OpenRouter présente une analyse approfondie des interactions avec les modèles de langage à grande échelle (LLM) basée sur plus de 100 000 milliards de tokens collectés dans des conditions réelles. Cette étude empirique examine les patterns d’utilisation à travers différentes tâches, zones géographiques et périodes temporelles, offrant des insights précieux sur l’adoption et l’usage des IA génératives par les développeur·se·s et utilisateur·rice·s.
Google is rolling out a new virtual try-on feature powered by Nano Banana and Gemini 2.5 Flash image models that can turn a simple selfie into a full-body avatar for trying on clothing.
Modern bionic hand prostheses nearly match their natural counterparts when it comes to dexterity, degrees of freedom, and capability. And many amputees who tried advanced bionic hands apparently didn’t like them. “Up to 50 percent of people with upper limb amputation abandon these prostheses, never to use them again,” says Jake George, an electrical and computer engineer at the University of Utah.
The main issue with bionic hands that drives users away from them, George explains, is that they’re difficult to control. “Our goal was making such bionic arms more intuitive, so that users could go about their tasks without having to think about it,” George says. To make this happen, his team came up with an AI bionic hand co-pilot.
Micro-management issues
Bionic hands’ control problems stem largely from their lack of autonomy. Grasping a paper cup without crushing it or catching a ball mid-flight appear so effortless because our natural movements rely on an elaborate system of reflexes and feedback loops. When an object you hold begins to slip, tiny mechanoreceptors in your fingertips send signals to the nervous system that make the hand tighten its grip. This all happens within 60 to 80 milliseconds—before you even consciously notice. This reflex is just one of many ways your brain automatically assists you in dexterity-based tasks.
Typically, if you happened across a walnut lying about, you might consider eating it or throwing it to a friendly squirrel. However, as [Penguin DIY] demonstrates, it’s perfectly possible to turn the humble nut into a clandestine surveillance device. It turns out the walnut worriers were right all along.
The build starts by splitting and hollowing out the walnut. From there, small holes are machined into the mating faces of the walnut, into which [Penguin DIY] glues small neodymium magnets. These allow the walnut to be opened and snapped shut as desired, while remaining indistinguishable from a regular walnut at a distance.
The walnut shell is loaded with nine tiny lithium-polymer cells, for a total of 270 mAh of battery capacity at 3.7 volts. Charging the cells is achieved via a deadbugged TP4056 charge module to save space, with power supplied via a USB C port. Holes are machined in the walnut shell for the USB C port as well as the camera lens, though one imagines the former could have been hidden purely inside for a stealthier look. The camera itself appears to be an all-in-one module with a transmitter built in, with the antenna installed in the top half of the walnut shell and connected via pogo pins. The video signal can be picked up at a distance via a receiver hooked up to a smart phone. No word on longevity, but the included batteries would probably provide an hour or two of transmission over short ranges if you’re lucky.
If you have a walnut tree in your backyard, please do not email us about your conspiracy theories that they are watching you. We get those more than you might think, and they are always upsetting to read. If, however, you’re interested in surveillance devices, we’ve featured projects built for detecting them before with varying levels of success. Video after the break.
OpenAI has long published research on the potential safety and economic impact of its own technology.
Now, Wired reports that the Sam Altman-led company is becoming more “guarded” about publishing research that paints an inconvenient truth: that AI could be bad for the economy.
The perceived censorship has become such a point of frustration that at least two OpenAI employees working on its economic research team have quit the company, according to four Wired sources.
One of these employees was economics researcher Tom Cunningham. In his final parting message shared internally, he wrote that the economic research team was veering away from doing real research and instead acting like its employer’s propaganda arm.
Shortly after Cunningham’s departure, OpenAI’s chief strategy officer Jason Kwon sent a memo saying the company should “build solutions,” not just publish research on “hard subjects.”
“My POV on hard subjects is not that we shouldn’t talk about them,” Kwon wrote on Slack. “Rather, because we are not just a research institution, but also an actor in the world (the leading actor in fact) that puts the subject of inquiry (AI) into the world, we are expected to take agency for the outcomes.”
The reported censorship, or at least hostility towards pursuing work that paints AI in an unflattering light, is emblematic of OpenAI’s shift away from its non-profit and ostensibly altruist roots as it transforms instead into a global economic juggernaut.
