Vue lecture

Il y a de nouveaux articles disponibles, cliquez pour rafraîchir la page.

Anthropic Makes Claude Opus 4.7 the Default in Claude Code Fast Mode

Anthropic has quietly made a massive move for developers. The company’s most powerful AI model, Claude Opus 4.7, is now the default model inside Claude Code fast mode. And the wild part?You only need one command: /fast That’s it. No extra setup. No complicated configuration. No additional payment tier. Developers using Claude Code can now […]

Claude Mythos 1 and Opus 4.8 Leak Online — Anthropic May Be Preparing Its Biggest AI Upgrade Yet

New leaks suggest Anthropic is getting ready to launch multiple next-generation AI systems, including the mysterious Claude Mythos 1 and an upgraded Claude Opus 4.8 model. The unreleased models recently appeared across several backend systems and internal UI references, hinting that Anthropic could be entering the final stages of rollout testing. Claude Mythos 1 Preview […]

Claude for Marketing Introduced: AI Can Now Create and Post Viral Content Automatically

Artificial intelligence is rapidly transforming marketing, and the latest announcement could change how brands manage social media forever. Claude has introduced Claude for Marketing, a new AI-powered workflow designed to help businesses deploy social media accounts, generate viral-ready content, and automatically publish posts — all from a single chat interface. The announcement highlights how marketers […]

LUV1 modular bike replaces your car for daily errands with 120L storage and swappable batteries

Most electric motorcycles still behave like motorcycles first and utility machines second. They chase performance numbers, oversized displays, or aggressive styling while ignoring a simple reality: most urban riders just want something practical enough to replace short car trips. The ANY LUV1 approaches the problem differently. Instead of behaving like a sportbike with batteries attached, it feels more like a compact urban tool designed around everyday life.

Created by Belgian startup ANY Mobility, LUV1 is sandwiched somewhere between an electric scooter, cargo bike, and lightweight motorcycle. The company calls it a “Life Utility Vehicle,” and the name makes sense once you look beyond the styling. Nearly every part of the vehicle revolves around usability, whether that means carrying groceries, office gear, camera equipment, or handling the kind of short-distance errands people usually default to using a car for.

Designer: ANY Mobility

That practicality starts with its packaging. The integrated cargo compartment offers 120 liters of storage, which is significantly more useful than the tiny under-seat compartments found on most scooters. It is large enough to carry shopping bags, delivery equipment, or a backpack and helmet without forcing riders to strap everything externally. Front and rear cargo racks expand that flexibility further, while configurable dividers allow owners to organize storage depending on the task at hand.

The modular approach is where the concept becomes genuinely interesting. Instead of locking owners into one fixed setup, the LUV1 can be customized with interchangeable body panels, seating layouts, storage accessories, and optional weather protection. One configuration can prioritize cargo hauling during the week while another leans toward casual commuting on weekends. It follows the same logic that made modular furniture and adaptable workspaces appealing: people increasingly want products that evolve with their routines rather than forcing routines around the product.

Visually, the bike avoids the exaggerated “future mobility” look many startups lean on. The clean bodywork and restrained surfacing come from Granstudio, the Italian design firm led by former Pininfarina design director Lowie Vermeersch. That design pedigree shows in the proportions and detailing. Even functional components like the storage compartments and structural frame feel integrated into the design rather than added as an afterthought.

Underneath the bodywork sits a modular aluminum chassis produced using high-pressure die-casting, a manufacturing method more commonly associated with larger automotive companies. The setup helps reduce complexity while providing the platform with enough flexibility to support various accessories and future configurations. Power comes from an 11 kW rear hub motor paired with dual swappable lithium-ion battery packs totaling 6.5 kWh. ANY Mobility claims a range of 68 to 87 miles, depending on use, while the top speed is rated at 62 mph, making the bike suitable for both dense city streets and suburban commuting. Charging the batteries through a standard 220V outlet reportedly takes under four hours.

The LUV1 also keeps accessibility in mind. It weighs around 352 pounds and features a relatively approachable 30.9-inch seat height, making low-speed maneuvering less intimidating for newer riders and shorter commuters alike. According to reports, the company expects pricing to fall between €7,000 and €10,000 (approximately $8,150 – $11,600) depending on configuration, and reservations have already opened ahead of production plans.

What makes the ANY LUV1 stand out is not raw performance or futuristic gimmicks, but how realistically it understands modern urban mobility. Most people are not looking for an electric motorcycle to replace weekend entertainment. They are looking for something convenient enough to replace unnecessary car usage, and the LUV1 feels designed precisely around that idea.

The post LUV1 modular bike replaces your car for daily errands with 120L storage and swappable batteries first appeared on Yanko Design.

Waymo’s Self-Driving Car Saw the Flood and Drove In Anyway. Here’s The Problem Plaguing Every Robotaxi.

Every sensor on a Waymo robotaxi sees the world in layers. The LiDAR maps it in three dimensions, radar bounces through it, and cameras read it in color and contrast, building a composite picture of the road that no human retina could match at the same fidelity. So when a Waymo encountered a flooded section of a 40 mph road in San Antonio on April 20, the car absolutely saw the water. It slowed down for it. Then it drove in anyway, floated off the road surface, and came to rest in Salado Creek. The voluntary recall Waymo filed with NHTSA on April 30, covering 3,791 vehicles, was triggered not by a sensor that missed a hazard, but by a software stack that saw the hazard clearly and still chose the wrong response.

You might be sitting in one of those 3,791 recalled vehicles right now, somewhere in Phoenix, Los Angeles, Austin, or Atlanta, and Waymo has confirmed the permanent software fix is still in development. Tesla’s Cybercab, entering production at Giga Texas, runs a supervised robotaxi service in Austin, Dallas, and Houston on a pure-vision architecture with no LiDAR whatsoever. Uber’s platform in Dallas is dispatching Avride-operated Hyundai Ioniq 5s that are currently under NHTSA investigation for 16 crashes involving lane changes and failure to stop for traffic ahead. Amazon’s Zoox uses cameras, LiDAR, radar, and long-wave infrared on every vehicle, the most sensor-redundant consumer-facing stack in the industry, and is still in limited city testing. Each platform has a different answer to what a self-driving car should do when it encounters something it cannot traverse, and after the San Antonio creek, all of those answers deserve a much closer look.

The NHTSA recall notice characterizes the flaw precisely: the software “may allow the vehicle to slow and then drive into standing water on higher speed roadways.” That is a classification error buried in the decision stack, not a sensor failure, and the distinction matters more than the recall number suggests. Waymo’s 5th-gen Jaguar I-Pace and 6th-gen Zeekr RT both carry LiDAR, radar, and cameras in overlapping fields of view, and the San Antonio car processed the flooded road accurately as a hazard worth responding to. The decision architecture, however, had no hard-stop condition for water on a 40 mph road, only a caution flag that reduced speed and left proceeding as an available output. A separate Waymo had already been stranded near McCullough Avenue in San Antonio roughly two weeks before the April 20 incident, confirming this was a repeatable failure mode across a fleet that was still carrying passengers in nine other cities.

Tesla’s Cybercab carries no LiDAR, putting its supervised fleet in Austin, Dallas, and Houston in a fundamentally different position when floodwater appears than Waymo’s overlapping sensor stack would. The platform relies on eight cameras and 4D millimeter-wave radar, meaning no independent depth-sensing channel exists to assess water severity when camera visibility degrades in heavy rain. A real-world FSD 14.3.1 test in April 2026 ended in manual takeover when the front bumper camera submerged, a precise illustration of where the vision-only approach runs out of information. Avride, dispatching Hyundai Ioniq 5s through Uber’s Dallas app since December, is under concurrent NHTSA investigation for 16 crashes involving lane changes and failures to stop for road hazards, all 16 occurring with a trained safety monitor seated in the vehicle. Amazon’s Zoox sits at the opposite end of the sensor redundancy spectrum, combining cameras, LiDAR, radar, and long-wave infrared in a 360-degree array with a human TeleGuidance fallback for scenarios the stack cannot resolve, though its commercial footprint remains a fraction of Waymo’s.

The Waymo recall, the Avride probe, and a dashcam video of a Waymo rolling through a red light on Irving Boulevard in Dallas all surfaced in the same seven-day window, collectively mapping the same design gap across three platforms: a perception-to-action pipeline that detects a hazard but generates the wrong response to it. Waymo’s OTA patch is deploying now, but the permanent fix remains in development, meaning every current ride runs on interim constraints rather than a finished solution. The San Antonio incident involved an empty car, and that is the only reason this story ends with a recovery operation rather than a casualty report. Each platform carrying passengers today is still writing its edge-case rulebook, publishing each new chapter only after something breaks on a live road. Knowing which system you are riding in, what its sensor stack can assess in a sudden storm, and whether its flood-detection logic has been patched from an interim fix to an actual solution is, I’d argue, the most practical safety question a passenger can ask in 2026.

The post Waymo’s Self-Driving Car Saw the Flood and Drove In Anyway. Here’s The Problem Plaguing Every Robotaxi. first appeared on Yanko Design.

Des faux installateurs Claude Code se baladent dans Google

Des faux installateurs Claude Code se baladent dans Google

Les chercheurs d'Ontinue, une boîte de cybersécurité, ont publié une analyse d'une campagne de vol de données qui vise directement les développeurs.

La méthode est en fait assez simple. Vous tapez "install claude code" dans Google, vous cliquez, plein d'innocence, sur le premier résultat sponsorisé, et là vous tombez sur une page qui ressemble trait pour trait à la doc officielle d'Anthropic. Sauf que le serveur derrière n'a rien à voir, et la commande d'installation à copier-coller a été modifiée.

L'astuce technique est intéressante. La commande PowerShell affichée n'est pas visible dans le code source qu'un scanner automatique pourrait analyser, elle est générée à la volée dans le HTML de la page. Du coup, les scans de sécurité voient du contenu légitime, l'utilisateur copie une commande malveillante, et paf, un loader PowerShell d'environ 600 Ko se lance discrètement.

Ce loader essaie ensuite quelque chose de carrément vicieux. Il abuse de IElevator2, une interface interne de Chromium (le moteur derrière Chrome, Edge, Brave, Vivaldi, Opera) lancée en janvier 2026 par Google pour mieux protéger les cookies et mots de passe avec un chiffrement renforcé. Le malware injecte un petit module dans le navigateur pour appeler cette interface depuis l'intérieur, ce qui lui permet de récupérer les clés et de déchiffrer toute la base. Cookies de session, mots de passe enregistrés, infos de paiement. Tout y passe.

Trois domaines pirates ont été enregistrés sur six jours en avril, tous derrière Cloudflare pour compliquer le takedown. Le bout de code utilisé ne correspond à aucune famille de malware connue (Lumma, StealC, Vidar). Le plus proche serait Glove Stealer, repéré en novembre 2024, mais avec une orchestration différente. Bref, ce serait un nouveau venu fabriqué exprès pour cette campagne.

Et la cible, c'est précisément les développeurs. Ce sont eux qui installent Claude Code (l'outil en ligne de commande d'Anthropic pour coder avec l'IA, concurrent direct de Cursor).

Et un développeur, dans son navigateur, c'est l'accès à GitHub, à AWS, à des dashboards de production, à des comptes Cloud, à des secrets qui se revendent potentiellement très cher. Les défenses classiques qui surveillent les exécutables natifs ne voient rien, tout se joue au niveau de l'appel COM (la mécanique Windows qui permet à des programmes de se parler entre eux) et de PowerShell.

Petit conseil : passez par anthropic.com directement, jamais par un résultat sponsorisé. Si vous avez collé une commande douteuse récemment, c'est le moment de regarder ce qui tourne sur votre machine.

Source : Infosecurity Magazine

Spotify génère maintenant vos podcasts tout seul

Spotify vient d'annoncer un truc intéressant je trouve, qui s'appelle les Personal Podcasts. Le principe c'est de demander absolument tout ce que vous voulez, par exemple un podcast sur un cours que vous venez de suivre, sur un bouquin, sur un article de Korben.info voire sur votre planning de la semaine, vos objectifs...etc etc... Peu importe... Spotify prend tout ça, et génère un épisode audio personnalisé raconté par une voix IA plus ou moins moche.

Et l'épisode apparaît alors directement dans votre bibliothèque, comme si quelqu'un avait fait un résumé audio de votre semaine.

Et pour faire ça, ils ont mis en ligne un repo GitHub save-to-spotify qui est un outil en ligne de commande permettant à des agents IA de créer ce podcast personnalisé tout ça accessible en ligne de commande ou via des agents comme Claude Code, OpenClaw ou Codex.

Que ce soit sous macOS ou Linux, ça s'installe en une ligne (allez lire le install.sh par sécurité quand même avant de l'exécuter) :

curl -fsSL https://saveto.spotify.com/install.sh | bash

Et si vous êtes sous Claude Code, c'est encore plus immédiat :

/plugin marketplace add spotify/save-to-spotify

Le skill se retrouve dans ~/.claude/skills/save-to-spotify/ et votre agent peut demander à Spotify de générer un podcast à la demande pour ensuite le pousser sur Spotify.

Rien de compliqué en fait !

Par contre, le podcast créé est 100% privé, donc vous ne pourrez pas le partager avec vos amis. Mais c'est pas bloquant non plus puisqu'il est toujours possible d'aller récupérer dans les dossiers temporaires de génération de l'émission les MP3 que ça vous crache pour ensuite les mettre ailleurs, soit sur votre site, soit les diffuser sur votre vrai podcast Spotify accessible à tous.

Maintenant, est-ce que je vais faire mon podcast pour raconter les actus que je mets sur ce site ?

Alors j'ai pas le temps mais je le ferai peut-être un jour si la qualité audio de l'IA est suffisante pour que ça ait l'air vraiment produit par un humain et pas par une machine. Pour vous donner une idée, voici ce que ça donne :

Donc c'est pas encore qualitatif... À voir en passant par des moteurs TTS comme ceux d'ElevenLabs... mais pour l'instant, c'est pas d'actualité pour moi. On verra bien... Je me suis quand même amusé à mettre les fichiers texte et JSON produits dans Notebook LM pour faire un autre type de podcast qui cette fois est un peu plus long et plus quali... Je vous mets ici.

Après, peu importe que vous le génériez via l'outil de Spotify ou autrement en passant par un autre outil, le CLI Save To Spotify, vous permettra ensuite de le mettre sur votre compte Spotify pour l'écouter par exemple dans la voiture ou dans les transports.

Bref, c'est disponible et si vous avez Claude Code sous la main, ça prend une ligne à installer. L'annonce complète est par ici .

DATO Ares Q4 - Le SSD aimanté qui se colle partout

Le truc cool avec un disque de backup, c'est quand vous arrêtez de le sortir du tiroir. Et c'est ce que permet de faire le DATO Ares Q4 (lien affilié) qui se trouve être aimanté et que je peux donc coller direct sur le côté de mon Mac Studio . Je l'ai depuis bientôt un an, et c'est devenu mon disque de Time Machine à résidence.

Il est trop mignon parce qu'il est tout petit, puisqu'il mesure 64 par 64 mm, 12 mm d'épaisseur, et 40 grammes, soit à peine plus gros qu'un bloc de petits Post-it, sauf que dedans y'a 4 To de NVMe en USB4. Les aimants sont prévus pour s'accrocher à un iPhone 15/16/17 Pro façon MagSafe, mais en pratique ils tiennent aussi nickel sur n'importe quelle machine, si vous collez le petit aimant-sticker livré avec.

Après je sais pas si coller un aimant sur un ordinateur c'est le move le plus intelligent que j'ai fait de ma vie mais je l'ai collé y'a 1 an, et ça bouge pas et j'ai aucun souci. Après c'est pas des GROS aimants non plus hein... C'est assez fin et pas très aimanté non plus. Et niveau connectique, c'est livré avec un câble USB-C rikiki de 10 cm.

Côté débits l'USB4 crache du 4000 Mo/s en lecture et 3600 Mo/s en écriture et comme c'est du natif, y'a pas de puce intermédiaire entre le NVMe et le port, donc on est max de ce qu'on peut avoir en vitesse avec ce genre de disque. Par contre, faut un port USB4 ou Thunderbolt 4/5 côté Mac, sinon ça retombe sur de l'USB 3.2 classique. C'est pas dramatique mais ça va 'achement moins vite quand même.

Et comme mon Mac Studio embarque 4 To en interne, c'est pile la taille du DATO que je dédie à ma sauvegarde Time Machine. Le sparsebundle peut donc grossir jusqu'à occuper tout le disque sans que je m'en occupe ! Et comme je ne le remplis jamais à 100%, je suis assez tranquille.

En plus de ce SSD, je fais toujours des backups sur mon NAS Synology avec Carbon Copy Cloner à côté parce que je suis parano et qu'on sait jamais...

Côté chauffe, la surprise est plutôt bonne, il reste froid / tiède même en plein transfert. Alors je suis pas sûr mais je me dis que le boîtier en alu du Mac Studio, joue peut-être un petit rôle de dissipateur thermique mais j'sais pas, peut-être pas... En tout cas, ça chauffe pas quoi.

Ce disque est compatible Mac, PC, Linux, et même iPhone Pro en USB-C comme ça si vous avez besoin de filmer de la 4K ProRes HDR direct sur un SSD plutôt que sur la mémoire interne, c'est possible ! Et il est garanti 5 ans, ce qui est toujours bon à prendre.

Voilà, si ça vous intéresse, le DATO Ares Q4 (lien affilié) est dispo sur Amazon en versions 1 To, 2 To ou 4 To .

This Ceramic Vase Is Actually a Phone Speaker That Needs No Power

The home has become increasingly cluttered with gadgets that need charging, pairing, and their own dedicated spaces. Even something as simple as playing music from a smartphone often involves a Bluetooth speaker sitting on a shelf, waiting for its battery to drain. There’s been a quiet counter-movement in product design, where objects do their jobs without power and sit in a room the way a vase or a mug would.

Kenji Abe’s ECHO is exactly that kind of object. It’s an analog speaker that amplifies smartphone audio simply by being set on top of the phone, requiring no power, no pairing, and no setup beyond placing it down. The concept takes its cues from wind instruments and seashells, two forms that have been shaping and projecting sound for centuries without the help of electricity.

Designer: Kenji Abe

The inside of ECHO works like a chamber, built to catch the phone’s audio and carry it outward in soft, diffused waves rather than projecting it directly. The geometry draws from the same logic as a cupped hand, but with more control over how sound travels. The result isn’t a dramatic volume boost so much as a room-filling quality that feels warmer than a powered speaker on a desk.

The choice of material makes as much of a statement as the form. Abe uses glazed ceramic, the same material found in vases, mugs, and tableware, giving ECHO a texture and presence that belongs in a home rather than on a tech shelf. It doesn’t look like an accessory. It looks like something that was always there, something that simply happened to be placed near a phone.

That quality matters when the phone is on the kitchen counter and you want music while cooking, or on a desk where you’d rather not have a speaker taking up permanent residence. ECHO doesn’t need to live next to a charging cable or be put away between uses. It sits on the table and becomes part of the room, as unobtrusive as any other ceramic piece nearby.

A guest walking in wouldn’t necessarily clock it as a tech product. That’s partly the point. The glazed surface catches light the way pottery does, and the form is quiet enough to sit beside books or plants without demanding attention. When a phone is slid underneath it, it starts doing its job. When the phone is gone, it just stays there, still looking like it belongs. The same underlying principle runs through the Battery-free Amplifying iSpeakers, where a Duralumin metal enclosure amplifies a smartphone’s audio without any power.

Abe designed ECHO to exist comfortably in a room even when it isn’t doing anything, a goal most speakers never consider. Most audio accessories announce themselves. This one quietly waits, and when a phone is close enough to fill the cavity with sound, the room gets a little warmer and a little fuller without anyone having to reach for a power button.

The post This Ceramic Vase Is Actually a Phone Speaker That Needs No Power first appeared on Yanko Design.

Nocturne’s Free App Turns Your Bricked Spotify Car Thing Into Something Better Than the Original

The open-source community has a long history of doing more with abandoned hardware than the original manufacturers ever did. The PSP got emulators Sony never approved. The Wii got homebrew loaders that ran software Nintendo pretended didn’t exist. The pattern repeats because the hardware is always fine; it’s the corporate support structure around it that evaporates. The Spotify Car Thing joined that lineage in December 2024 when Spotify killed server-side authentication and turned every unit into an expensive knob with a screen attached.

Nocturne picked up where Spotify dropped off. The project launched in October 2024, anticipating the shutdown, and has shipped four major versions since. V4.0.0, currently in beta with a public release imminent, finally delivers true Bluetooth connectivity without phone tethering, a companion app, and a feature set that makes the original Spotify firmware look like a rough draft.

Developer: Nocturne Team (Brandon Saldan, Dominic Frye, and contributors)

The Car Thing runs a 512MB RAM, 4GB storage Amlogic S905D2 SoC, which is a polite way of saying it has the processing power of a mid-range router from 2015. Early versions of Nocturne required a Raspberry Pi as a co-processor just to get the thing online, which was a heroic workaround but not exactly mainstream-friendly. V3 replaced that with Bluetooth tethering through your phone’s hotspot. V4 cuts the tether entirely, pairing directly via Bluetooth through the new Nocturne Companion app, which requires a Nocturne+ subscription to fund the team’s continued development.

What the photos make immediately clear is how clean the UI actually looks in practice. The now-playing screen pulls album art and renders it as a full bleed gradient background, the same visual logic Spotify used but executed with noticeably more polish in edge cases. The typography is large and glanceable. The playlist browser view is dense but organized, using album thumbnails and track titles in a layout that navigates naturally with the knob. Image 3 shows a subtle ambient lighting effect around the screen border, a rainbow glow that responds to the current track, which is the kind of detail you wouldn’t expect from a community firmware project running on 2021 budget hardware.

The gesture control, OTA updates, Spotify Connect device switcher, podcast support, local file playback, and DJ mode all carried over from V3. The V4 architecture also bakes in full offline functionality, meaning the firmware survives without Spotify’s servers being cooperative, which was precisely the failure mode that bricked every original unit in the first place.

Nocturne’s GitHub currently lists V3.0.0 as the stable release, with V4.0.0 accessible to donors via Discord while the team finishes the public build. If you’ve got a Car Thing in a drawer, the installation guide at usenocturne.com is the next tab you should open.

The post Nocturne’s Free App Turns Your Bricked Spotify Car Thing Into Something Better Than the Original first appeared on Yanko Design.

❌