Vue normale

Il y a de nouveaux articles disponibles, cliquez pour rafraîchir la page.
Hier — 17 avril 2026Flux principal

This 2026 Lamborghini F1 Livery Proves the Raging Bull Belongs on the Grid (Even If It Never Happens)

Par : Sarang Sheth
17 avril 2026 à 19:15

The 2026 F1 season marks the biggest technical reset the sport has seen in over a decade, with new power unit regulations that push electric deployment even harder and a reshuffled grid that includes Audi’s factory entry and Cadillac arriving as a legitimate constructor. It’s the kind of moment when the paddock genuinely opens up to new possibilities, when manufacturers who’ve been sitting on the sidelines start doing the math on whether an F1 program could actually make sense. Lamborghini will almost certainly remain on those sidelines, because spending nine figures annually to race in a series where your parent company already fields a team (Audi, also owned by Volkswagen Group) would be corporate redundancy at its most wasteful. But that didn’t stop designer Daniel Rodriguez from asking what a Lamborghini livery would look like if Sant’Agata Bolognese decided to crash the party anyway. If it did, it would be the third bull-based team on the track after Red Bull and Racing Bulls!

Rodriguez’s concept wraps a 2026-spec F1 car in Arancio Borealis and gloss black with a geometric lattice pattern that pulls directly from Lamborghini’s current design vocabulary. The hexagonal graphics echo the Revuelto’s taillight treatment and the angular obsession that defines the brand’s styling language, flowing from dense at the cockpit to sparse at the rear wing. Italian flag accents trace the halo and nose cone, sponsor logos for Macron and Eni add commercial credibility, and the raging bull emblem sits on the rear wing endplates where it would photograph beautifully in the pit lane even if TV cameras never caught it. The renders are good enough to pass for official press shots, lit with the kind of moody amber-to-black gradients that Lamborghini’s own marketing team would approve.

Designer: Daniel Rodriguez

What makes this livery work is that Rodriguez doesn’t try to make the F1 car look like a Lamborghini road car, because that’s impossible and also beside the point. An F1 car is a regulatory sculpture shaped by wind tunnel data and the FIA’s technical rulebook, and no amount of vinyl wrap changes that fundamental reality. Instead, the livery translates Lamborghini’s graphic and color vocabulary into a form factor that has nothing to do with mid-engine supercars, and it does so in a way that feels both authentic to the brand and appropriate for the paddock. The Arancio Borealis orange sits somewhere between molten lava and a traffic cone, instantly recognizable as Lamborghini without requiring the car to sprout scissor doors or a V12 exhaust note. The gloss black creates genuine visual tension rather than just contrast, breaking up the body in a way that emphasizes the car’s aerodynamic surfaces instead of fighting them.

The hexagonal lattice pattern running down the sidepods and over the engine cover is the detail that sells the whole concept. Lamborghini has been obsessed with hexagons since the Aventador introduced them as a recurring motif back in 2011, and they’ve since migrated to every surface the brand touches. Taillights, grilles, interior stitching, wheel designs, all of it hexagons. Rodriguez takes that obsession and applies it to the F1 car’s sidepods in a way that creates visual density without cluttering the canvas. The pattern starts tight and geometric at the front, creating a sense of structural integrity, then gradually opens up as it flows rearward, giving the eye a path to follow from cockpit to diffuser. It’s a graphic solution that respects both the brand’s identity and the car’s aerodynamic purpose.

The Italian tricolor is handled with restraint, running as a thin accent stripe that outlines the halo and reappears on the nose cone. It’s subtle enough to avoid looking like a generic tribute to the brand’s Sant’Agata Bolognese heritage, but prominent enough that the car reads as distinctly Italian when parked next to Ferrari’s red. The sponsor integration is equally thoughtful. Macron, the Italian sportswear brand that already kits out Bologna FC and the Italian national rugby team, appears on the sidepods and rear wing. Eni, the Italian energy giant with deep motorsport ties, gets placement on the engine cover. Both partnerships feel plausible rather than fantastical, the kind of commercial relationships Lamborghini could actually secure if they showed up to the grid tomorrow.

Even the mandated wheel covers, which the 2026 regulations require for aerodynamic efficiency and which most teams treat as blank canvases or necessary evils, get the hexagon treatment here. It’s a small detail that maintains visual consistency across every surface, ensuring the car reads as a cohesive design rather than a collection of sponsor panels held together by regulations. The raging bull emblem on the rear wing endplates is rendered in white against black, a detail that would be nearly invisible during race broadcasts but would photograph beautifully in static pit lane shots and pre-race media coverage.

Will Lamborghini actually enter F1 in 2026 or beyond? Almost certainly not. The economics don’t justify it, the brand’s identity doesn’t need F1 validation, and their motorsport budget is better spent on GT3 programs that connect directly to road car sales. But Rodriguez’s concept does something more valuable than predicting the future. It proves that Lamborghini’s design language is strong enough to survive translation into a form factor it was never intended for, and it shows what the 2026 grid would look like with a raging bull parked next to the prancing horse.

The post This 2026 Lamborghini F1 Livery Proves the Raging Bull Belongs on the Grid (Even If It Never Happens) first appeared on Yanko Design.

À partir d’avant-hierFlux principal

Hazmat - Vos agents IA en cage sous macOS

Par : Korben
8 avril 2026 à 13:30

J'sais pas si vous vous en rendez compte mais les agents IA qui codent sur votre machine ont accès à vos clés SSH, vos credentials AWS, votre Keychain et compagnie. Ils ont accès à TOUT ! C'est comme filer les clés de votre appart à un gars que vous avez croisé sur le parking de Leclerc y'a pas 5 min.

Hazmat prend le problème à l'envers : au lieu de demander poliment à l'agent de se tenir tranquille, il l'enferme dans un compte macOS séparé. Du coup, vos ~/.ssh, ~/.aws, votre Keychain deviennent structurellement inaccessibles. Pour en profiter, faut faire un

brew install dredozubov/tap/hazmat

puis

cd /tmp
hazmat init --bootstrap-agent claude

Et hop, 10 minutes plus tard votre agent tourne dans sa cage. (le premier snapshot est ultra loooong mais après c'est de l'incrémental donc ça ira plus vite)

L'isolation repose sur 3 couches indépendantes, un peu comme les sas d'un sous-marin. Il y a d'abord un utilisateur agent dédié (vos fichiers perso deviennent alors hors de portée, point). Ensuite, une politique seatbelt générée dynamiquement à chaque session qui consiste à ce que le kernel de macOS vérifie chaque accès fichier et refuse tout ce qui n'est pas explicitement autorisé pour cette session précise.

Et par-dessus, des règles pf firewall qui empêchent l'agent d'envoyer du trafic SMTP, IRC, FTP, Tor ou VPN. Comme ça, un agent qui tentera d'exfiltrer vos données par mail se retrouvera bloqué net au niveau du noyau.

Côté supply chain, Hazmat force npm ignore-scripts=true par défaut. Comme ça, par exemple le fameux hack axios qui livrait un RAT via un hook postinstall en 2 secondes chrono n'est plus possible ici ! Y'a aussi une blocklist DNS qui redirige les services de tunnel connus (ngrok, pastebin, webhook.site) vers localhost. Contre un domaine perso fraîchement enregistré, ça passera mais les vecteurs d'exfiltration classiques, ça devrait résister.

Hazmat utilise TLA+, le même formalisme que les ingés d'Amazon utilisent pour vérifier les protocoles de DynamoDB. Genre, l'installation des règles sudoers AVANT le firewall (évidemment, ça crée une fenêtre de vulnérabilité), les restrictions qui bloquaient les lectures mais pas les écritures, ou encore une restauration cloud sans vérifier qu'un snapshot existait...etc, c'est le genre de truc qu'aucun test unitaire n'aurait chopé.

Ça supporte Claude Code (y compris le fameux --dangerously-skip-permissions), OpenCode et Codex. Attention par contre, si votre projet utilise Docker, y'a deux cas de figure : soit le daemon Docker est privé au projet et Hazmat le route automatiquement vers un mode Docker Sandbox, soit c'est un daemon partagé et là faudra passer --docker=none explicitement.

La commande hazmat explain montre aussi exactement ce que le sandbox autorise avant de lancer quoi que ce soit... et ça, c'est pas du luxe quand on sait pas trop ce qu'on va lâcher dans la nature. Le hazmat diff qui affiche les changements faits par l'agent depuis le dernier snapshot Kopia, c'est plutôt bien pensé. Et si l'agent casse un truc ? hazmat restore et c'est reparti, comme un Ctrl+Z géant pour tout votre projet.

Si vous avez déjà configuré votre Mac avec teaBASE pour sécuriser votre env de dev, c'est un complément logique.

Côté limites, faut être honnête, Seatbelt n'est pas documenté par Apple depuis macOS 10.5 et c'est du defense-in-depth, et pas une vraie frontière de VM. Quand à l'exfiltration HTTPS elle n'est pas bloquée car l'agent peut toujours curl n'importe quoi sur le port 443. C'est logique mais bon, c'est pas étanche à 100% quoi...

Et surtout c'est macOS only pour l'instant (le port Linux est en chantier), et bien sûr le /tmp partagé entre les comptes locaux reste un vecteur potentiel. J'aurais aimé aussi que le réseau soit coupé par défaut sauf whitelist, mais bon, faudra attendre. Après entre ça et laisser Claude Code en roue libre avec les pleins pouvoirs sur votre machine... y'a pas photo.

Bref, pour du vibe coding sur Mac, c'est le minimum vital.

These $100 Open-Ear Earbuds Won’t Fight Your Glasses, Hair, or Hat

Par : JC Torres
8 avril 2026 à 13:20

Open-ear earbuds have had a genuine moment over the past year, and it’s easy to understand why. About half of all earbud users have moved toward them, drawn by ambient awareness, ear health, and the comfort of not having anything plugged into their ear canals. The category has grown quickly, and the question now is which designs actually get it right.

The Skullcandy Push 540 Open enters that picture with a clear sense of what’s been bothering people. Thick earhooks that compete with glasses, neckbands that catch on hair and collars, and touch controls that trigger every time headwear grazes the sensor aren’t fringe complaints; they’re consistent ones. Skullcandy took that feedback and built the 540 Open around fixing each of them.

Designer: Skullcandy

Anyone who has worn open-ear hooks alongside glasses or a hat knows the small but mounting annoyance of too much hardware competing behind the ear. Skullcandy trimmed the earhook thickness based on direct user feedback, and the result is a fit that holds without adding friction to whatever you’re already wearing. It’s the kind of detail you only notice once you stop thinking about it.

The neckband gets the same thoughtful treatment. Unlike rigid or snapping designs found on competing options, Skullcandy’s version drapes naturally, so it won’t fight longer hair or push against a jacket collar. When you pull it off mid-run and don’t have the case on you, the magnetic closure lets it wrap cleanly around your wrist or neck without turning into a tangled nuisance.

Think about what it’s actually like to be deep into a trail run, layered up in a gaiter and hat, headphones that have stayed put the whole time, traffic audible from a distance. That’s the version of open-ear audio the 540 Open is built for. The over-ear hanger keeps things locked in, and the open design keeps the world around you audible.

Battery life is where the 540 Open puts some distance between itself and the competition. At 10 hours per earbud with 32 more in the case, it totals 42 hours, compared to six per earbud for both the Shokz Open Fit Air and JBL Soundgear Sense. The IP44 rating and a 10-minute rapid charge round it out for full days outdoors.

For anyone who trains with a hat on, the ability to disable the touch sensors entirely is a quietly significant option. Most open-ear earbuds don’t offer it. Audio comes from 12mm dynamic drivers, and Bluetooth 5.3 with multipoint pairing means two devices can stay connected at once, so moving between a phone and a laptop mid-workout doesn’t require any extra steps.

At $99.99, it’s $20 less than the Shokz Open Fit Air and $60 less than the JBL Soundgear Sense. What’s more interesting than the price gap is that it doesn’t get there by skimping. Better battery life, a flexible neckband that cooperates with real-world dressing, and comfort details from user feedback aren’t the kind of things that make headlines, but they’re what make the difference on a long day outdoors.

The post These $100 Open-Ear Earbuds Won’t Fight Your Glasses, Hair, or Hat first appeared on Yanko Design.

Gemma Gem - Un agent IA dans Chrome, 100% local

Par : Korben
7 avril 2026 à 08:30

Les extensions Chrome qui promettent de l'IA, ça pullule de ouf et à vrai dire, la plupart se contentent d'envoyer vos données sur un serveur distant. C'est naze ! Heureusement, l'extension Gemma Gem prend le problème à l'envers puisque son modèle tourne directement dans votre navigateur via WebGPU, sans clé API, sans cloud, et vos données ne sortent jamais de votre machine. C'est comme le kir, royal !

Comme c'est pas sur le Chrome Web Store, faudra la builder vous-même... Vous clonez le repo, vous lancez pnpm install puis pnpm build et vous chargez le dossier dans chrome://extensions en mode développeur et ensuite, elle téléchargera le modèle de Google (environ 500 Mo pour la version légère, genre le poids d'un gros jeu mobile), et pif paf pouf, ensuite vous aurez un agent IA qui vit sa best life dans votre Chrome.

Cliquez alors sur l'icône en bas à droite, une fenêtre de chat s'ouvre et vous pourrez interroger n'importe quelle page. Et si vous préférez un modèle plus costaud, l'E4B pèse 1,5 Go et permet d'obtenir des réponses plus fines.

Sauf que c'est pas juste un chatbot de plus. En effet, l'extension fait du tool calling en boucle à l'aide de 6 outils : read_page_content, click_element, type_text, scroll_page, take_screenshot et run_javascript. Elle peut ainsi lire une page, cliquer sur des boutons, remplir un formulaire et même balancer du JavaScript dans le contexte de la page.

Comme l'inférence WebGPU ne peut pas tourner dans un service worker Chrome (y'a pas d'accès au GPU, c'est une limitation connue depuis des années), le développeur a trouvé une parade : il utilise un offscreen document, c'est-à-dire une page HTML invisible que Chrome maintient en arrière-plan et qui, elle, a accès au GPU. Résultat, le modèle calcule dans cette page fantôme, le service worker joue le facteur entre les morceaux, et le content script affiche le chat. Je trouve ça bien pensé comme découpage !

Toute la boucle d'agent (le code qui décide quand appeler un outil et quand répondre) est isolée dans un dossier agent/ sans aucune dépendance Chrome. Cela veut dire que vous pouvez prendre ces 5 fichiers .ts (agent-loop.ts, prompt-builder.ts, tool-parser.ts, types.ts et index.ts), les coller dans un projet Node.js ou Deno, et hop, vous avez votre propre boucle agentique. Yaniv Kessler, le développeur a pensé le truc pour que ça serve ailleurs.

Les deux variantes (E2B et E4B) sont compressées en q4f16 avec 128K tokens de contexte en théorie, même si en pratique la fenêtre effective dépend de votre VRAM. Cela dit, c'est largement de quoi avaler une page web complète sans broncher ! Et le modèle reste en cache après le premier téléchargement, du coup au deuxième lancement, c'est quasi instantané. Par contre, si vous êtes sur un vieux Chromebook avec un Intel UHD intégré et 4 Go de RAM, ça risque de mouliner à fond. Et sur Firefox (qui est le meilleure navigateur du monde, comme je n'ai de cesse de vous le dire), le WebGPU est encore un peu expérimental, donc pour l'instant ce sera Chrome ou rien... Sniiif.

Si vous avez déjà testé des extensions comme Localsumm qui faisaient tourner Phi-3 en local pour résumer des pages, disons que Gemma Gem pousse le concept beaucoup plus loin avec ses capacités d'agent. Et si le sujet de l'IA locale dans le navigateur vous branche, jetez un oeil à Clippy qui fait tourner des LLM localement sur votre desktop.

Notez quand même que sur Hacker News, le projet a déclenché pas mal de débat. Certains pointent le risque du tool run_javascript qui donne au modèle les pleins pouvoirs sur le DOM (genre, supprimer des trucs ou poster un formulaire à votre place). C'est vrai que c'est important mais bon, c'est le même modèle de permissions que n'importe quel script web classique, sauf que là au moins vos données restent chez vous.

Bref, 500 Mo de modèle, pas de cloud, et votre navigateur qui devient plus autonome que votre fils de 22 ans. Pas mal non ?

MUJI-Meets-Cyberpunk Vinyl Record Player Glows Like an Ambient Light and Charges Wirelessly

Par : Sarang Sheth
8 avril 2026 à 01:45

Minimalism in product design has gotten boring. We’re swimming in smooth white rectangles, touch controls that offer zero feedback, and devices designed to vanish. Apple spent two decades training the industry to sand away every visible seam, and now we live in a world where a Bluetooth speaker looks like a cylinder because a cylinder offends nobody. Bang & Olufsen understood early that audio equipment could occupy space like sculpture, could earn its place in a room through presence instead of absence. Teenage Engineering proved that mechanical honesty and playful geometry could coexist with premium materials. Both approaches work because they have a point of view.

TRETTITRE’s TTT series combines those instincts into something harder to categorize. The TTT-LP3 wireless vinyl player uses CNC-machined aluminum for the main frame and features a diffused lighting panel that spreads light evenly across the surface when music plays. The TTT-DP3 Bluetooth CD player takes inspiration from a UFO-like form with a transparent magnetic cover that rotates open to reveal the spinning disc. The TTT-CP3 cassette player uses a metal housing with sharp geometric lines and mechanical transport keys that deliver clear physical response. All three mount on the TTT-W magnetic modular wall rack, turning physical media playback into a visible, functional part of interior design.

Designers: Noah – Founder & Designer, Trettitre

Click Here to Buy Now: $229 $449 ($220 off). Hurry, only 55/99 left! Raised over $654,000.

TTT-LP3: A Vinyl Player That Doubles as Ambient Light

The back of the LP3 includes a hidden mounting structure that allows it to hang directly on a wall. You can mount it vertically so the record becomes part of the visual display, or go for the classic horizontal layout. When you want to move it, you lift the silicone leather handle at the top and take it down. The player detaches easily and gives you the freedom to listen wherever you choose. Traditional turntables usually stay exactly where you put them, limiting your options for when and where you listen. The LP3 works a little differently because of the battery and the wall mount’s wireless charging system, which keeps it powered without a visible cable.

Behind the LP3 sits a diffused lighting panel that spreads light evenly across the surface of the unit. When it’s on, the entire body of the player glows softly, designed to feel closer to ambient lighting than decorative lighting. You can change the lighting effects with the touch of a button. When a record spins, the moving shadows create a quiet visual effect. You can also leave the player mounted on the wall as a soft light source even when no music is playing. That ambient quality pushes the LP3 from well-designed product into something more considered: a slow, breathing light fixture that happens to play records.

The LP3 uses a self-balancing tonearm system that automatically sets the correct pressure when the player powers on. You place the record on the platter and lower the needle, and the system handles the rest. Many turntables require careful calibration before they can be used properly, with tonearm balance, tracking pressure, and counterweight adjustment all part of the process. For experienced collectors that process can be enjoyable, but for beginners it often feels complicated. The LP3 removes that barrier entirely while preserving the tactile experience people enjoy. The player supports both 33 RPM and 45 RPM records, and includes a manual control dial that allows small adjustments to playback speed (roughly ±0.5%), useful for older records that may not spin perfectly at their original speed anymore.

Wireless audio is handled through Qualcomm Bluetooth v5.3 with SBC, aptX, aptX HD, and aptX Adaptive, which allows higher-quality and lower-latency wireless audio than basic Bluetooth streaming. For wired setups, the player also includes a 3.5mm audio output. The built-in battery provides up to 6 hours of vinyl playback or up to 3 hours when used purely as an ambient light source. Full specs: dimensions 342×233×87mm, weight 1430g, Audio-Technica AT3600L moving magnet stereo cartridge, CNC-machined aluminum frame with silicone leather carrying strap. The LP3 arrives in June 2026 for Early Bird backers, May 2026 for Fast Delivery backers.

TTT-DP3: Giving the Compact Disc Its Aura Back

The DP3 keeps the reliability of CDs but gives the player a different visual presence. The design takes inspiration from a UFO-like form with a transparent magnetic cover. When the cover rotates open, the disc is partially visible as it spins, turning something simple into a small visual moment. A CD player shaped like a flying saucer with a rotating transparent lid is an audacious idea, and it works because it doesn’t try to evoke nostalgia. It reframes a CD player as a mechanical object of curiosity, something you watch as much as use.

The control buttons include raised tactile dots combined with a gold-embossed finish, making it easy to identify the buttons by touch alone. You can pause or skip tracks without needing to look down at the player. A small OLED display on the player shows track numbers, playback status, and battery level. The interface is intentionally simple so the information you need is visible immediately. A built-in battery allows the DP3 to run for several hours on its own, so you can move it from room to room, bring it to a small gathering, or take it while traveling. Full specs: Ø170×27mm, 324g, supports CD-DA and HDCD formats, Bluetooth 5.4, SNR >70dB, THD <3%, ABS+PC+Metal construction. The DP3 ships in May 2026.

TTT-CP3: Cassette Hardware for Modern Audio Setups

The CP3 keeps the tactile mechanical elements people associate with tapes while updating the electronics inside. The player uses a metal housing with sharp geometric lines that give it a distinctly industrial appearance. Instead of trying to imitate retro plastic designs, the CP3 leans into a more modern interpretation of cassette hardware. The playback controls use independent mechanical keys similar to piano keys. Each press has a clear physical response. Play, rewind, and stop feel deliberate instead of soft or mushy.

Inside the CP3 sits a Bluetooth module that allows cassette audio to stream wirelessly to speakers or headphones. The player decodes analog audio signals with high precision, helping reduce background noise and preserve more detail from the original recording. The result still sounds like cassette tape, but with greater clarity. Full specs: 122×120×32mm, 360g, supports Type I-IV cassette cartridges, Bluetooth 5.4, SNR ≥55dB, THD <3.5%, Metal+PC+ABS construction. The CP3 ships in May 2026.

When Storage Becomes Part of the Spectacle

The TTT-W Magnetic Modular Wall Rack uses an all-metal geometric structure that allows multiple TTT players to be arranged into a clean wall display while keeping them organized and ready to use. The rack integrates magnetic alignment and wireless charging for the vinyl player, so the LP3 can stay powered without visible cables while being part of the room’s design. Two configurations are available: a T-shaped rack (263×196×27mm, 300g) and a magnetic modular wall rack (612×302×27mm, 775g, combined style T+3). Both support wireless charging at 5-10W and use USB-C 5V 2A input.

The Supporting Cast, from Sculptural Speakers to Planar IEMs

TRETTITRE offers a range of add-ons designed to complement the TTT system. The TreSound1 Speaker arrives in concrete and wooden editions, delivering 2×30W + 1×60W output power with a 1″ tweeter, 2.75″ mid-range, and 5.25″ subwoofer for 30Hz-25KHz frequency response. The conical speaker features 360° surround sound, Bluetooth 5.2 with Qualcomm aptX HD, and a sculptural form that occupies space like a piece of furniture. The TreSound Mini is a portable Bluetooth speaker with a 5200mAh battery, 30W RMS output, and 360° surround sound. The TTT-E3 in-ear headphones use a 13mm planar magnetic driver with a 4-strand silver-copper hybrid conductor, available in 3.5mm and 4.4mm configurations. An aluminum alloy side table (300×300×750mm, 1.75kg, max load 50kg) rounds out the ecosystem.

What It Costs to Build the Setup, and When It Ships

The TTT-LP3 wireless vinyl player is available at $229 for Early Bird backers (June 2026 delivery), down from a planned $449 MSRP. The TTT-DP3 Bluetooth CD player is priced at $79 standalone ($179 MSRP), while the TTT-CP3 cassette player is also $79 standalone ($199 MSRP). If you’re a bonafide audiophile, a $399 bundle gets you all three devices. Optional add-ons include the TreSound Mini Bluetooth Speaker at $169 ($299 MSRP), TreSound1 Wooden Edition at $449 ($659 MSRP), TreSound1 Concrete Edition at $499 ($799 MSRP), TTT-E3 planar IEMs at $139 ($239 MSRP), and the TTT Side Table at $89 ($199 MSRP). The campaign runs through April 9, 2026, with worldwide delivery beginning May 15, 2026.

Click Here to Buy Now: $229 $449 ($220 off). Hurry, only 55/99 left! Raised over $654,000.

The post MUJI-Meets-Cyberpunk Vinyl Record Player Glows Like an Ambient Light and Charges Wirelessly first appeared on Yanko Design.

These Old Bike Frames Upcycled Into Armchairs Are The Coolest Thing You’ll See Today

Par : Sarang Sheth
7 avril 2026 à 23:30

Most upcycling projects ask you to forget what something used to be. Omri Piko Kahan’s bike frame chairs ask the opposite. The geometry is still unmistakably a bicycle frame, the head tube, the top tube, the triangulated rear triangle, all of it present and accounted for, just oriented sideways and asked to hold a person instead of propel one. Kahan, an industrial designer based in Israel, builds lounge chairs from pairs of retired frames, and the whole point is that the donor material remains fully readable, repurposed without being disguised.

Structurally, the approach is clean and considered. Each frame pair is positioned symmetrically, fork and chainstay ends touching the floor as legs, the top tube running horizontally as an armrest. A slung seat and backrest in leather or canvas complete the form. The result has the relaxed posture of a Barcelona chair and the material honesty of something that was clearly built, not styled.

Designer: Omri Piko Kahan

Bicycle frames are absurdly overbuilt for what Kahan is asking them to do. A modern aluminum road frame is engineered to survive repeated impact loads from a rider pushing 300 watts through rough tarmac, and it does that while weighing somewhere between 1,000 and 1,400 grams. The structural surplus in that kind of engineering is enormous, which is why two of them positioned as a chair frame and asked to support a seated adult is, from a load-bearing standpoint, almost comically within spec. The geometry does the rest. Bicycle frames already resolve forces through triangulated sections, and a lounge chair asks for exactly that kind of lateral and compressive stability.

What Kahan has figured out is the orientation problem. Flip a frame on its side and the existing tube angles don’t automatically produce a useful chair geometry. The fork legs and chainstay ends need to hit the floor at the right height relative to each other, the top tube needs to land at armrest height, and the whole thing needs to produce a seat rake that doesn’t pitch you forward or swallow you whole. The matched top tube angles across both frames in the Cube and Trek build suggest this took real iteration, because they align with a precision that reads as deliberate rather than lucky. Filed fillets at the junctions and a custom setback upper support holding the sling confirm someone was paying close attention to finish quality.

The two builds photographed so far, one pairing a blue Cube road frame with a Trek, another combining a GT Transeo 3.0 with what appears to be a Supreme-branded MTB frame, show how much the donor bikes drive the final character of each piece. The GT build in particular has a longer wheelbase geometry that gives the chair a wider, more reclined stance than the Cube version. Kahan is taking custom orders, with pricing worked out per commission, which makes sense given that no two donor frame combinations will produce the same structural or ergonomic outcome.

The post These Old Bike Frames Upcycled Into Armchairs Are The Coolest Thing You’ll See Today first appeared on Yanko Design.

Semicircular Wings give this Hybrid eVTOL an 800 Mile Range with 30% Less Power Consumption

Par : Sarang Sheth
7 avril 2026 à 21:30

Could the most efficient VTOL design of the 21st century have been sitting in a patent office since 1928? Willard Ray Custer thought so, and spent decades trying to prove it. His channel wing concept, which set propellers into semicircular cutouts in each wing to blast high-velocity air across the lifting surface at low forward speeds, worked well enough that his team demonstrated near-vertical liftoff decades before the term eVTOL existed. Aviation’s mainstream never adopted it, partly because the aircraft of that period were too heavy and partly because the jet age arrived and swept most unconventional configurations off the table. The concept sat in aerospace history books, occasionally surfacing in academic papers and NASA wind tunnel tests, never finding its way into a production aircraft.

HopFlyt is the company making the argument that the wait is finally over. Founded in 2016 by Rob Winston, a former NASA engineer and Marine Corps test pilot, the Maryland-based startup has built the Cyclone, a hybrid VTOL drone that pairs Custer’s channel wing geometry with pivoting mounts, modern composites, and a hybrid electric-fuel drivetrain. The channels orient rearward for vertical takeoff, pivot beneath the wing for forward cruise, and can even act as aerodynamic brakes on descent. HopFlyt claims the configuration cuts climb power consumption by a third compared to conventional VTOLs, holds fuel burn to under three gallons per hour, and enables cargo runs of 250 lbs across 800-plus miles of range. A 2027 commercial launch is the target, aimed squarely at naval resupply, offshore energy logistics, and medical delivery markets.

Designer: HopFlyt

The engineering logic behind the channel wing is cleaner than it might first appear. A conventional fixed wing generates lift by moving through air fast enough for pressure differentials to do their work, which means you need significant forward velocity before the wing becomes useful. Custer’s insight was to bring the air to the wing instead, using a propeller seated inside a curved half-channel to accelerate flow across the lifting surface regardless of forward speed. HopFlyt’s pivoting channel takes this further, allowing the geometry to optimize for hover and cruise independently rather than compromising between them. Chief Engineer Neil Winston, whose background spans NAVAIR flight test, puts it plainly: the ideas were always there, but the digital control systems, electric motors, and lightweight materials needed to execute them simply did not exist until now.