When OpenAI was founded in 2016, it championed open-source AI and research. Today its models are close-sourced, and the company has restructured itself into a for-profit, public benefit corporation. Exactly when is unclear, but reports also suggest that the private entity is planning to go public at a $1 trillion valuation, anticipated to be one of the largest initial public offerings of all time.
Though its non-profit arm remains nominally in control, OpenAI has garnered billions of dollars in investment, has signed deals that could bring in hundreds of billions of more, while also entering contracts to spend just as dizzying amounts of money. OpenAI gets AI chipmaker to agree to invest up to $100 billion in it on one end, and says it will pay Microsoft up to $250 billion for its Azure cloud services on the other.
With that sort of money hanging in the balance, it has billions of reasons why it wouldn’t want to release findings that shake the public’s already wavering belief in its tech — as many fear its potential to destroy or replace jobs, not to mention talk of an AI bubble or existential risks to humankind from the tech.
OpenAI’s economic research is currently overseen by Aaron Chatterji, According to Wired, Chatterji led a report released in September which showed how people around the world used ChatGPT, framing it as proof of how it created economic value by increasing productivity. If that seems suspiciously glowing, an economist who previously worked with OpenAI and chose to remain anonymous alleged to Wired that it was increasingly publishing work that glorifies its own tech.
Cunningham isn’t the only employee to leave the company over ethical concerns of its direction. William Saunders, a former member of OpenAI’s now-defunct “Superalignment” team, said he quit after realizing it was “prioritizing getting out newer, shinier products” over user safety. After departing last year, former safety researcher Steven Adler has repeatedly criticized OpenAI for its risky approach to AI development, highlighting how ChatGPT appeared to be driving its users into mental crises and delusional spirals. Wired noted that OpenAI’s former head of policy research Miles Brundage complained after leaving last year that it became “hard” to publish research “on all the topics that are important to me.”
The gaming industry witnesses revolutionary transformation as blockchain infrastructure enables sophisticated virtual economies supporting millions of concurrent transactions.
High speed networks capable of processing thousands of operations per second provide the technological foundation for sustainable in-game marketplaces, seamless player-to-player trading, and complex economic systems rivaling real-world markets.
Understanding performance indicators including sol price, currently trading between $130-140 in December 2025, becomes essential for developers evaluating network capabilities when building virtual worlds requiring instant transaction finality and minimal operational costs.
Network Performance Requirements for Gaming Applications
Game developers require blockchain infrastructure delivering transaction speeds comparable to traditional centralized gaming servers while maintaining decentralization benefits. Solana emerged as the leading blockchain for gaming applications by processing over 65,000 transactions per second with confirmation times under one second. This performance capability proves critical when thousands of players simultaneously execute in-game purchases, trade digital assets, or participate in marketplace activities requiring instant confirmation.
Traditional blockchains struggle with gaming demands, as Bitcoin requires approximately 10 minutes for transaction confirmation while Ethereum averages 12-15 seconds. These delays create unacceptable user experiences in fast paced gaming environments where players expect instant feedback comparable to Web2 applications.
Solana's Proof-of-History mechanism combined with parallel transaction processing through Sealevel enables the network to handle gaming workloads without compromising decentralization or security.
Player Ownership and NFT Integration
Blockchain technology fundamentally transforms player relationships with digital assets by establishing verifiable ownership through non-fungible tokens. Unlike traditional games where developers retain ultimate control over in-game items, NFT-based assets provide players true ownership recorded on immutable public ledgers.
Players can freely trade weapons, character skins, virtual land, and other game elements on open marketplaces, monetising their gameplay achievements outside publisher-controlled ecosystems.
Star Atlas exemplifies sophisticated NFT integration within gaming, allowing players to own spaceships, weapons, and territory as tradable blockchain assets. The game combines strategy, role-playing, and simulation mechanics with play to earn models where players earn cryptocurrency by completing missions and conquering new areas.
Mini Royale: Nations demonstrates NFT interoperability by integrating Ready Player Me avatars, enabling players to use characters across different platforms while maintaining ownership verification through blockchain records.
Decentralised In-Game Economics
Blockchain enables developers to construct player driven economies operating independently from centralized publisher control. Smart contracts facilitate secure peer-to-peer transactions where players directly trade assets without intermediary involvement, reducing friction and enabling instant settlements.
These decentralised marketplaces operate 24/7 with transparent pricing mechanisms visible to all participants, establishing fair market values through organic supply and demand dynamics.
Gaming tokens serve as native currencies within virtual worlds while maintaining convertibility to other cryptocurrencies and fiat money through external exchanges. This dual functionality allows players to monetize gaming time by earning tokens through gameplay achievements, then exchanging accumulated wealth for real-world value.
The blockchain and NFTs in gaming market projects expansion to $2,721 billion by 2035, demonstrating institutional recognition of sustainable economic models emerging within virtual environments.
Transaction Cost Considerations
Network fees significantly impact gaming economic viability, particularly for high-frequency microtransactions characteristic of modern multiplayer games. When evaluating sol price movements and network economics, developers recognise that Solana's average transaction cost remains below $0.001, enabling profitable business models even with numerous small value transactions.
This cost efficiency contrasts sharply with networks charging several dollars per transaction, which prove prohibitive for gaming applications requiring thousands of daily player interactions.
Affordable transaction fees make Solana ideal for play to earn models, NFT minting, and in-game marketplace operations without imposing excessive overhead on players or developers. Games like STEPN, Aurory, and Honeyland leverage Solana's low cost infrastructure to create engaging experiences where players earn rewards without transaction fees consuming their earnings.
The economic sustainability of these models depends directly on network efficiency, positioning high-performance blockchains as essential infrastructure for next-generation gaming.
Cross-Game Asset Interoperability
Blockchain technology facilitates asset portability across different gaming platforms, creating interconnected virtual ecosystems. NFT standards enable developers to recognise and integrate assets from other games, allowing players to utilise characters, items, or currencies across multiple titles.
This interoperability enhances asset value by expanding utility beyond single game contexts while building cohesive metaverse environments where digital identities and possessions persist across virtual worlds.
Echoes of Eldoria implements sophisticated NFT systems where each hero possesses unique attributes and battle histories permanently recorded on-chain. Players can trade these characters in dedicated marketplaces while maintaining complete ownership history and achievement records.
The platform's NFT staking, character fusion mechanics, and limited edition releases demonstrate how blockchain infrastructure supports complex game mechanics impossible within traditional centralised architectures.
Comment pouvons-nous interagir avec les ordinateurs en utilisant uniquement notre esprit ? Les interfaces cerveau-ordinateur existantes sont souvent limitées par leur taille et leur invasivité,...
Csikszentmihalyi drew this finding from over 100 interviews with exceptional people—engineers, writers, historians, chemists, musicians, business people and more—and his lifetime of research into psychology, happiness, and flow. Alongside dismantling several myths surrounding creativity, such as the tortured genius, Mihalyi explores the central idea that creative people do not lean only toward one side or the other on a range of personality traits. Instead, creative people will display quite opposite characteristics at different times, or even at the same time.
The 10 Contradictory Traits of Creative People
Creative people often display both sides of these traits:
Energetic and Restful
Smart and Naive
Playful and Disciplined
Imaginative and Realistic
Extroverted and Introverted
Humble and Proud
Feminine and Masculine
Traditional and Rebellious
Passionate and Objective
Sensitive and Joyful
The Importance of Contradiction
This idea of contradictory personality traits was exciting to me when I read about it. I was a young design and engineering graduate learning and applying the Theory of Inventive Problem-Solving (TRIZ) for what seemed like creative solutions. I was fascinated with where ideas come from, what creative people do differently, and how I could become one.
A fundamental idea in TRIZ is that solutions evolve through the resolution of contradictions. For example, you want a tent to be strong and sturdy so it doesn’t blow over or break. The standard approach for this makes it heavier. But you also want it to be lightweight to carry. How can it be both strong and lightweight? Enter creative solutions.
Compromise and trade-off are sometimes necessary, but whenever you find what seems a truly creative solution, you’ll usually find that someone’s resolved a contradiction at its core.
And here was Mihaly Csikszentmihalyi finding that creative people benefit not just from being imaginative, but also, at times, from being realistic. