaborative album. “We have a few ideas,” Finneas teased. “Nothing concrete yet, but I’m always writing for her.” Billie added that they expect to reunite in the studio later this year. “We can’t stay apart for too long. It’s like a gravitational pull.”

In an industry where family partnerships often break under pressure, Billie Eilish and Finneas stand out as a rare example of resilience and love. Their story resonates with fans who see them as not just artists, but as a real-life example of what it means to support each other unconditionally. As Billie put it, “We’re not just collaborators. We’re family. And that’s forever.”

The Elle interview offers a deeper look into their dynamic, beyond the rumors and headlines. It reveals a partnership built on trust, humor, and an unshakeable belief in each other’s talent. Whether on stage together or apart, Billie Eilish and Finneas O’Connell continue to write their legacy—one note, one fight, one hug at a time.

Source: Promiflash.de News

Das Szenario RCP8.5 galt lange als das Doomsday-Szenario der Klimaforschung. Es sah vor, dass die globale Durchschnittstemperatur um bis zu 4,8 Grad ansteigt, was zu verheerenden Folgen wie Massenmigration, wirtschaftlichen Schäden in Billionenhöhe und einem Zusammenbruch von Ökosystemen geführt hätte. Nun hat der IPCC eingeräumt, dass diese Entwicklung extrem unwahrscheinlich ist. Stattdessen gilt das Szenario RCP4.5 mit einem Anstieg um etwa drei Grad als wahrscheinlicher. Auch dies ist erheblich, aber kein Weltuntergang.

Deutsche Klimapolitik auf dem Prüfstand

Besonders in Deutschland und der Europäischen Union beruhen weite Teile der Klimapolitik auf den Annahmen des RCP8.5-Szenarios. Dazu gehören die Senkung des Energieverbrauchs um 26,5 Prozent bis 2030 sowie die Reduzierung der CO₂-Emissionen auf null bis 2050, in Deutschland sogar bis 2045. Diese Maßnahmen haben erhebliche wirtschaftliche Kosten verursacht, das Wachstum gebremst und teilweise den Aufstieg populistischer Parteien wie der AfD befördert.

Kritiker argumentieren, dass nun eine Revision dieser Politik möglich sei. Klimaschutz müsse zwar ein wichtiges Ziel bleiben, aber nicht mehr der absolute Primat, dem alles andere untergeordnet wird. Die ursprüngliche Argumentation fußte auf der Annahme, dass die Menschheit vor einem unmittelbaren Kollaps stehe – eine Annahme, die sich als übertrieben herausgestellt hat.

Allerdings fand dieser Kurswechsel in den deutschsprachigen Medien nur wenig Widerhall. Außer in der „Welt“, dem „Cicero“ und der „Bild“ wurde kaum über die Rücknahme von RCP8.5 berichtet. Auch die Politik zeigte sich nicht erleichtert. Das liegt nach Ansicht von Beobachtern daran, dass viele Akteure in Medien, NGOs und Politik ihre Karrieren auf diesen Horrorszenarien aufgebaut haben. Sie seien zu „Zeugen Gretas“ geworden – in Anlehnung an die schwedische Aktivistin Greta Thunberg, die mit hysterischen Auftritten und radikalen Forderungen bekannt wurde.

Die Rolle von Greta Thunberg und die ideologische Färbung

Greta Thunberg selbst hatte stets betont, dass sie nur der Wissenschaft folge. Ihre „Fridays for Future“-Bewegung forderte einen sofortigen und radikalen Umbau der Gesellschaft. Doch die Wissenschaft hat sich weiterentwickelt. Während Thunberg und ihre Anhänger weiterhin auf die düstersten Prognosen verweisen, hat der IPCC selbst die Risiken nach unten korrigiert. Dies führte zu einer Spaltung: Auf der einen Seite stehen diejenigen, die an den alten Horrorszenarien festhalten, auf der anderen Seite diejenigen, die eine differenziertere Betrachtung fordern.

Die Bezeichnung „Zeugen Gretas“ spielt auf die Zeugen Jehovas an, die ebenfalls mit apokalyptischen Prophezeiungen begannen, die nie eintraten. Heute stehen sie in Fußgängerzonen und wirken oft wie Relikte einer vergangenen Zeit. Die Analogie soll verdeutlichen, dass die radikale Klimabewegung Gefahr läuft, sich selbst zu isolieren, wenn sie nicht bereit ist, ihre Positionen zu aktualisieren.

Internationale Vergleiche: Technische Lösungen statt Radikalumbau

Andere Länder verfolgen einen pragmatischeren Ansatz. Die USA senkten unter Präsident Obama ihre CO₂-Emissionen, indem sie von Kohle auf Gas umstiegen. China strebt bis 2060 eine Energieversorgung an, die zu 80 Prozent aus Erneuerbaren und zu 20 Prozent aus Kernkraft besteht. Beide Länder setzen auf technische Lösungen, ohne die gesamte Wirtschaftsordnung infrage zu stellen. Deutschland und Europa hingegen haben einen Kurs eingeschlagen, der in den Augen vieler Ökonomen Wohlstand vernichtet und Extremisten stärkt.

Die Rücknahme von RCP8.5 ist ein wichtiger Einschnitt. Sie zeigt, dass die Klimaforschung selbstkritisch genug ist, um Modelle zu korrigieren, wenn die Realität anders verläuft. Diese Fähigkeit zur Selbstkorrektur ist eine Stärke der Wissenschaft. Die Politik sollte dieser Entwicklung folgen und ihre Maßnahmen an die neue Erkenntnislage anpassen.

Es bleibt abzuwarten, ob die deutsche Bundesregierung und die EU-Kommission den Mut aufbringen, ihre ehrgeizigen, aber teuren Klimaziele zu überprüfen. Bisher signalisieren sie eher Kontinuität. Die Diskussion um den richtigen Weg wird die politische Landschaft noch lange prägen.

Source: Ruhrbarone News

The Sighting: A Night Out in LA

The outing took place in the bustling streets of Los Angeles, a city accustomed to celebrity sightings. However, the combination of the two famous couples — one rumored, one confirmed — created a media frenzy. Jacob Elordi, best known for his roles in Euphoria and Saltburn, was driving while Kendall, the supermodel and reality star, sat beside him. In the back, Kylie, the cosmetics mogul, and Timothée, the Oscar-nominated actor of Dune fame, appeared relaxed and affectionate. The presence of Medrano, a renowned fashion photographer, added an element of creative camaraderie to the group.

Witnesses noted that the atmosphere was light-hearted. All four seemed to be enjoying each other's company, laughing and chatting as they navigated the city. The images captured a rare moment of unguarded celebrity interaction, free from the usual staged red-carpet poses. For fans, it was a glimpse into the private lives of individuals who are often guarded about their personal relationships.

Comparisons to Y2K Paparazzi Culture

Social media users were quick to draw parallels between this new set of photos and a legendary paparazzi moment from 2006. That year, Britney Spears, Lindsay Lohan, and Paris Hilton were photographed together in a convertible, laughing and riding through the streets of Los Angeles. The image became an iconic symbol of mid-2000s celebrity culture — a time when tabloids thrived on candid shots of young stars living recklessly. The 2025 snapshot of the Jenners, Elordi, and Chalamet evoked that same sense of carefree rebellion, albeit with a modern twist.

The comparison highlights how celebrity dynamics have evolved. While Britney, Lindsay, and Paris were often portrayed as 'it girls' navigating fame's chaotic early days, Kendall, Kylie, Jacob, and Timothée represent a new generation of stars who wield more control over their narratives through direct social media engagement. Yet the candid paparazzi moment remains a timeless fascination, reminding the public that even the most polished celebrities have unguarded moments.

The Rumored Couple: Kendall and Jacob

The most compelling aspect of the sighting is the continued speculation about a romantic relationship between Kendall Jenner and Jacob Elordi. The two were first linked in March 2025, when they attended the Vanity Fair Oscar Party together. Paparazzi captured them in close conversation, and insiders hinted that they had been seeing each other quietly. Later, they were spotted at the Coachella music festival, where they appeared to be especially comfortable in each other's presence.

Since then, both have remained silent about the nature of their relationship. Kendall has a long-established pattern of keeping her love life out of the spotlight. She previously dated NBA player Devin Booker and Latin music star Bad Bunny, but rarely spoke publicly about those relationships. Jacob, on the other hand, has been linked to several costars, including Zendaya and Olivia Jade, but has never confirmed a serious relationship publicly. According to reports, Jacob is not a fan of reality TV culture, which could be a complicating factor given Kendall's deep roots in the Keeping Up with the Kardashians franchise.

Despite the silence, the chemistry between them is hard to ignore. Body language experts analyzing the latest photos note that Kendall's relaxed posture in the passenger seat and their shared smiles suggest intimacy. Whether this is a casual fling or a budding romance remains to be seen, but the rumour mill shows no signs of slowing down.

The Established Couple: Kylie and Timothée

In contrast, Kylie Jenner and Timothée Chalamet have been a known quantity for over three years. They began dating in early 2023 and have since been photographed at numerous events, including the 2025 Vanity Fair Oscar Party. Their relationship has been characterized by low-key public appearances and occasional Instagram posts, but both have avoided oversharing. Kylie, who was previously engaged to rapper Travis Scott, seems to have found a stable partnership with the actor. Timothée, known for his privacy, has adapted to the intense scrutiny that comes with dating a Kardashian-Jenner. Fans have noted that the couple appears genuinely happy, and their presence on this double date only confirms their comfort with each other.

In the photos, Kylie and Timothée are seen in the back seat, smiling and seemingly engaged in conversation with the front passengers. Their body language indicates a settled, affectionate bond. For many, it's a refreshing reminder that even in the whirlwind of Hollywood, enduring relationships can flourish.

A Look at Kendall's Relationship History

To understand the significance of this potential new relationship, it helps to review Kendall's romantic past. She has often been linked to high-profile athletes and musicians. Her most serious relationship to date was with Devin Booker, a star player for the Phoenix Suns. The two dated on and off from 2020 to 2022, and even lived together for a time. However, they ultimately parted ways, with sources citing conflicting schedules and different priorities. After Booker, Kendall was involved with Bad Bunny, the Puerto Rican reggaeton sensation. That relationship was more private, though they were spotted at various events before reportedly ending in early 2024.

Kendall has stated in interviews that she values her independence and isn't in a rush to settle down. In a 2025 interview with Vogue France, she said: "The most important thing is to stay true to myself and continue having a good time. That might sound like a cliché, but it's very important to me." She also mentioned that while she wants children someday, it's not an immediate priority: "I want to make sure I can dedicate a lot of time to them, and right now I'm too busy for that."

Jacob Elordi, for his part, has his own career trajectory. Rising to fame through The Kissing Booth trilogy, he later earned critical acclaim for his role in Euphoria and the film Saltburn. He has spoken about the challenges of fame and his desire for authenticity. Whether he and Kendall are a match beyond the surface level is a question that only time can answer.

Why the Silence? The Celebrity Code

Neither Kendall nor Jacob has commented on the double date or their relationship status. This silence is strategic. In an era where every social media post is dissected, many celebrities choose to protect their privacy by saying nothing. Kendall, in particular, has mastered this art. She rarely confirms or denies rumors, allowing speculation to fuel public interest without committing to a narrative. This approach keeps fans guessing and maintains a sense of mystery.

Jacob's reticence might also stem from his personal dislike of reality TV. Growing up in Australia, he was not immersed in the Kardashian phenomenon, and sources close to him suggest he finds the constant attention overwhelming. He prefers to let his work speak for itself. For now, the pair seems content to enjoy each other's company away from the glare of official confirmation.

The viral nature of the photos, driven by accounts like DeuxMoi, shows that the public's appetite for celebrity gossip remains insatiable. The comparison to the 2006 Britney-Lindsay-Paris photo underscores a cultural shift: while that earlier image was seen as a sign of wild youth, the 2025 version is interpreted through a lens of curated relationships and brand management. Yet at the core, the fascination is the same — a desire to see icons as relatable humans.

Kendall Jenner's words from the Vogue France interview resonate here: she wants to keep things "light and easy" in 2026. Whether that involves a full-fledged romance with Jacob Elordi or simply a close friendship, the Jenner sisters and their circle continue to captivate audiences worldwide. The double date may be just one night in LA, but it has already become a defining pop culture moment of the year.

Source: Promiflash.de News

Wien – Seit 15 Jahren verkörpert Adele Neuhauser (67) die beliebte Ermittlerin Bibi im „Tatort“ aus Wien. Die Fans lieben ihre coole Art und ihr Gespür für Gerechtigkeit. Doch nun zeigt die österreichische Schauspielerin eine völlig neue Seite: Im ZDF-Zweiteiler „Mama ist die Best(i)e“ (Montag, 20.15 Uhr) schlüpft sie in die Rolle der verurteilten Mörderin Gloria Almeda. Eine Rolle, die weit entfernt ist von den moralisch einwandfreien Ermittlern, die sie so lange geprägt hat.

Die Handlung: Gloria, einst steinreiche Society-Lady, sitzt im Gefängnis – verurteilt für den Mord an ihrem Ehemann Viktor. Doch sie beteuert ihre Unschuld. Nach ihrer Entlassung kehrt sie mit einer elektronischen Fußfessel in ihr eigenes Schloss zurück und beginnt, ihre Familie zu verdächtigen. Denn Viktor war ein Despot, den wirklich jeder hasste – Kinder, Verwandte, Angestellte. Alle hatten ein Motiv. Gloria ist entschlossen, den wahren Täter zu finden, und geht dabei gnadenlos vor.

Neuhauser beschreibt ihre Figur als „wunderbar schrullig“ und als „Bestie in der Vergangenheit, die nach der Haft die Chance hat, die Beste zu werden“. Diese Ambivalenz fasziniert die Schauspielerin. „Hand aufs Herz: Die Gloria-Charaktereigenschaften sind mir privat fremd. Aber sie regen durchaus meine Fantasie an“, erklärt sie. Die dunklen Seiten der Geschichte – Intrigen, Verrat, Geldgier – seien ein toxischer Mix, der sie reize. Für Neuhauser ist der Zweiteiler wie ein Krimi-Spiel „Cluedo“, bei dem jeder der Mörder sein könnte.

Hintergrund: Adele Neuhausers Karriere

Adele Neuhauser wurde am 17. Januar 1959 in Athen geboren, wuchs aber in Wien auf. Ihre Schauspielkarriere begann sie am Theater, bevor sie im Fernsehen bekannt wurde. Internationale Bekanntheit erlangte sie durch ihre Rolle als Inspektorin Bibi Fellner im Wiener „Tatort“, den sie seit 2006 gemeinsam mit Harald Krassnitzer (als Moritz Eisner) spielt. Die Reihe zählt zu den erfolgreichsten des deutschen Krimis, und Neuhauser wurde mehrfach ausgezeichnet, unter anderem mit dem Deutschen Fernsehpreis. Doch 2025 gab sie bekannt, dass sie den „Tatort“ nach 15 Jahren verlassen wird – eine Entscheidung, die sie aus freien Stücken traf. „Ich möchte mich neuen Herausforderungen widmen“, sagte sie damals. Der Zweiteiler „Mama ist die Best(i)e“ ist einer dieser Schritte ins Neuland.

Die Schauspielerin hat sich immer wieder als wandlungsfähig erwiesen. Neben Krimis spielte sie auch in Komödien und Dramen. Besonders schätzt sie Rollen, die ihr erlauben, die dunklen Abgründe menschlicher Natur zu erkunden. Im Gespräch mit BILD betonte sie: „Ich liebe es, wenn eine Figur nicht nur gut oder böse ist, sondern beides in sich vereint. Gloria ist genau das: Eine Frau, die aus Not zu extremen Mitteln greift, aber dennoch eine gewisse Verletzlichkeit zeigt.“

Die Handlung von „Mama ist die Best(i)e“

Der Zweiteiler spielt auf einem abgelegenen Schloss in Österreich, das Gloria nach ihrer Entlassung aus dem Gefängnis bewohnt. Die Familie hat sich nach Viktors Tod zerstritten, und jeder scheint etwas zu verbergen. Glorias Kinder – ein erfolgreicher Geschäftsmann und eine etwas naive Tochter – misstrauen ihr. Auch die Schwester des Verstorbenen taucht auf und verdächtigt Gloria. Unterdessen versucht Gloria, mit Hilfe ihrer ehemaligen Mithäftlinge – der gewalttätigen Maria (gespielt von Edita Malovčić) und der klugen Henny (Lara Mandoki) – die Wahrheit ans Licht zu bringen. Die drei Frauen haben unterschiedliche Motive, und die Dynamik zwischen ihnen sorgt für spannende Wendungen.

Neuhauser lobt die schwarze Komödie als „erfrischend anders“. Anders als klassische Krimis, in denen die Ermittler stets die Guten sind, steht hier die vermeintliche Täterin im Mittelpunkt. Der Zuschauer wird mit ihrer Perspektive konfrontiert und fragt sich: Ist Gloria wirklich schuldig? Oder nur das Opfer einer Intrige? Die Inszenierung nutzt diese Unsicherheit geschickt aus und lässt die Zuschauer mitfiebern.

Neue Rollen nach dem „Tatort“-Aus

Für Neuhauser bedeutet der Zweiteiler einen Neuanfang. Nach dem freiwilligen Abschied vom „Tatort“ hat sie mehr Zeit für andere Projekte. Sie plant, sich künftig auf Rollen mit moralischer Ambivalenz zu konzentrieren. „Es gibt so viele interessante Geschichten, die darauf warten, erzählt zu werden. Ich möchte nicht nur die Heldin spielen, sondern auch die Frau, die Fehler macht und sich irrt.“ Dazu gehören auch die dunklen Seiten des Lebens: „Intrigen, Verrat, Geldgier – das sind Themen, die mich faszinieren, weil sie so menschlich sind.“

Auch privat legt Neuhauser den Fokus neu. Sie ist stolze Großmutter zweier Enkelkinder und möchte sich mehr Zeit für die Familie nehmen. „Die Rolle als Oma ist für mich die wichtigste im Leben“, sagt sie. Doch sie betont, dass die Schauspielerei weiterhin eine Leidenschaft bleibt. „Ich werde nicht aufhören, aber ich wähle bewusster aus.“

Was würde Neuhauser ins Gefängnis mitnehmen?

Auf die Frage, was sie selbst mit in den Knast nehmen würde, antwortet sie konkret: „Das Buch von Dimitré Dinev, `Zeit der Mutigen’.“ Ein Bestseller über die Kernfrage: Was macht den Menschen aus? Wie übersteht er Jahre der Unterdrückung und Gewalt? Dazu einen angenehmen Pyjama und einen guten „Polster“ – das österreichische Wort für Kissen. Diese Antwort zeigt ihre pragmatische und dennoch tiefgründige Art. Sie möchte im Gefängnis nicht nur überleben, sondern auch wachsen.

Der Film „Mama ist die Best(i)e“ ist nicht nur ein Kriminalfall, sondern auch eine schwarze Komödie über Familiengeheimnisse, Macht und Vergeltung. Adele Neuhauser spielt die Rolle mit sichtlicher Freude an der Boshaftigkeit, ohne dabei die menschliche Seite zu vernachlässigen. Am Montag werden Millionen Zuschauer sie in dieser neuen Rolle sehen – und vielleicht eine Seite an ihr entdecken, die sie bisher nicht kannten.

Der Zweiteiler läuft am Montag um 20.15 Uhr im ZDF. Wiederholungen sind in der Mediathek verfügbar. Die Besetzung umfasst neben Neuhauser auch Harald Krassnitzer in einer kleinen Gastrolle sowie Edita Malovčić, Lara Mandoki und weitere bekannte Gesichter. Der Regisseur, Markus Imboden, hat bereits mehrere erfolgreiche Krimis inszeniert und betont, dass der Film bewusst mit den Erwartungen des Publikums spielt: „Es geht darum, Vorurteile zu hinterfragen. Ist eine verurteilte Mörderin automatisch böse? Oder kann sie auch das Opfer sein?“ Für Neuhauser ist diese Frage der Kern der Geschichte: „Gloria gibt uns die Chance, über Schuld und Unschuld nachzudenken. Und dabei können wir auch über uns selbst lachen.“

Adele Neuhauser hat mit dieser Rolle bewiesen, dass sie mehr kann als nur die Ermittlerin. Ihre schauspielerische Bandbreite reicht von komödiantisch bis tief dramatisch. Der Zweiteiler ist ein gelungener Beweis dafür, dass sie auch ohne „Tatort“ weiterhin eine feste Größe im deutschen Fernsehen sein wird.

Source: bild.de News

A user on X has shared a screenshot suggesting Google may be testing stricter usage tracking and possible weekly limits inside Gemini. The screenshot shows a new section that explains, “Plan limits determine how much you can use Gemini over time.” This means Google could be preparing a more aggressive system that measures how frequently free users interact with Gemini, especially when using heavier AI models.

The screenshot also includes a usage bar that tracks how much of the quota has already been consumed. In this particular case, the user had reportedly used around 5% of the available allowance, with the limit resetting later in the day. While that may not sound alarming yet, it does point toward Gemini becoming far more structured about how much free access people actually get.

This was always inevitable

Running large AI models is absurdly expensive. Every prompt, generated image, or long conversation costs money in computing power, and tech companies have spent the last few years conditioning users to expect near-unlimited AI for free. That honeymoon phase was never going to last forever. Google, like practically every other AI company right now, ultimately wants people to pay for premium access. The challenge is figuring out how hard it can push before users simply move elsewhere. Because, unlike traditional software lock-ins, AI tools are painfully easy to abandon. If Gemini suddenly feels restrictive, people can switch to ChatGPT, Claude, or another free alternative within minutes.

Google’s move comes amid a broader industry trend where AI providers are tightening free usage. OpenAI, for instance, introduced rate limits for ChatGPT’s free tier earlier this year, capping the number of messages per hour for heavy users. Anthropic’s Claude also imposes usage caps based on the complexity of conversations. These measures are seen as necessary to manage server costs and ensure quality of service for paying customers. However, they also risk alienating the very user base that companies hope to convert into subscribers.

The economics behind AI chatbots are fundamentally different from traditional software. While cloud services like Google Drive or Dropbox rely on storage costs that have steadily declined, AI inference requires massive computational resources that scale with usage. Each interaction with a large language model consumes significant GPU time, and the demand for these models has exploded since the launch of ChatGPT in late 2022. According to estimates from industry analysts, a single query to a model like Gemini Ultra might cost a few cents in compute, which adds up quickly across millions of daily active users.

For Google, the stakes are particularly high. The company has invested tens of billions of dollars in AI research and infrastructure, including custom tensor processing units (TPUs) designed to accelerate model training and inference. While Gemini is integrated into products like Google Workspace, Android, and Google Cloud, the free tier remains a critical entry point for consumer adoption. Yet without some form of usage management, the free tier could become a financial drain that undermines the viability of the entire ecosystem.

That said, it is important not to overreact just yet. At the moment, this appears to be limited to a single user report, and Google has not officially announced weekly caps for Gemini’s free tier. There is always the possibility that this is part of a small-scale test or an experimental rollout that never expands further. Still, Google has a long history of quietly testing features with limited audiences before rolling them out more broadly. So even if this is only visible to a handful of users today, it would not be surprising to see stricter Gemini limits slowly appear for more people over the coming months. The bigger question is whether users will tolerate it once it happens. Because people have gotten very comfortable treating AI chatbots like infinite digital assistants. The moment those assistants start saying, “You’ve hit your limit for the week,” the relationship between users and AI platforms could start to feel very different.

The psychological shift is significant. Over the past two years, free AI chatbots have become deeply integrated into workflows for everything from writing and coding to research and entertainment. Users have developed habits that assume always-on availability. A weekly limit, even if generous, introduces a new friction that could change how people approach AI interaction. Instead of asking multiple follow-up questions to refine an idea, users might hesitate, saving queries for more important tasks. This could reduce the stickiness of the platform and open the door for competitors that maintain more lenient policies.

Google is not alone in facing this dilemma. Major players like Microsoft (with Copilot), Amazon (with Alexa+), and Apple (with future Siri upgrades) are all grappling with similar cost-reward calculations. The difference lies in how each company chooses to balance free access with premium upsells. Google’s advantage is its massive ecosystem—Gmail, YouTube, Google Maps, and Chrome all feed data and contexts that can enhance Gemini’s responses. But that advantage only matters if users remain engaged with the free tier long enough to appreciate the value of going paid.

Another factor to consider is the rise of open-source models. Communities around models like Llama, Mistral, and Falcon allow users to run AI locally on their own hardware, bypassing any commercial limits entirely. While these local models are not as powerful as state-of-the-art commercial ones, they are improving rapidly and offer privacy and control that cloud services cannot match. If Google makes its free tier too restrictive, some users may migrate to self-hosted alternatives, further fragmenting the market.

From a technical perspective, implementing weekly limits is relatively straightforward. Google already tracks usage metrics for its API services, where developers are charged per token. Extending that logic to consumer-facing interactions requires only a simple counter and a reset timer. The challenge is choosing thresholds that deter abuse without discouraging legitimate use. Heavy users—such as writers who draft dozens of emails daily or students who use Gemini for research—could hit caps quickly, leading to frustration. Meanwhile, casual users might never notice the limits, making the change almost invisible to the majority.

Google’s current Gemini pricing tiers offer a clue to its thinking. The free tier provides access to Gemini Pro for most tasks, while Gemini Advanced (part of the Google One AI Premium plan at $20/month) unlocks Gemini Ultra and deeper integration with Google products. The reported test might be a way to push free users toward that premium tier by making the free experience slightly less frictionless. However, the risk is that users perceive the limits as a bait-and-switch, damaging trust in the brand.

In the broader landscape, AI companies are racing to find a sustainable business model. Subscription revenues are growing but still represent a small fraction of the overall user base. Advertising, enterprise licenses, and API fees are other revenue streams, but none have yet proven sufficient to cover the enormous costs of training and running frontier models. Until a breakthrough in hardware efficiency or algorithmic optimization occurs, usage limits will likely become a standard feature of free AI services.

For now, the reported Gemini limits remain unconfirmed by Google. The company has not responded to requests for comment, and no official documentation has been published. Yet the pattern is familiar: a single user report surfaces a new feature, followed by a gradual rollout that eventually touches millions. As with many technology shifts, the early warning signs are often subtle. This screenshot may be the first crack in the wall of unlimited free AI access—a wall that never had a chance of standing forever.

Source: Digital Trends News

According to early reports, some users are now spotting a new “Thinking Level” option inside the Gemini app. The feature reportedly appears within Gemini’s existing model picker, where users already choose between options like Fast, Thinking, Pro, or Google AI Plus. Instead of simply choosing which model you want, Google also appears to be testing how deeply that model reasons through a task. The report says the new Thinking Level option currently shows up when selecting Fast (Gemini 3 Flash) or Gemini 3.1 Pro with thinking enabled. For now, the rollout seems extremely limited, likely confined to a small percentage of users for early feedback.

Things are getting a little more nuanced

If this sounds familiar, that is because Google AI Studio already offers similar controls with Low, Medium, and High reasoning levels. Bringing that flexibility into the regular Gemini app feels like the next obvious step, especially as AI companies increasingly compete on how “thoughtful” or agentic their assistants can feel. The move reflects a broader industry trend: giving users granular control over AI behavior. OpenAI, for instance, recently introduced a “reasoning effort” parameter in its API, and Anthropic has long offered temperature and top-p controls. However, Google’s approach is more user-friendly, embedded directly in the app interface rather than requiring developer-level tweaks.

The Thinking Level feature essentially allows Gemini to allocate more or fewer computational resources to analyzing a query. At the lowest level, the AI might generate a quick, surface-level response based on pattern matching. At the highest level, it could engage in multi-step reasoning, checking its own outputs and exploring alternative solutions before settling on an answer. This is particularly useful for complex tasks like mathematical proofs, logical puzzles, or detailed research questions where accuracy matters more than speed. Conversely, for everyday queries like “What’s the weather today?” or “Set a timer for 10 minutes,” a lower thinking level can save time and battery life on mobile devices.

Honestly, I feel this could be more useful than it sounds

Not every AI request needs maximum reasoning power. Sometimes you just want a quick answer without waiting several extra seconds while the model overanalyzes your grocery list, as if it were preparing a PhD thesis. Giving users control over that balance between speed and deeper reasoning could make Gemini feel much more flexible day to day. For power users, the high thinking level might enable Gemini to tackle problems that previously required specialized AI tools. For casual users, the low level ensures the assistant remains snappy and responsive. This tiered approach also aligns with Google’s broader vision of making AI ubiquitous: by letting the user decide how much “brainpower” to apply, the assistant can adapt to context without manual model switching.

From a technical perspective, the Thinking Level likely adjusts the number of tokens used in the chain-of-thought reasoning process. Gemini’s architecture supports chain-of-thought prompting, where the model explicitly lists intermediate steps before arriving at a conclusion. More thinking means longer chains and more self-correction loops. Google has not disclosed exact parameters, but similar systems in AI Studio offer three presets: Low (fast, minimal reasoning), Medium (balanced), and High (thorough, multi-check). The app version may adopt the same scale or offer a slider. Early UI screenshots show a simple dropdown, suggesting Google is prioritizing ease of use over complexity.

The timing of this leak is no coincidence. Google I/O 2026 is expected to focus heavily on Gemini’s evolution from a chatbot into a full-fledged digital assistant capable of handling complex, multi-step tasks. The Thinking Level feature fits precisely into that narrative: it gives users the ability to command the AI with varying levels of autonomy and depth. This could be a key differentiator against competitors like ChatGPT and Claude, which currently offer fewer such controls in their consumer apps. While developers have access to reasoning parameters via APIs, typical users are left with binary choices like “use GPT-4 or GPT-3.5.” Google’s approach democratizes that flexibility.

Beyond the Thinking Level, Google also appears to be expanding Gemini’s growing ecosystem of third-party app integrations. Right now, Gemini already works with apps and services like GitHub, OpenStax, Spotify, and WhatsApp. However, support documentation reportedly hints that integrations for Canva, Instacart, and OpenTable are also on the way. None of these integrations appears to be live yet, but the timing makes sense. Google I/O is usually where the company shows off Gemini becoming less of a chatbot and more of a proper digital assistant that can actually do things across apps and services. For instance, imagine asking Gemini to create a presentation on Canva, order groceries via Instacart, and book a dinner reservation on OpenTable — all in one conversation, with the AI handling the logic and sequencing.

These integrations would work through Gemini’s recently launched “Extensions” system, which allows the AI to connect to external services via a unified interface. Each extension handles authentication and data retrieval, while Gemini orchestrates the overall task. The addition of Canva, Instacart, and OpenTable would significantly expand Gemini’s utility beyond information retrieval into action taking. It positions Gemini closer to a virtual assistant like Alexa or Siri, but with far deeper reasoning capabilities — especially when combined with the Thinking Level control.

However, this evolution also raises questions about privacy and data handling. Third-party integrations require sharing user requests with external services, and Google has been cautious about transparency. The company recently published a detailed privacy policy for Gemini Extensions, explaining that data will be used only for fulfilling the immediate request and not for training models. Still, users who enable high thinking levels for tasks involving sensitive information might be wary of the increased processing. Google will need to clearly communicate how reasoning depth affects data retention and security.

Another angle is the impact on subscription tiers. Google currently offers Gemini with a free tier that includes limited access to advanced models. If Thinking Level becomes a premium feature — restricting high reasoning to Gemini Advanced subscribers — it could drive more users toward paid plans. Alternatively, Google might keep low and medium levels free while charging for high level, akin to its current model where “Pro” models are locked behind a subscription. This would mirror the AI Studio pricing, where more compute-intensive reasoning incurs higher costs. The leaked support documents did not clarify pricing, but given Google’s history, monetization is inevitable.

Meanwhile, Google is also pushing the boundaries of AI beyond text. The company recently announced that Project Genie can now use real-world imagery from Google Street View to generate interactive virtual environments. This “world model” technology allows users to walk through AI versions of real places, blending reality with AI-generated styles. While separate from Gemini, it demonstrates Google’s broader investment in creating immersive, reasoning-driven experiences. Similarly, Google wants to reinvent the TV remote with Gemini and pointer controls, turning televisions into interactive AI hubs. These efforts all point to a future where AI does not just answer questions but actively participates in shaping digital experiences.

Despite the excitement, not all changes are welcome. Google’s Gemini might be testing weekly limits, and free users won’t love it. A recent screenshot suggests stricter usage tracking and possible caps on interactions, especially when using heavier AI models. This could be a way to manage server costs as more users adopt the Thinking Level feature — higher reasoning levels consume more compute, and unlimited free access would be unsustainable. Google will need to strike a balance between retaining users and covering infrastructure expenses.

At this point, Gemini’s evolution feels less about smarter answers alone and more about turning the app into something that quietly handles parts of your digital life in the background — ideally without making everything feel unnecessarily complicated. The Thinking Level feature, combined with expanding integrations, represents a step toward that vision. By giving users control over reasoning depth, Google acknowledges that intelligence is not one-size-fits-all. Some days you need a quick ‘yes’ or ‘no’ other days you need a detailed analysis. With the new option, Gemini aims to offer both — and let you choose which you need right now.

As with any rumored feature, plans could change before the official announcement at Google I/O. The limited rollout suggests Google is still testing the waters; feedback from early users will likely shape the final implementation. If successful, Thinking Level could become a standard feature in AI assistants, forcing competitors to follow suit. For now, the news gives us a glimpse of how Google envisions the next generation of AI interaction — one where the user is in control of the machine’s mind.

Source: Digital Trends News

The Long Road to Siri 2.0

Siri has been a staple of Apple devices since its introduction with the iPhone 4S in October 2011. At that time, the assistant was labeled a beta product, and Apple quietly removed the branding in 2013. However, despite improvements over the years, Siri has consistently faced criticism for lagging behind competitors like Google Assistant and Amazon Alexa in terms of accuracy, natural language understanding, and third-party integration. The gap has only widened as the AI industry has shifted toward large language models and generative AI.

Apple's AI strategy, branded as Apple Intelligence, was first announced at WWDC 2024 but immediately encountered skepticism due to its limited scope and phased rollout. Features like improved Siri, on-device processing, and privacy-focused AI were promised, but many have been delayed. The most ambitious piece — a fully revamped Siri with chatbot capabilities — was originally expected in 2024 but has now reportedly slipped to 2026, with iOS 27 as its launch vehicle. Even then, the beta label indicates that Apple may consider it a work in progress.

What the New Siri Could Look Like

According to reports, the redesigned Siri will be rebuilt as a conversational assistant capable of handling ongoing dialogues, maintaining contextual memory, and integrating deeply with third-party apps. The update is also expected to introduce a standalone Siri app, similar to ChatGPT's interface, with chat-style interactions replacing the current command-based system. Additionally, the Dynamic Island on supported iPhones could be used to display Siri's status or responses without taking over the full screen.

Privacy remains a key concern for Apple. Sources indicate that the company is adding stronger privacy controls, such as optional auto-delete settings for conversation history, to reassure users that their data is not being stored unnecessarily. This aligns with Apple's long-standing stance that user privacy is a fundamental right, even as it adds more AI features to its devices.

The question is whether these features will be enough to win back users who have already migrated to more advanced assistants. Google's Gemini, integrated into Android and the Google app, offers real-time summarization, image generation, and proactive task assistance. ChatGPT, meanwhile, has become a household name for everything from creative writing to code debugging. Apple's slower, more deliberate approach may produce a polished product, but it risks being irrelevant by the time it arrives.

Why the Beta Label Matters Now

If Apple does opt to release the new Siri as a beta in iOS 27, it would serve multiple purposes. First, it provides Apple with flexibility to continue refining the assistant after launch, allowing the company to roll out updates quickly based on user feedback. Second, it helps manage expectations regarding bugs, hallucinations, or missing features that are common with early-stage AI. Third, it allows Apple to get the feature into users' hands sooner rather than waiting for a fully polished version — a departure from its typical strategy of only releasing products that are ready for the mass market.

However, labeling a feature as beta for years can also backfire. The original Siri beta lasted over two years, and many users today still perceive Siri as unfinished or unreliable despite the removal of the label. If the new Siri launches as beta, it may reinforce the narrative that Apple is constantly playing catch-up in AI rather than leading.

Competitive Landscape: Apple vs. the AI Giants

The AI assistant market has evolved dramatically since Siri's debut. Google Assistant, Amazon Alexa, and Microsoft's Cortana (now discontinued) each made strides in specific areas. But the real game-changer was the release of OpenAI's ChatGPT in late 2022, which popularized conversational AI with near-human responses. Google responded with Bard (now Gemini), and a wave of AI applications followed, from image generation to code assistance.

Apple, meanwhile, has been cautious. Its focus on on-device processing and privacy means that many AI features cannot rely on cloud servers, limiting their complexity. The new Siri may leverage a combination of on-device and cloud-based AI to deliver more advanced capabilities, but Apple is unlikely to compromise its privacy stance. For example, conversations may be anonymized and processed in batches, with users able to opt out of data sharing entirely.

Yet the challenge remains: even with the best privacy protections, users expect their assistants to work seamlessly. If Siri cannot parse complex queries or hold a thread of conversation like Gemini or ChatGPT, the privacy advantage may not matter to the average consumer.

What's Next: WWDC and Beyond

Apple is expected to provide more details about Siri's redesign and its broader AI roadmap at WWDC 2027, which typically takes place in June. Developer beta versions of iOS 27 will likely be the first opportunity for the public to test the new Siri experience. The announcement will also clarify whether the beta label will apply immediately or only during the testing phase.

Other features expected in iOS 27 include enhanced Apple Intelligence capabilities across apps, deeper integration with the Vision Pro, and possible expansion of the Dynamic Island functionality. However, the spotlight will undoubtedly be on Siri, as it represents Apple's biggest bet on catching up in the AI race.

The larger question is whether Apple's slower, more cautious AI rollout can still compete in a market where rivals have spent two years aggressively deploying generative AI into mainstream consumer products. For now, Siri's overhaul appears less like a finished comeback and more like Apple finally arriving at the AI race — still mid-development, with a beta tag to remind everyone of the work that remains.

Source: Digital Trends News

The auto-delete feature would allow users to set a timer for their Siri conversations: automatically erasing them after 30 days, after one year, or never. This mirrors the existing auto-delete system in Apple’s Messages app, where users can choose to delete messages after 30 days, one year, or forever. The integration of such controls directly into Siri’s core experience represents a shift from the industry norm, where privacy settings are often buried in menus.

Apple Is Rebuilding Siri Around Privacy-Friendly AI Conversations

For years, Siri has lagged behind competitors like ChatGPT, Google Gemini, and Amazon Alexa in terms of conversational ability. The iOS 27 update is expected to change that by introducing a standalone Siri app, allowing users to interact with the assistant through text input in addition to voice commands. A new “Search or Ask” mode will let users seamlessly switch between traditional web searches and Siri conversations.

Under the hood, Siri will gain the ability to maintain conversational context and remember previous interactions, a feature that modern AI chatbots already rely on heavily. However, Apple’s approach is more cautious. Rather than storing conversation histories indefinitely for personalization or model training, the company is reportedly building tighter limits around memory retention and data handling. This cautious approach is consistent with Apple’s longstanding commitment to user privacy, but it has also contributed to the company’s slower progress in the AI race.

The reported auto-delete feature is part of a broader strategy to make privacy a central selling point. Unlike many AI chatbots that offer temporary or incognito modes as optional settings, Apple appears to be integrating these controls directly into the user experience. This could mean that users are prompted to choose a deletion timeline when they first start using the new Siri, rather than having to dig through settings later.

Privacy as Apple’s Main AI Differentiator

Apple has spent over a decade positioning privacy as one of its core competitive advantages. This approach has helped the company differentiate itself from ad-driven giants like Google and Meta, but it has also slowed down its AI initiatives. While OpenAI, Google, and Anthropic have focused on building larger models and more advanced reasoning capabilities, Apple has been investing heavily in on-device processing and private cloud infrastructure.

The new Siri will continue to rely on Apple’s Private Cloud Compute system, which processes requests in a secure environment that does not store user data. At the same time, reports suggest that Apple may also tap into Google’s Gemini infrastructure to improve Siri’s underlying capabilities. This creates an unusual dynamic: Apple wants Siri to compete with cutting-edge AI chatbots, but without fully adopting the data collection practices that many competitors depend on.

The auto-delete feature is a direct response to growing consumer concerns about AI privacy. Many users are unaware that their conversations with chatbots like ChatGPT and Gemini are often stored indefinitely and used to improve models. Apple’s approach gives users more transparency and control, potentially attracting customers who are wary of how their data is being used.

Industry analysts believe this could be a significant differentiator in the consumer market. While tech enthusiasts might prioritize advanced reasoning and large model sizes, mainstream users often care more about trust and safety. By emphasizing automatic deletion, Apple can position Siri as the “safer” AI assistant for everyday tasks.

How the Auto-Delete Feature Compares to Competitors

Most major AI chatbots already offer some form of privacy controls. For example, ChatGPT has a “temp chat” mode that does not save conversations, and Google Gemini allows users to delete chat history manually. However, these are typically optional settings that users must enable each time, or they require navigating complex menus. Apple’s reported approach is different because it makes auto-deletion a core part of the setup process.

If implemented as described, users would likely choose their preferred deletion schedule when they first activate Siri. The options of 30 days, one year, or permanent retention give users flexibility while encouraging them to think about their privacy preferences upfront. This stands in contrast to competitors that default to saving all conversations and only delete them upon explicit request.

Another important aspect is that Apple is reportedly designing the feature to work across all Siri interactions, including voice queries and text-based conversations. This means that even if a user asks Siri a simple question, the answer will be governed by the same deletion rules. Such uniformity could reduce the risk of accidental data retention.

Data security experts have praised the move but note that the real test will be in implementation. For example, Apple must ensure that deleted conversations are truly removed from all servers and backups, and that users can easily verify this. The company’s reputation for privacy could be damaged if any loopholes are discovered.

Historical Context: Siri’s Long Road to Modern AI

Siri was first introduced as a standalone app in 2010 and acquired by Apple shortly thereafter. It debuted as a flagship feature of the iPhone 4S in 2011, and at the time, it was considered revolutionary. However, over the years, Siri fell behind competitors like Google Assistant and Amazon Alexa, which benefited from access to vast amounts of user data and cloud computing resources.

Apple attempted to catch up by acquiring AI startups, hiring top researchers, and investing in on-device machine learning. Yet the core Siri experience remained relatively unchanged for years. The introduction of Apple Intelligence in 2024 marked a turning point, but the rollout has been slow and incremental. iOS 27 is expected to be the most dramatic overhaul since Siri’s inception.

The auto-delete feature is part of a larger effort to give users granular control over their data. Apple has already implemented similar controls in iCloud, Photos, and other services. Extending this to Siri’s AI conversations is a natural next step, but it also highlights the tension between privacy and functionality. AI systems often improve by learning from user interactions, and limiting data retention could slow down Siri’s development.

Nevertheless, Apple seems willing to accept this trade-off. The company’s marketing materials for the new Siri are expected to emphasize that users do not have to sacrifice privacy to use a modern AI assistant. This messaging could resonate strongly in a market where data breaches and AI scandals are increasingly common.

Broader Implications for the AI Industry

If Apple’s auto-delete feature is well-received, it could pressure other AI companies to adopt similar privacy-first designs. Currently, many chatbots rely on continuous data collection to improve their models, and transparency about data usage is often lacking. Apple’s move could set a new standard for what consumers expect from AI assistants.

Regulators around the world are also paying attention. The European Union’s AI Act and similar legislation in other regions require companies to be more transparent about data handling and to give users more control. Apple’s approach aligns well with these regulatory trends, potentially giving the company an advantage in markets with strict privacy laws.

On the other hand, some critics argue that auto-deletion could hinder Siri’s ability to provide personalized responses. Without long-term memory, the assistant may struggle to learn user preferences and habits. Apple is reportedly addressing this by storing some context on-device, but the details remain unclear. The company may also allow users to opt into longer retention for specific purposes, such as smart home automation or frequent reminders.

The successful implementation of iOS 27 Siri will be closely watched by competitors and consumers alike. If Apple can deliver a conversational AI that rivals ChatGPT while maintaining its privacy commitments, it could reshape the entire industry. The auto-delete feature is just one piece of that puzzle, but it might be the most important one.

Source: Digital Trends News

Under the program, residents registered with Malta’s digital identity system will gain access to ChatGPT Plus after completing a government-backed AI training course focused on practical and responsible AI usage. The rollout begins this month and also includes Maltese citizens living abroad.

This partnership marks a significant shift in how AI is being integrated into society. Just a few years ago, ChatGPT was primarily a productivity tool for students, coders, and office workers. Now, entire nations are exploring nationwide AI access programs. Malta’s move positions the country as a testbed for large-scale AI adoption, potentially setting a precedent for other nations.

OpenAI wants Malta to become a nationwide AI adoption experiment

The AI literacy course is designed to equip citizens with the skills needed to use AI responsibly and effectively. It covers topics such as prompt engineering, ethical considerations, data privacy, and the limitations of AI models. By requiring completion of this course before accessing the premium subscription, OpenAI and the Maltese government aim to ensure that users are informed and capable rather than just passive consumers.

The partnership also raises important questions about the role of private companies in public education and infrastructure. While OpenAI stands to gain widespread user adoption and valuable data, the government of Malta gains a competitive edge in AI readiness. Other countries are watching closely. The UAE, for example, has been working with OpenAI through its massive Stargate UAE infrastructure partnership, and reports suggest nationwide ChatGPT access is being explored there as well.

Historically, similar public-private partnerships have occurred in areas like internet connectivity and digital identity. However, the AI sector is unique because the underlying technology is evolving rapidly and the implications for jobs, education, and democracy are profound. OpenAI’s move is reminiscent of when Microsoft partnered with governments to provide Office 365 to schools, but the stakes are higher here because AI is not just a productivity tool—it can generate content, make decisions, and influence opinions.

This is starting to feel less like software and more like digital infrastructure

What makes this deal interesting is how quickly AI tools are evolving from consumer products into something governments increasingly view as public infrastructure. Just a couple of years ago, ChatGPT was mostly a productivity tool for students, coders, and office workers. Now, entire countries are discussing nationwide AI access programs.

And honestly, that shift should probably make people pause a little. Once governments start integrating specific AI platforms into education, workplaces, and public services, these tools stop being optional conveniences and start becoming deeply embedded digital dependencies. For OpenAI, this is brilliant positioning, but if entire countries eventually begin relying on one company’s AI ecosystem, this stops being about chatbots and starts looking a lot more like infrastructure control.

The concept of digital infrastructure is not new. In the early 2000s, governments around the world worked with companies like Google and Microsoft to provide email services and cloud storage to citizens. However, those services were relatively passive—they stored data and facilitated communication. AI, on the other hand, is active; it can generate text, analyze data, and even automate decision-making. When a government’s public services rely on a single AI provider, that provider effectively governs the algorithms that shape how citizens interact with their government.

Malta, a small island nation with a population of just over 500,000, is an ideal candidate for such an experiment. Its size allows for controlled implementation and easy monitoring. The country already has a robust digital infrastructure, including a national digital identity system that is used for everything from tax filing to healthcare. By adding AI access to this system, Malta is effectively embedding AI into the daily lives of its citizens.

Privacy experts have raised concerns about data handling in such a partnership. While OpenAI has stated that the program respects data protection regulations, the fact that a private company will have access to the interactions of an entire nation’s citizens is unprecedented. The AI literacy course includes modules on privacy, but the underlying tension remains: can a company whose business model relies on data truly be neutral when serving as a public utility?

On the other hand, proponents argue that democratizing access to advanced AI is essential for economic competitiveness. In an era where AI skills are increasingly valued in the job market, providing free access to a premium AI tool could level the playing field for citizens in small countries like Malta. The government has positioned the initiative as a way to prepare its workforce for the future, reduce digital divides, and attract tech companies to the region.

The University of Malta’s involvement adds academic credibility to the program. The university will help design and deliver the AI literacy course, ensuring it is pedagogically sound and up-to-date. It will also conduct research on the impact of nationwide AI access, providing valuable data for other governments considering similar initiatives.

Competitors like Anthropic (maker of Claude) and Google (with Gemini) are also pursuing government partnerships. Google’s Project Genie, for instance, uses Street View data to create interactive AI-generated worlds, which could be used for education and tourism. However, OpenAI’s head start with Malta gives it a strong foothold in the government sector. If the Malta experiment succeeds, other nations—especially those with smaller populations or developing economies—may follow suit, creating a network effect that solidifies OpenAI’s dominance.

Yet there are potential pitfalls. The one-year subscription period may create a dependency that is hard to break. Once citizens become accustomed to the premium features, they may pressure the government to continue the subscription or find it difficult to switch to a different AI platform. This “lock-in” effect is a common concern in technology policy and could result in a de facto monopoly on AI services in Malta.

Furthermore, the program relies on Malta’s digital identity system, which itself raises questions about surveillance and data centralization. Combining digital identity with AI access means that every interaction with the AI can be traced back to a specific individual. While the government assures that data will be anonymized and used only for program improvement, the potential for misuse is significant.

Despite these concerns, the partnership represents a bold step forward in AI policy. It acknowledges that AI literacy is as important as digital literacy was two decades ago. Governments can no longer afford to let AI adoption happen organically; they must actively shape how their citizens learn about and use these tools. Malta’s approach could serve as a model for other nations, but it also highlights the need for robust regulatory frameworks to prevent over-reliance on a single private company.

As the program rolls out, all eyes will be on Malta. Will citizens embrace the AI tool? Will it boost productivity and innovation? Or will it create new inequalities between those who complete the course and those who do not? The answers to these questions will shape the future of government-AI partnerships for years to come.

Source: Digital Trends News

Training efficiency isn’t about squeezing GPUs harder; it’s about spending smarter for the same accuracy. The following methods focus on training-time cost levers, changes inside the loop that cut waste without touching your model architecture. All code examples below are available in the accompanying Green AI Optimization Toolkit repository.

The compute levers: Taking weight off the chassis

The easiest way to speed up a race car is to take weight off the chassis. In Deep Learning, that weight is precision. For years, 32-bit floating point (FP32) was the default. But today, switching to Mixed-Precision Math (FP16/INT8) is the highest ROI change a practitioner can make. On hardware with dedicated tensor units, like NVIDIA Ampere/Hopper, AMD RDNA 3 or Intel Gaudi 2, mixed precision can increase throughput by 3x or more.

However, this isn’t a magic wand for everyone. If you are running on pre-2019 GPUs (like the Pascal architecture) that lack Tensor Cores, you might see almost no speed gain while risking numerical instability. Similarly, compliance workloads in finance or healthcare that require bit-exact reproducibility may need to stick to FP32. But for the 90% of use cases involving memory-bound models (ResNet-50, GPT-2, Stable Diffusion), the shift is essential. It also unlocks Gradient Accumulation, allowing you to train massive models on smaller, cheaper cards by simulating larger batch sizes.

 From 'green-ai-optimization-toolkit/01_mixed_precision.py'

import torch
from torch.cuda.amp import autocast, GradScaler

 Simulate a Batch Size of 64 using a Micro-Batch of 8
eff_batch_size = 64
micro_batch = 8
accum_steps = eff_batch_size // micro_batch 

scaler = GradScaler()  Prevents gradient underflow in FP16

for i, (data, target) in enumerate(loader):
     1. The Toggle: Run forward pass in FP16
    with autocast():
        output = model(data)
        loss = criterion(output, target)
        loss = loss / accum_steps  Normalize loss
    
     2. Scale gradients and accumulate
    scaler.scale(loss).backward()
    
     3. Step only after N micro-batches
    if (i + 1) % accum_steps == 0:
        scaler.step(optimizer)
        scaler.update()
        optimizer.zero_grad()

The data levers: Feeding the beast

If your GPU utilization is hovering around 40%, you aren’t training a model; you are burning cash. The bottleneck is almost always the data loader. A common mistake is treating data preprocessing as a per-epoch tax. If you use expensive text tokenizers (like Byte-Pair Encoding) or complex image transforms, cache pre-processed data. Tokenize or resize once, store the result and feed it directly.

Furthermore, look at your file formats. Reading millions of small JPEG or CSV files over a network file system kills I/O throughput due to metadata overhead. Instead, stream data via archives. Sharding your dataset into POSIX tar files or binary formats like Parquet/Avro allows the OS to read ahead, keeping the GPU hungry. Watch out for storage ballooning: Caching pre-processed data can triple your storage footprint. You are trading storage cost (cheap) for compute time (expensive). Also be careful with over-pruning: While data deduplication is excellent for web scrapes, be careful with curated medical or legal datasets. Aggressive filtering might discard rare edge cases that are critical for model robustness.

The operational levers: Safety and scheduling

The most expensive training run is the one that crashes 99% of the way through and has to be restarted. In the cloud, spot instances (or pre-emptible VMs) offer discounts of up to 90%. To use them safely, you must implement robust checkpointing. Save the model state frequently (every epoch or N steps) so that if a node is reclaimed, you lose minutes of work, not days.

Open-source orchestration frameworks like SkyPilot have become essential here. SkyPilot abstracts away the complexity of Spot Instances, automatically handling the recovery of reclaimed nodes and allowing engineers to treat disparate clouds (AWS, GCP, Azure) as a single, cost-optimized resource pool. You should also implement early stopping. There is no ROI in “polishing noise”. If your validation loss plateaus for 3 epochs, kill the run. This is especially potent for fine-tuning tasks, where most gains arrive in the first few epochs. However, be cautious if you are using curriculum learning, where loss might naturally rise before falling again as harder examples are introduced.

The “smoke test” protocol

Finally, never launch a multi-node job without a dry run. A simple script that runs two batches on a CPU can catch shape mismatches and OOM bugs for pennies. Here is an implementation in Python:

 From 'green-ai-optimization-toolkit/03_smoke_test.py'
def smoke_test(model, loader, device='cpu', steps=2):
    """
    Runs a dry-run on CPU to catch shape mismatches 
    and OOM bugs before the real run starts.
    """
    print(f"💨 Running Smoke Test on {device}...")
    model.to(device)
    model.train()
    
    try:
        for i, (data, target) in enumerate(loader):
            if i >= steps: break
            data, target = data.to(device), target.to(device)
            output = model(data)
            loss = output.sum()
            loss.backward()
        print("✅ Smoke Test Passed. Safe to launch expensive job.")
        return True
    except Exception as e:
        print(f"❌ Smoke Test Failed: {e}")
        return False

The rapid-fire checklist: 10 tactical quick wins

Beyond the major architectural shifts, there is a long tail of smaller optimizations that, when stacked, yield significant savings. Here is a rapid-fire checklist of tactical wins.

1. Dynamic batch-size auto-tuning

The tactic: Have the framework probe VRAM at launch and automatically choose the largest safe batch size. Best for: Shared GPU clusters (Kubernetes/Slurm) where free memory swings wildly. Watch out: Can break real-time streaming SLAs by altering step duration.

2. Continuous profiling

The tactic: Run lightweight profilers (PyTorch Profiler, NVIDIA Nsight) for a few seconds per epoch. Best for: Long jobs (>30 mins). Finding even a 5% hotspot pays back the profiler overhead in a day. Watch out: I/O-bound jobs. If GPU utilization is <20%, a profiler won’t help; fix your data pipeline first.

3. Store tensors in half-precision

The tactic: Save checkpoints and activations in FP16 (instead of default FP32). Best for: Large static embeddings (vision, text). It halves I/O volume and storage costs. Watch out: Compliance workloads requiring bit-exact auditing.

4. Early-phase CPU training

The tactic: Run the first epoch on cheaper CPUs to catch gross bugs before renting GPUs. Best for: Complex pipelines with heavy text parsing or JSON decoding. Watch out: Tiny datasets where the data transfer time exceeds the compute time.

5. Offline augmentation

The tactic: Pre-compute heavy transforms (Mosaic, Style Transfer) and store them, rather than computing on-the-fly. Best for: Heavy transforms that take >20ms per sample. Watch out: Research that studies augmentation randomness; baking it removes variability.

6. Budget alerts & dashboards

The tactic: Stream cost metrics per run and alert when burn-rate exceeds a threshold. Best for: Multi-team organizations to prevent “runaway” billing. Watch out: Alert Fatigue. If you ping researchers too often, they will ignore the notifications.

7. Archive stale artifacts

The tactic: Automatically move checkpoints >90 days old to cold storage (Glacier/Archive tier). Best for: Mature projects with hundreds of experimental runs. Watch out: Ensure you keep the “Gold Standard” weights on hot storage for inference.

8. Data deduplication

The tactic: Remove near-duplicate samples before training. Best for: Web scrapes and raw sensor logs. Watch out: Curated medical/legal datasets where “duplicates” might actually be critical edge cases.

9. Cluster-wide mixed-precision defaults

The tactic: Enforce FP16 globally via environment variables so no one “forgets” the cheapest knob. Best for: MLOps teams managing multi-tenant fleets. Watch out: Legacy models that may diverge without specific tuning.

10. Neural architecture search (NAS)

The tactic: Automate the search for efficient architectures rather than hand-tuning. Best for: Long-term production models where efficiency pays dividends over years. Watch out: Extremely high upfront compute cost; only worth it if the model will be deployed at massive scale.

You don’t need to wait for an H100 allocation to make your AI stack efficient. By implementing mixed precision, optimizing your data feed and adding operational safety nets, you can drastically reduce both your carbon footprint and your cloud bill. The most sustainable AI strategy isn’t buying more power, it’s wasting less of what you already have.

Source: InfoWorld News

We often talk about “accuracy” and “engagement” as the north stars of AI. But recently, a new metric has become just as critical: efficiency. At Meta, I worked on the infrastructure powering Instagram Reels recommendations. We were dealing with a platform serving over a billion daily active users. At that scale, even a minor inefficiency in how data is processed or stored snowballs into megawatts of wasted energy and millions of dollars in unnecessary costs. We faced a challenge that is becoming increasingly common in the age of generative AI: how do we make our models smarter without making our data centers hotter?

The answer wasn’t in building a smaller model. It was in rethinking the plumbing — specifically, how we computed, fetched and stored the training data that fueled those models. By optimizing this “invisible” layer of the stack, we achieved over megawatt-scale energy savings and reduced annual operating expenses by eight figures. Here is how we did it.

The hidden cost of the recommendation funnel

To understand the optimization, you have to understand the architecture. Modern recommendation systems generally function like a funnel. At the top, you have retrieval, where we select thousands of potential candidates from a pool of billions of media items. Next comes early-stage ranking, a high-efficiency phase that filters this large pool down to a smaller set. Finally, we reach late-stage ranking. This is where the heavy lifting happens. We use complex deep learning models — often two-tower architectures that combine user and item embeddings — to precisely order a curated set of 50 to 100 items to maximize user engagement.

This final stage is incredibly feature-dense. To rank a single Reel, the model might look at hundreds of “features.” Some are dense features (like the time a user has spent on the app today) and others are sparse features (like the specific IDs of the last 20 videos watched). The system doesn’t just use these features to rank content; it also has to log them. Why? Because today’s inference is tomorrow’s training data. If we serve you a video and you “like” it, we need to join that positive label with the exact features the model saw at that moment to retrain and improve the system.

This logging process — writing feature values to a transient key-value (KV) store to wait for user interaction — was our bottleneck.

The challenge of transitive feature logging

To understand why this bottleneck existed, we have to look at the microscopic lifecycle of a single training example. In a typical serving path, the inference service fetches features from a low-latency feature store to rank a candidate set. However, for a recommendation system to learn, it needs a feedback loop. We must capture the exact state of the world (the features) at the moment of inference and later join them with the user’s future action (the label), such as a “like” or a “click.”

This creates a massive distributed systems challenge: stateful label joining. We cannot simply query the feature store again when the user clicks, because features are mutable — a user’s follower count or a video’s popularity changes by the second. Using fresh features with stale labels introduces “online-offline skew,” effectively poisoning the training data.

To solve this, we use a transitive key-value (KV) store. Immediately after ranking, we serialize the feature vector used for inference and write it to a high-throughput KV store with a short time-to-live (TTL). This data sits there, “in transit,” waiting for a client-side signal. If the user interacts, the client fires an event, which acts as a key lookup. We retrieve the frozen feature vector from the KV store, join it with the interaction label and flush it to our offline training warehouse (e.g., Hive/Data Lake) as a “source-of-truth” training example. If the user does not interact, the TTL expires, and the data is dropped to save costs.

This architecture, while robust for data consistency, is incredibly expensive. We were essentially continuously writing petabytes of high-dimensional feature vectors to a distributed KV store, consuming massive network bandwidth and serialization CPU cycles.

Optimizing the “head load”

We realized that our “write amplification” was out of control. In the late-stage ranking phase, we typically rank a deep buffer of items — say, the top 100 candidates — to ensure the client has enough content cached for a smooth scroll. The default behavior was eager logging: We would serialize and write the feature vectors for all 100 ranked items into the transitive KV store immediately.

However, user behavior follows a steep decay curve. A user might only view the first 5–6 items (the “head load”) before closing the app or refreshing the feed. This meant we were paying the serialization and I/O cost to store features for items 7 through 100, which had a near-zero probability of generating a positive label. We were effectively DDoSing our own infrastructure with “ghost data.”

We shifted to a “lazy logging” architecture. First, we reconfigured the serving pipeline to only persist features for the Head Load (e.g., top 6 items) into the KV store initially. Second, as the user scrolls past the Head Load, the client triggers a lightweight “pagination” signal. Only then do we asynchronously serialize and log the features for the next batch (items 7–15). This change decoupled our ranking depth from our storage costs. We could still rank 100 items to find the absolute best content, but we only paid the “storage tax” for the content that actually had a chance of being seen. This reduced our write throughput (QPS) to the KV store significantly, saving megawatts of power previously wasted on serializing data that was destined to expire untouched.

Rethinking storage schemas

Once we reduced what we stored, we looked at how we stored it. In a standard feature store architecture, data is often stored in a tabular format where every row represents an impression (a specific user seeing a specific item). If we served a batch of 15 items to one user, the logging system would write 15 rows. Each row contained the item features (which are unique to the video) and the user features (which are identical for all 15 rows). We were effectively writing the user’s age, location and follower count 15 separate times for a single request.

We moved to a batched storage schema. Instead of treating every impression as an isolated event, we separated the data structures. We stored the user features once for the request and stored a list of item features associated with that request. This simple de-duplication reduced our storage requirement by more than 40%. In distributed systems like the ones powering Instagram or YouTube, storage isn’t passive; it requires CPU to manage, compress and replicate. By slashing the storage footprint, we improved bandwidth availability for the distributed workers fetching data for training, creating a virtuous cycle of efficiency throughout the stack.

Auditing the feature usage

The final piece of the puzzle was spring cleaning. In a system as old and complex as a major social network’s recommendation engine, digital hoarding is a real problem. We had over 100,000 distinct features registered in our system. However, not all features are created equal. A user’s “age” might carry very little weight in the model compared to “recently liked content.” Yet, both cost resources to compute, fetch and log.

We initiated a large-scale feature auditing program. We analyzed the weights assigned to features by the model and identified thousands that were adding statistically insignificant value to our predictions. Removing these features didn’t just save storage; it reduced the latency of the inference request itself because the model had fewer inputs to process. This kind of pruning is a continuous process, as new features are regularly added by different teams. Establishing a culture of data hygiene and regular audits is essential for maintaining long-term efficiency.

The energy imperative

As the industry races toward larger generative AI models, the conversation often focuses on the massive energy cost of training GPUs. Reports indicate that AI energy demand is poised to skyrocket in the coming years. For instance, a single training run of a large language model can consume as much electricity as hundreds of households use in a year. But for engineers on the ground, the lesson from my time at Meta is that efficiency often comes from the unsexy work of plumbing. It comes from questioning why we move data, how we store it and whether we need it at all.

By optimizing our data flow — lazy logging, schema de-duplication and feature auditing — we proved that you can cut costs and carbon footprints without compromising the user experience. In fact, by freeing up system resources, we often made the application faster and more responsive. Sustainable AI isn’t just about better hardware; it’s about smarter engineering. These techniques are applicable beyond social media; any large-scale ML system that relies on feature logging and label joining can benefit from similar redesigns. As the demand for personalized recommendations continues to grow, energy-aware engineering will become a competitive advantage, both financially and environmentally.

This approach also aligns with broader industry trends such as green computing and sustainable software development. Companies like Google and Microsoft have committed to carbon-negative or carbon-free operations, and optimizing recommendation pipelines is a concrete step toward those goals. The methods described here are not proprietary; they are sound engineering practices that any team can adopt if they are willing to look beyond model architecture and pay attention to the data infrastructure that supports it.

Source: InfoWorld News

Open Source Becomes Dull and Essential

The romantic notion of open source as a developer-led revolution has given way to a more mundane reality: open source is where infrastructure hardens into standards. The Cloud Native Computing Foundation (CNCF) now hosts over 230 projects with more than 300,000 contributors worldwide. Its 2025 survey found that 98% of organizations have adopted cloud-native techniques, and 82% of container users run Kubernetes in production. GitHub’s Octoverse report for 2025 recorded 1.12 billion contributions, over 180 million developers, and 518.7 million merged pull requests. The Apache Software Foundation remains steady with 9,905 committers across 295 projects and 1,310 software releases in fiscal year 2025. These numbers confirm that open source engagement is alive and well, but concentrated in the infrastructure layers that matter most for AI and scale.

Who Contributes and Why

The list of top contributors reveals a clear corporate strategy. According to CNCF Devstats for 2025, Red Hat led with 194,699 contributions, followed by Microsoft (107,645) and Google (91,158). Independent contributors came fourth at 52,404, proving that community still plays a role, but the center of gravity is unmistakably corporate. Red Hat’s dominance stems from its Kubernetes-centric product OpenShift, making its contributions a direct product strategy rather than charity. Microsoft, once an open source skeptic, now invests heavily in projects like OpenTelemetry, which saw a 39% rise in commits and a contributor base growing from 1,301 to 1,756 in a single year. This is not philanthropy—it’s a land grab for observability standards. Splunk and other vendors similarly invest to normalize interfaces and shape operational assumptions.

Cilium, a networking and security project, saw contributing companies rise 90% after joining CNCF, from 533 to 1,011, and individual contributors jumped from 1,269 to 4,464. Google, Datadog, and Cloudflare expanded their contributions as the project matured. Cilium sits at the intersection of networking, observability, and security—precisely the categories that become mission-critical when workloads are distributed, latency-sensitive, and expensive. Nvidia, despite its massive cash reserves, ranked 14th in Kubernetes contributions with 5,892 contributions and has open sourced the KAI Scheduler, a Kubernetes-native GPU scheduler. Nvidia also describes itself as a key contributor to Kubeflow. This indicates that even dominant hardware companies recognize the need to influence the orchestration and workflow layers that determine how their chips are used in real-world AI systems.

AI Driving Open Infrastructure

AI workloads are accelerating the importance of open infrastructure. CNCF reports that 66% of organizations hosting generative AI models now use Kubernetes for some or all inference workloads, calling Kubernetes the de facto operating system for AI. This claim is plausible given the need for scalable, portable, and observable infrastructure for training and inference. Kubeflow, built on Kubernetes, is becoming central to AI workflows. By contributing to these projects, companies ensure that AI systems remain governable, visible, and efficient—qualities that proprietary stacks often lack. The open source model allows organizations to inspect, influence, and adapt the infrastructure to their specific needs, reducing lock-in and fostering innovation.

The shift is not without trade-offs. Open source is increasingly about control—not proprietary control, but control over the layers where ecosystems harden into standards. The companies investing upstream are not doing it out of civic virtue but because whoever shapes the substrate typically gains leverage over everything built on top of it. This pragmatic approach has made open source more essential than ever, but also less romantic. The era of open source as a moral crusade is over. It has been replaced by a cold, strategic calculus where investment in open source is a necessary cost of doing business in the cloud-native and AI age.

In conclusion, open source has not died; it has matured. It has become the control plane for AI and modern infrastructure. The numbers and trends from CNCF, GitHub, and Apache confirm that engagement is shifting to the most impactful layers. AI is making open infrastructure more important because few organizations want to build their future on opaque, inescapable platforms. The future of open source is dull, essential, and heavily funded—exactly what the industry needs to support the next wave of AI innovation.

Source: InfoWorld News

According to IBM’s The Enterprise in 2030 report, a head-turning 79% of C-suite executives expect AI to boost revenue within four years, but only about 25% can pinpoint where that revenue will come from. This disconnect fosters unrealistic expectations and creates pressure to deliver quickly on initiatives that are still experimental or immature. This phenomenon, widely referred to as the AI hype hangover, mirrors earlier technology cycles such as the dot-com boom, cloud computing, and digital transformation, where initial exuberance gave way to a sobering realization of the hard work required to achieve meaningful outcomes.

The way AI dominates discussions at conferences is in stark contrast to its slower progress in the real world. New capabilities in generative AI and machine learning show promise, but moving from pilot to impactful implementation remains challenging. Many experts describe this as an AI hype hangover, in which implementation challenges, cost overruns, and underwhelming pilot results quickly dim the glow of AI’s potential. Similar cycles occurred with cloud and digital transformation, but this time the pace and pressure are even more intense. The stakes are higher because AI is expected to transform not just IT operations but entire business models.

Use cases vary widely

AI’s greatest strengths, such as flexibility and broad applicability, also create challenges. In earlier waves of technology, such as ERP and CRM, return on investment was a universal truth. AI-driven ROI varies widely—and often wildly. Some enterprises can gain value from automating tasks such as processing insurance claims, improving logistics, or accelerating software development. However, even after well-funded pilots, some organizations still see no compelling, repeatable use cases. This variability is a serious roadblock to widespread ROI. Too many leaders expect AI to be a generalized solution, but AI implementations are highly context-dependent. The problems you can solve with AI (and whether those solutions justify the investment) vary dramatically from enterprise to enterprise.

This leads to a proliferation of small, underwhelming pilot projects, few of which are scaled broadly enough to demonstrate tangible business value. In short, for every triumphant AI story, numerous enterprises are still waiting for any tangible payoff. For some companies, it won’t happen anytime soon—or at all. The historical context is telling: during the early days of cloud computing, many organizations ran small test workloads without achieving significant cost savings or agility until they committed to full-scale migration and modernization. AI requires a similar leap of faith combined with rigorous evaluation.

The cost of readiness

If there is one challenge that unites nearly every organization, it is the cost and complexity of data and infrastructure preparation. The AI revolution is data hungry. It thrives only on clean, abundant, and well-governed information. In the real world, most enterprises still wrestle with legacy systems, siloed databases, and inconsistent formats. The work required to wrangle, clean, and integrate this data often dwarfs the cost of the AI project itself. Beyond data, there is the challenge of computational infrastructure: servers, security, compliance, and hiring or training new talent. These are not luxuries but prerequisites for any scalable, reliable AI implementation.

In times of economic uncertainty, most enterprises are unable or unwilling to allocate the funds for a complete transformation. Many leaders have said that the most significant barrier to entry is not AI software but the extensive, costly groundwork required before meaningful progress can begin. For example, a manufacturer wanting to use AI for predictive maintenance must first unify data from multiple plant floors, standardize sensor formats, and ensure real-time data pipelines—all of which can take months or years. The same struggles occur in financial services, healthcare, and retail. Without this foundational investment, even the most advanced AI models will fail to deliver value, reinforcing the hangover effect.

Three steps to AI success

Given these headwinds, the question isn’t whether enterprises should abandon AI, but rather, how can they move forward in a more innovative, more disciplined, and more pragmatic way that aligns with actual business needs? The first step is to connect AI projects with high-value business problems. AI can no longer be justified because everyone else is doing it. Organizations need to identify pain points such as costly manual processes, slow cycles, or inefficient interactions where traditional automation falls short. Only then is AI worth the investment. This requires a shift from technology-first thinking to problem-first thinking, where business leaders articulate the desired outcome and data scientists evaluate whether AI is the right tool.

Second, enterprises must invest in data quality and infrastructure, both of which are vital to effective AI deployment. Leaders should support ongoing investments in data cleanup and architecture, viewing them as crucial for future digital innovation, even if it means prioritizing improvements over flashy AI pilots to achieve reliable, scalable results. This includes building data lakes or meshes, implementing data governance frameworks, and ensuring compliance with regulations such as GDPR or CCPA. Without such investments, AI projects will remain stuck in pilot purgatory, consuming resources without delivering business impact.

Third, organizations should establish robust governance and ROI measurement processes for all AI experiments. Leadership must insist on clear metrics such as revenue, efficiency gains, or customer satisfaction and then track them for every AI project. By holding pilots and broader deployments accountable for tangible outcomes, enterprises will not only identify what works but will also build stakeholder confidence and credibility. Projects that fail to deliver should be redirected or terminated to ensure resources support the most promising, business-aligned efforts. This disciplined approach helps avoid the trap of spreading investments too thinly across many unproven initiatives.

The road ahead for enterprise AI is not hopeless, but will be more demanding and require more patience than the current hype would suggest. Success will not come from flashy announcements or mass piloting, but from targeted programs that solve real problems, supported by strong data, sound infrastructure, and careful accountability. For those who make these realities their focus, AI can fulfill its promise and become a profitable enterprise asset. The cure for the AI hype hangover is simply a return to fundamentals: align technology with business value, build the necessary foundations, and measure outcomes rigorously.

Source: InfoWorld News

The Economics of Contribution Have Broken

The core problem is a brutal asymmetry in effort. A developer can prompt an AI agent to generate changes across dozens of files in under a minute. But a maintainer must spend an hour carefully reviewing those changes for correctness, edge cases, and long-term alignment with the project's vision. When hundreds of contributors all use AI tools, the result is not a better project but a maintainer who walks away.

Mitchell Hashimoto, founder of HashiCorp, recently indicated he might close external pull requests entirely. He cited the flood of AI-generated submissions as a primary reason. Flask creator Armin Ronacher has described this phenomenon as "agent psychosis," where developers become addicted to the dopamine hit of agentic coding, spinning up agents that run wild through projects.

Real-World Impact on Projects

The OCaml community experienced a vivid example when maintainers rejected an AI-generated pull request containing more than 13,000 lines of code. They cited copyright concerns, lack of review resources, and the long-term maintenance burden. One maintainer warned that such low-effort submissions create a real risk of bringing the pull request system to a halt.

Even GitHub, the host of the world's largest code forge, is exploring tighter pull request controls and UI-level deletion options. This platform-level response indicates that the problem is no longer a niche annoyance but a structural shift in how open source is made.

Small Libraries Face Obsolescence

Beyond the PR flood, AI is also changing the incentives for using small open source libraries. Nolan Lawson, author of the blob-util JavaScript library with millions of downloads, argues that the era of small, low-value utility libraries is over. In the age of Claude and GPT-5, developers can simply ask an AI to generate a utility function in milliseconds, eliminating the need to take on a dependency.

Lawson points out that these libraries served an educational purpose, allowing developers to learn by reading others' code. When replaced with ephemeral AI-generated snippets, that teaching function is lost. The community trades understanding for instant answers.

The Bifurcation of Open Source

This crisis is leading to a bifurcation in the open source world. On one side are massive, enterprise-backed projects like Linux or Kubernetes. These have the resources to build their own AI-filtering tools and the organizational weight to ignore noise. They can afford to maintain high barriers to contribution.

On the other side are "provincial" projects run by individuals or small cores. Many of these are simply stopping external contributions altogether. The irony is that AI was supposed to make open source more accessible, but by lowering the barrier to contribution, it has lowered the value of each contribution. When everyone can contribute, nobody's contribution is special.

Redefining Openness

Open source is not dying, but the meaning of "open" is being redefined. The era of radical transparency, where anyone could contribute, is giving way to an era of radical curation. The future of open source may belong to the few, not the many. The community was always more myth than reality, but AI has made the myth unsustainable.

The most successful open source projects moving forward will be those that are hardest to contribute to. They will demand a high level of human effort, context, and relationship. They will reject the slop loops in favor of slow, deliberate, and deeply personal development. The bazaar was a fun idea while it lasted, but it could not survive the arrival of the robots.

In sum, the open source world does not need more code; it needs more care. Care for the humans who shepherd communities and create code that will endure beyond a simple prompt. The shift may produce a quieter, smaller, and more exclusive ecosystem, but perhaps that is the only way open source survives the age of AI agents.

Source: InfoWorld News

Seit 1. Mai 2025 ist das Parken in der Park-and-Ride-Anlage Fölling im Grazer Norden kostenlos. Eigentlich ein starkes Angebot, um Pendler aus der Umgebung zum Umstieg auf öffentliche Verkehrsmittel zu bewegen. Doch der erhoffte Zuspruch bleibt aus. Ein Lokalaugenschein offenbart eine frustrierende Realität: Die Hochgarage, die 2010 um 2,7 Millionen Euro errichtet wurde, steht an einem ganz normalen Werktag nahezu leer. Von den rund 160 Stellplätzen sind nur eine Handvoll belegt. Das Bild der Einsamkeit wird durch die wenigen Autos unterstrichen, die wie verloren auf den weiten Betonflächen parken.

Die Idee der Pendler: Schienenersatzverkehr bis zur Garage

Die geringe Nutzung ist kein Zufall. Die Anlage liegt zwar verkehrsgünstig an der B 375, aber der Umstieg auf die Straßenbahn ist umständlich. Von der Garage muss man einen etwa 800 Meter langen Fußweg zur Haltestelle Fölling der Linie 1 zurücklegen – eine Strecke, die bei schlechtem Wetter oder schwerem Gepäck sehr unattraktiv ist. Genau hier setzen die Rufe der Pendler an: „Warum fährt nicht einfach der Schienenersatzverkehr bis Fölling?“, fragen sich viele, die auf die öffentlichen Verkehrsmittel umsteigen müssten. Die Idee klingt einfach: Wenn die Straßenbahnlinie 1 wegen Bauarbeiten durch Busse ersetzt wird, könnten diese Busse doch direkt bis vor die Parkhaustüre fahren. Die Pendler müssten dann nicht mehr den beschwerlichen Fußmarsch auf sich nehmen.

Die beharrliche Ablehnung der Graz Linien

Doch so einfach ist die Umsetzung offenbar nicht. Die Graz Linien verweisen auf mehrere Hürden. Erstens sei die Trasse der Straßenbahn nicht mit Bussen befahrbar, da die Gleise zwischen Haltestelle Fölling und der Garage eine Wendeschleife der Bahn darstellen. Busse könnten dort nur mit erheblichem Aufwand wenden. Zweitens seien die Fahrplanzeiten für den Schienenersatzverkehr exakt auf die Bauarbeiten abgestimmt; eine Verlängerung bis zur Garage würde zusätzliche Fahrzeit bedeuten, die den gesamten Takt durcheinanderbringen könnte. Drittens gebe es rechtliche und genehmigungstechnische Fragen, da die Garage auf einem Grundstück stehe, das nicht zum Liniennetz gehöre. Die Verantwortlichen betonen, dass man die Situation analysiere, aber eine einfache Lösung stehe nicht zur Verfügung.

Ein teures Projekt ohne Nutzen?

Die Geschichte des Parkhauses Fölling ist geprägt von Fehlplanungen und unerfüllten Erwartungen. Die Hochgarage wurde 2010 mit dem Ziel gebaut, den Parkdruck in den Wohngebieten von Mariatrost zu verringern und eine Park-and-Ride-Option für Pendler aus dem Norden von Graz zu schaffen. Damals gab es auch die Idee, eine zusätzliche Haltestelle der Straßenbahn direkt bei der Garage zu errichten, was allerdings aus Kostengründen verworfen wurde. Stattdessen wurde der vorhandene, aber unattraktive Fußweg als ausreichend erachtet. Nach der Fertigstellung blieb die Garage jahrelang ein „Murksprojekt“, wie es Stadtrat Kurt Hohensinner (ÖVP) nannte. Die Auslastung lag jahrelang unter 20 Prozent. Die Stadt Graz versuchte dann im Mai 2025 mit der Gratisparken-Aktion einen neuen Anlauf, aber offenbar ohne durchschlagenden Erfolg.

Die Kosten für die Stadt: Millionen ohne Einnahmen

Durch die kostenlose Parkmöglichkeit entgehen der Stadt nicht nur ursprüngliche Parkgebühren, sondern sie muss auch die Betriebskosten der Garage tragen. Dazu gehören Strom, Reinigung, Wartung und eventuelle Sicherheitsmaßnahmen. Schätzungen aus dem Jahr 2024 gingen von jährlichen Kosten in Höhe von rund 50.000 Euro aus, die nun ohne Gegenwert anfallen. Die 2,7 Millionen Euro Baukosten sind längst abgeschrieben, doch der Wert der Anlage bleibt ein Prestigeprojekt, das keiner nutzt. Die Opposition im Grazer Gemeinderat fordert daher seit Jahren entweder eine massive Attraktivierung der Anbindung oder sogar die Schließung der Garage.

Die größere Bedeutung: Park-and-Ride in Graz

Der Fall Fölling steht stellvertretend für die Herausforderungen von Park-and-Ride-Konzepten in einer wachsenden Stadt. Graz hat in den letzten Jahren mehrere große Park-and-Ride-Anlagen errichtet, insbesondere an den Hauptverkehrsachsen: im Süden bei Webling, im Osten an der Liebenauer Tangente und im Osten beim Messegelände. Während die Anlagen an den Endhaltestellen der Straßenbahnlinien 4, 5 und 6 stark frequentiert sind, bleiben die dezentralen Standorte wie Fölling untergenutzt. Der Grund liegt in der direkten Anbindung: Wer in Webling parkt, steigt direkt in die Bim um und ist in zehn Minuten in der Innenstadt. In Fölling dagegen muss man nach dem Parken noch 15 Minuten zu Fuß zur Haltestelle gehen und dann 20 Minuten mit der Straßenbahn fahren – insgesamt länger als die direkte Autofahrt in die Stadt. Solange die letzte Meile nicht attraktiv gestaltet wird, wird das Konzept nicht aufgehen.

Die Stimmen der Pendler: Frustration und Resignation

Vor dem Parkhaus, das an diesem Vormittag fast menschenleer wirkt, treffen wir einen Pendler, der gerade aus seinem Auto aussteigt. „Ich bin heute das erste Mal hergekommen, weil ich gehört habe, dass es gratis ist“, sagt der 34-jährige Bankangestellte. „Aber nachdem ich gesehen habe, wie weit die Haltestelle weg ist, bin ich doch lieber mit dem Auto weiter in die Stadt gefahren. Das lohnt sich einfach nicht.“ Andere Pendler, die auf dem Weg zur Garage vorbeikommen, äußern ähnliche Gefühle. „Ich habe schon vor Jahren vorgeschlagen, dass man einen kleinen Shuttle-Bus von der Garage zur Haltestelle einsetzt“, meint eine 45-jährige Lehrerin. „Aber die Stadt hat nicht reagiert. Jetzt ist es gratis, aber deswegen parke ich dort trotzdem nicht. Der Zeitverlust ist es mir nicht wert.“

Die Suche nach Lösungen: Was könnte funktionieren?

Experten für Verkehrsplanung nennen mehrere erfolgversprechende Ansätze für eine Belebung des Standorts. Eine Möglichkeit wäre die Etablierung einer regelmäßigen Buslinie, die die Garage direkt an die Straßenbahnhaltestelle anbindet. Dies könnte als Zubringer oder sogar als Verlängerung einer bestehenden Buslinie realisiert werden. Zudem könnte eine bessere Beschilderung und Wegweisung den Autofahrern den Standort ins Bewusstsein rufen. Kritiker weisen jedoch darauf hin, dass eine dauerhafte Busanbindung Kosten verursache, die in keinem Verhältnis zum Nutzen stünden, solange die Garage so wenige Nutzer habe. Eine weitere Idee ist die Verlegung der Straßenbahnhaltestelle näher zur Garage – ein Projekt, das schon 2010 erwogen, aber wegen der hohen Kosten von rund einer Million Euro verworfen wurde. Die aktuellen Planungen seitens der Stadt Graz sehen immerhin vor, die Verkehrssituation im gesamten Bereich Mariatrost zu überprüfen, was möglicherweise zu einer Neuausrichtung führt.

Die Zukunft des Standorts: Hoffnung auf Besserung

Die Grazer Stadtregierung, die im Herbst 2025 eine Gemeinderatswahl vor sich hat, ist bemüht, den Missstand zu beheben. Die zuständige Stadträtin Judith Schwentner (Grüne) kündigte an, dass die Verwaltung konkrete Vorschläge erarbeiten solle, wie die Anbindung verbessert werden könne. Man wolle auch mit den Graz Linien über eine mögliche Verlängerung von Schienenersatzverkehren während der Sommerbaustellen verhandeln. Bis dahin bleiben die meisten Parkplätze leer, und die 2,7 Millionen Euro teure Garage wirkt wie ein Mahnmal verfehlter Stadtplanung. Während andere Städte wie Wien oder München erfolgreiche Park-and-Ride-Konzepte betreiben, zeigt das Beispiel Graz, wie wichtig es ist, die Schnittstelle zwischen Auto und öffentlichem Verkehr wirklich zu durchdenken – nicht nur mit dem Bau einer Garage, sondern mit dem entscheidenden letzten Schritt zur Haltestelle.

Source: Kleinezeitung News

Die angespannte Stimmung vor der Hochzeit

Palastinsider berichteten, dass die Stimmung im Vorfeld der Hochzeit ziemlich angespannt und stressig gewesen sei. Die Autorin Tina Brown (72) bezeichnete die Situation in ihrem Buch "The Palace Papers" als "ein einziges Chaos" und zitierte eine Quelle mit den Worten, es sei "die ganze Zeit über nichts als Drama" gewesen. Mitarbeiter hätten die ständigen Forderungen von Harry und Meghan "satt" gehabt. Diese Darstellung deckt sich mit vielen Berichten, die in den folgenden Jahren ans Licht kamen. Der Königshaus-Experte Tom Bower (79) behauptet in seiner Biografie "Revenge", Meghan sei "zu anspruchsvoll" gewesen. Vor allem der Umgang mit der königlichen Etikette und den Traditionen des Palastes habe zu Konflikten geführt.

Der Tiara-Streit mit der Queen

Ein besonders prominenter Konflikt drehte sich um die Hochzeitstiara. Meghan hatte sich angeblich einen Kopfschmuck mit Smaragden gewünscht, und Harry unterstützte seine Frau in diesem Wunsch. Doch Angela Kelly (68), die persönliche Modeberaterin der verstorbenen Queen Elizabeth II. (1926–2022), riet wegen der russischen Herkunft der Schmucksteine ab. Die Queen musste schließlich persönlich eingreifen. Laut "The Times" soll sie Harry deutlich in seine Schranken gewiesen haben: "Meghan kann nicht alles haben, was sie will. Sie bekommt die Tiara, die ich ihr gebe." Meghan entschied sich am Ende für das "Queen Mary's Diamond Bandeau", ein elegantes Diadem aus Diamanten. Dieser Vorfall zeigt, wie tief die Spannungen bereits vor der Hochzeit waren – es ging nicht nur um ein Schmuckstück, sondern um die Frage, wer in der königlichen Hierarchie das Sagen hat.

Der Kleider-Streit mit Prinzessin Kate

Ein weiteres Drama soll sich um die Kleider der Blumenmädchen abgespielt haben. Laut Bower war Harrys Schwägerin, Prinzessin Kate (44), der Meinung, dass die Mädchen der royalen Tradition folgen und Strumpfhosen tragen sollten. Meghan bestand jedoch offenbar darauf, es anders zu handhaben. Außerdem soll es eine Diskussion um die Länge von Prinzessin Charlottes (11) Rock gegeben haben: Kate sei dieser für ihre damals dreijährige Tochter zu kurz gewesen. Meghans Assistentin Melissa Toubati und die Designerin Clare Waight Keller gaben an, "miterlebt zu haben, wie Meghan Kates Einwand entschieden zurückwies", so Bower. Kate sei daraufhin in Tränen ausgebrochen. Meghan wurde in der Presse als "Diva und Bridezilla" bezeichnet.

Die Herzogin von Sussex legte 2021 in ihrem TV-Interview mit Oprah Winfrey (72) ihre Sicht der Dinge dar. Sie behauptete, sie habe bei dem Vorfall geweint, nicht Kate. "Ein paar Tage vor der Hochzeit war sie wegen etwas aufgebracht – ja, das Thema stimmte –, das mit den Kleidern der Blumenmädchen zu tun hatte, und das brachte mich zum Weinen und verletzte meine Gefühle wirklich", erzählte Meghan Oprah. "Was schwer zu verkraften war, war, für etwas verantwortlich gemacht zu werden, das ich nicht nur nicht getan hatte, sondern das mir widerfahren war." Dieser Disput symbolisiert die tiefen Missverständnisse zwischen den Schwägerinnen, die nie vollständig ausgeräumt wurden.

Waren Harry und Meghan schon vorher verheiratet?

In dem Interview mit Oprah Winfrey behauptete Meghan außerdem, dass sie Harry drei Tage vor ihrer großen Hochzeit in Windsor bereits heimlich geheiratet hatte. "Niemand weiß das, aber wir haben den Erzbischof angerufen und einfach gesagt: 'Hören Sie, diese Sache, dieses Spektakel ist für die Welt, aber wir wollen, dass unsere Verbindung nur zwischen uns besteht.'" Meghan erzählte dann, dass sie die private Zeremonie im Garten von Nottingham Cottage auf dem Gelände des Kensington-Palastes abgehalten hätten. Offizielle Dokumente und auch der damalige Erzbischof Justin Welby (70) selbst widersprachen dieser Behauptung allerdings. Laut "BBC" soll es sich bei der Zeremonie vor der Hochzeit lediglich um "im privaten Rahmen ausgetauschte persönliche Gelübde" gehandelt haben. Diese Episode zeigt, wie sehr das Paar versuchte, seine Beziehung von der öffentlichen Inszenierung zu trennen – ein Motiv, das später auch ihren Rückzug aus dem Königshaus prägte.

Meghans Vater sorgt für einen Skandal

Für den größten Skandal sorgte vor der Hochzeit von Harry und Meghan aber der Vater der Braut. Die königliche Familie hatte angekündigt, dass Meghans Vater, Thomas Markle (81), seine Tochter zum Altar führen wird. Zwei Tage vor der Hochzeit veröffentlichte Meghan jedoch eine Erklärung, in der sie mitteilte: "Leider wird mein Vater nicht an unserer Hochzeit teilnehmen. Ich habe mich immer um meinen Vater gekümmert und hoffe, dass ihm der Freiraum gegeben wird, den er braucht, um sich auf seine Gesundheit zu konzentrieren." Die Zeitung "Mail on Sunday" hatte zuvor enthüllt gehabt, dass Markle, der damals in Mexiko lebte, dabei geholfen hatte, Paparazzi-Fotos von sich selbst bei den Vorbereitungen für die Hochzeit zu inszenieren.

Thomas Markle erklärte dann kurz vor der Hochzeit der US-Promi-Seite "TMZ", er habe einen Herzinfarkt erlitten. Meghan erfuhr angeblich erst durch diesen Bericht von der Einweisung ihres Vaters ins Krankenhaus. Prinz Harrys Vater, König Charles (77), führte Meghan schließlich zum Altar. Zu ihrem Vater soll sie den Kontakt abgebrochen haben. Diese Episode belastete die Beziehung zwischen Meghan und der königlichen Familie zusätzlich, da die ständigen Enthüllungen aus der Familie der Braut als peinlich empfunden wurden.

Die Vorgeschichte: Meghan und die königliche Familie

Um die Dramen rund um die Hochzeit vollständig zu verstehen, muss man einen Blick auf die Vorgeschichte werfen. Meghan Markle war bereits vor der Verlobung eine erfolgreiche Schauspielerin, bekannt durch die Serie "Suits". Ihr Eintritt in die königliche Familie war von Anfang an von kulturellen und medialen Spannungen begleitet. Als Frau mit afroamerikanischen Wurzeln und geschieden stand sie unter besonderer Beobachtung. Harry selbst hatte in Interviews später betont, dass er sich von der Presse und dem Palast im Stich gelassen fühlte. Schon bei der Verlobungsbekanntgabe im November 2017 gab es Berichte über Reibereien: Meghan soll sich gegen die traditionelle Verlobungsfotosession gewehrt haben, bei der das Paar vor laufenden Kameras posieren musste.

Der Palast war bemüht, Meghan in die königlichen Abläufe zu integrieren, aber es gab immer wieder Missverständnisse. Meghans Hang zu selbstbestimmten Entscheidungen – etwa als sie in der Öffentlichkeit die Tür ihres Autos selbst öffnete oder unkonventionelle Kleidung trug – stieß im konservativen Umfeld des Hofes oft auf Unverständnis. Parallel dazu wuchs die Kritik an der Doppelmoral der britischen Presse, die Meghan und Kate völlig unterschiedlich behandelten. Während Kates Fehltritte meist mit Nachsicht bedacht wurden, wurde Meghan schnell als anspruchsvoll und schwierig abgestempelt.

Die Hochzeit selbst: Ein Spektakel für die Welt

Die Hochzeit am 19. Mai 2018 in der St. George’s Chapel auf Schloss Windsor war ein Medienereignis erster Güte. Rund 19 Millionen Briten verfolgten die Übertragung live, weltweit waren es schätzungsweise zwei Milliarden Menschen. Der Gottesdienst wurde von Erzbischof Justin Welby geleitet, und das Paar tauschte persönlich geschriebene Gelübde aus. Die Predigt hielt der US-amerikanische Bischof Michael Curry, der mit leidenschaftlichen Worten über die Macht der Liebe sprach. Meghan trug ein schlichtes, elegantes Kleid von Givenchy, entworfen von Clare Waight Keller. Die Hochzeit war betont modern und inklusiv: Ein Gospelchor sang, und die afroamerikanische Bürgerrechtlerin Martin Luther Kings Zitate wurden zitiert. Doch hinter der glanzvollen Fassade brodelte es weiter.

Die Folgen: Der Bruch mit der Monarchie

Nur zwei Jahre nach der Hochzeit, im Januar 2020, verkündeten Harry und Meghan überraschend ihren Rückzug als hochrangige Mitglieder der königlichen Familie. Sie zogen nach Kanada, später nach Kalifornien. Die Beziehung zu den Windsors war zerrüttet. In ihrem Oprah-Interview im März 2021 sprachen die beiden offen über die Schwierigkeiten, die sie erlebt hatten: rassistische Untertöne in der Presse, mangelnde Unterstützung durch den Palast und die Isolation, die Meghan in eine tiefe Krise gestürzt hatte. Der Streit um die Tiara und die Blumenmädchen-Kleider erschien da nur als Vorgeschmack auf die tiefer liegenden Konflikte.

Analysen und Einschätzungen von Experten

Königshaus-Experten wie Tom Bower und Tina Brown haben in ihren Büchern detailliert die Machtkämpfe innerhalb des Palastes beschrieben. Bower sieht in Meghan eine ehrgeizige Frau, die die starren Regeln des Königshauses nicht akzeptieren wollte. Brown hingegen kritisiert das System des Palastes, das nicht in der Lage war, mit modernen Herausforderungen umzugehen. Die Hochzeit war aus ihrer Sicht der Höhepunkt einer jahrelangen Eskalation. Andere Beobachter wie der Historiker Robert Hardman betonen, dass Harry und Meghan selbst durch ihr Verhalten – wie die geheime Vermählung drei Tage vorher – das Vertrauen des Palastes verspielt hätten. Fakt ist: Die Hochzeit, die als Märchen begann, entwickelte sich zu einer der größten Zerreißproben für das britische Königshaus seit der Abdankungskrise 1936.

Die Frage, ob die Skandale um die Hochzeit von Meghan und Harry übertrieben dargestellt wurden, bleibt umstritten. Während einige Medien die Berichte über Bridezilla-Verhalten kolportierten, betonen Unterstützer des Paares, dass Meghan von Anfang an mit unfairen Maßstäben gemessen wurde. Die Wahrheit liegt vermutlich in der Mitte: Ein Zusammenprall zweier Welten – der traditionellen, formellen Monarchie und einer selbstbewussten, modernen Frau – war fast unvermeidlich. Die Hochzeit war der Moment, in dem dieser Konflikt erstmals öffentlich sichtbar wurde, auch wenn die Kameras der Welt nur das Lächeln einfingen.

Die Rolle der Medien

Ein Aspekt, der bei der Betrachtung der Hochzeitsdramen nicht vernachlässigt werden darf, ist die Rolle der britischen Boulevardpresse, insbesondere der "Mail on Sunday" und der "Sun". Sie waren es, die die Skandale um Meghans Vater aufdeckten, und sie waren es, die die Geschichten über die Auseinandersetzungen zwischen Meghan und Kate verbreiteten. Harry hat mehrfach betont, dass die Medienhetze gegen seine Frau ein wesentlicher Grund für den Auszug aus Großbritannien war. Der Rechtsstreit gegen die "Mail on Sunday" wegen der Veröffentlichung eines privaten Briefes von Meghan an ihren Vater endete 2022 mit einem Teilerfolg für Meghan. Der Fall verdeutlichte, wie tief das Misstrauen zwischen den Sussexes und der Presse war.

Die Hochzeit von Harry und Meghan bleibt ein faszinierendes Ereignis, das viele Jahre später immer noch nicht alle Geheimnisse preisgegeben hat. Die Enthüllungen in Büchern und Interviews haben das Bild einer perfekten royalen Hochzeit nachhaltig zerstört. An ihre Stelle trat die Einsicht, dass hinter den Kulissen des britischen Königshauses oft das gleiche Chaos und die gleichen menschlichen Schwächen herrschen wie in jeder anderen Familie. Nur dass hier die ganze Welt zusieht.

Source: GMX News News