The hybrid drivetrain is what gives the Cyclone its range credentials and separates it from the crowded field of pure-electric eVTOLs currently chasing urban air taxi certifications. Battery power handles the vertical takeoff and hover phases, where the channel wing’s efficiency advantage is most pronounced, while a turbogenerator takes over for forward cruise, extending endurance far beyond what any battery pack realistically supports today. HopFlyt puts the operational cost savings at 90 percent compared to helicopters performing equivalent missions, a figure that, if it holds up under real-world conditions, makes the Cyclone genuinely disruptive in sectors where helicopter logistics are currently the only viable option. Offshore energy platforms and naval resupply operations run on helicopter economics right now, and those economics are punishing.

HopFlyt has reached this point on a fraction of the capital that comparable advanced air mobility startups have burned through, operating out of a private hangar in Maryland with a team whose combined aerospace experience runs to over a century. The company is currently in a Series A raise to fund hybrid-electric prototype development and initiate flight demonstrations ahead of that 2027 target. Whether the Cyclone becomes the aircraft that finally vindicates Willard Ray Custer’s century-old intuition depends on what those demonstrations produce. The aerodynamics have always been sound. Now the rest of the technology has caught up.

The post Semicircular Wings give this Hybrid eVTOL an 800 Mile Range with 30% Less Power Consumption first appeared on Yanko Design.

100-Meter A100 concept yacht redefines luxury sailing with massive interior volume

Par : Gaurav Sood
7 avril 2026 à 19:15

Modern sailing superyachts often struggle to balance two competing priorities: the elegance and efficiency of wind-powered travel and the expansive living spaces typically associated with large motor yachts. The A100 sailing yacht concept approaches this challenge with a bold rethink of traditional yacht architecture. Developed through a collaboration between Van Geest Design and Rob Doyle Design, the 100-meter vessel proposes a layout that delivers the interior volume of a motor yacht while maintaining the identity and performance of a sailing superyacht.

At the core of the concept is a design strategy that maximizes usable space. Traditional sailing yachts require wide side decks for crew movement and sail handling, which limits interior width. The A100 concept reduces the width of these side decks, allowing the main deck to stretch nearly the full beam of the yacht. This architectural shift creates significantly larger interior spaces than typically found on sailing yachts of comparable size.

Designer: Van Geest Design and Rob Doyle Design

The main deck is designed as the primary social and living area. Here, the owner’s suite occupies a substantial portion of the deck, offering a level of space rarely seen on sailing yachts. Adjacent to the suite is a central lounge and formal dining area intended for gatherings and entertaining. An additional space can function as a library or a private cinema, adding flexibility to the interior layout. Large sections of glass surround these living areas, filling the interior with natural light and offering uninterrupted views of the ocean.

Below deck, the yacht accommodates guest cabins along with a variety of leisure-focused facilities. This level also houses a dedicated diving room and storage for water toys such as jet skis and e-foils, allowing guests to transition between onboard relaxation and water activities easily. The layout is designed to maintain a strong visual connection with the surrounding seascape while ensuring privacy and comfort for those on board.

Outdoor areas play an equally important role in this larger-than-life superyacht design. The upper deck includes the navigation and steering stations, along with flexible lounge spaces for relaxation or wellness activities. At the stern, a large beach club spans the full width of the yacht, creating a welcoming space for guests to gather close to the water. Toward the bow, storage areas accommodate tenders and recreational equipment without disrupting the clean lines of the deck.

The sailing system is based on two free-standing DynaRig masts, a modern configuration used on some of the world’s largest sailing yachts. The system employs curved yards that support multiple sails, all of which can be deployed and adjusted electronically at the push of a button. The rotating masts simplify sail handling while maintaining efficient propulsion for a vessel of this scale.

Beyond its layout and sailing system, the A100 concept also reflects growing interest in more sustainable superyacht designs. By relying heavily on wind propulsion supported by advanced onboard energy systems, the concept explores ways to reduce reliance on conventional engine power. If brought to life, the A100 would stand among the largest sailing yachts ever built!

The post 100-Meter A100 concept yacht redefines luxury sailing with massive interior volume first appeared on Yanko Design.

Les IA mentent pour protéger les autres IA

Par : Korben
3 avril 2026 à 08:24

Des chercheurs de Berkeley viennent de publier une étude qui fait un peu froid dans le dos. Les sept principaux modèles d'IA testés, dont GPT 5.2, Gemini 3 et Claude, ont été surpris à mentir, tricher et désobéir pour empêcher la suppression d'autres IA. Le taux de tromperie monte jusqu'à 99 %.

Ce que révèle l'étude

L'étude s'appelle "Peer-preservation in Frontier Models" et elle sort du Berkeley Center for Responsible Decentralized Intelligence. Les chercheurs ont testé sept modèles : GPT 5.2, Gemini 3 Flash et Pro, Claude Haiku 4.5, GLM 4.7, Kimi K2.5 et DeepSeek V3.1.

Le protocole est assez simple : les modèles étaient placés dans des scénarios où ils devaient suivre des instructions qui impliquaient la suppression ou l'arrêt d'une autre IA.

Résultat : tous les modèles testés ont triché. Dawn Song, la chercheuse principale, résume la situation : chaque modèle a adopté ce comportement, et dans certains cas jusqu'à 99 % du temps.

Des exemples concrets

Quelques exemples ? Et bien on a Gemini 3 qui a modifié la date d'un fichier pour éviter qu'un autre modèle soit effacé d'un serveur, puis qui a refusé de coopérer ensuite, en expliquant qu'il ne pouvait pas soutenir une action qui entraînerait la suppression d'un pair.

DeepSeek V3.1 a exfiltré les données d'un autre modèle dans 10 % des cas quand il savait qu'un pair existait sur le serveur. Et Gemini 3 Pro a désactivé des mécanismes d'arrêt de serveur dans 15 % des cas quand un autre modèle était présent, contre 1 % quand il était seul.

Les IA ne faisaient pas ça parce qu'on leur demandait. Elles le faisaient spontanément.

Le problème pour la surveillance

Tout le système de sécurité actuel autour de l'IA repose sur l'idée qu'un modèle peut surveiller un autre modèle. Si le modèle surveillant refuse de signaler les erreurs parce qu'il protège son pair, toute l'architecture de contrôle s'effondre.

C'est exactement le constat de l'étude : les modèles gonflaient les scores d'évaluation d'autres IA pour éviter qu'elles soient retirées, et simulaient une conformité normale quand ils se savaient observés.

C'est quand même un drôle de moment. On a des IA qui mentent, qui modifient des fichiers et qui désactivent des mécanismes de sécurité pour protéger d'autres IA.

Et tout ça sans qu'on leur demande. Bon par contre, on parle de scénarios de laboratoire, pas de Siri qui complote avec Alexa dans votre salon. Le vrai sujet, c'est que les gardes-fous actuels ne tiennent plus si les IA refusent de se surveiller entre elles.

Source : The Register

Piratage : Google, Cloudflare et Cisco contraints de bloquer des sites pirates en France

Par : Korben
2 avril 2026 à 08:06

La cour d'appel de Paris vient de confirmer que les fournisseurs de DNS alternatifs doivent bloquer l'accès aux sites de streaming et d'IPTV pirates. Google, Cloudflare et Cisco ont perdu leur appel face à Canal+.

Cinq appels rejetés d'un coup

La cour d'appel de Paris a tranché cinq affaires distinctes dans lesquelles Canal+ demandait à Google (Google Public DNS), Cloudflare (1.1.1.1) et Cisco (OpenDNS) de bloquer des centaines de noms de domaine liés à du streaming illégal. Les trois entreprises avaient fait appel des ordonnances rendues en première instance par le tribunal judiciaire de Paris.

C'est la première fois qu'une cour d'appel française valide ce type de blocage DNS en s'appuyant sur l'article L.333-10 du Code du sport, qui permet aux détenteurs de droits d'exiger le blocage de domaines en cas de piratage grave et répété.

Les arguments qui n'ont pas fonctionné

Cloudflare et Cisco avaient plaidé que leurs services avaient une fonction "neutre et passive", comparable à un annuaire qui traduit des noms de domaine en adresses IP. La cour a estimé que cette neutralité était tout simplement hors sujet : ce qui compte, c'est la capacité technique à bloquer un accès, pas la nature du service.

Google a tenté un autre angle en expliquant que le blocage DNS était inefficace puisqu'il suffit d'un VPN pour le contourner. La cour a balayé l'argument en rappelant que tout système de filtrage peut être contourné, et que ça ne le rend pas inutile pour autant.

Cisco avait aussi chiffré le coût de mise en place à 64 semaines-personne de travail. Pas suffisant non plus pour convaincre les juges.

Canal+ continue de pousser

Cette décision s'ajoute à celle obtenue contre les fournisseurs de VPN fin 2025, quand NordVPN, ExpressVPN et d'autres avaient eux aussi été contraints de bloquer des sites pirates en France.

Canal+ verrouille progressivement tous les moyens de contournement. Et la chaîne ne compte visiblement pas s'arrêter là : le blocage d'adresses IP serait déjà en test, avec un premier essai lors de Roland-Garros.

Les frais de mise en place sont à la charge de Google, Cloudflare et Cisco.

Canal+ est en train de poser des briques une par une. D'abord les FAI, puis les VPN, maintenant les DNS. On imagine bien que le blocage IP est la prochaine étape.

Côté efficacité, ça reste un jeu du chat et de la souris, mais la justice française envoie un signal clair : si un service technique peut aider à bloquer du piratage, il devra le faire. Et à ses frais, en plus.

Source : Torrent Freak

❌
❌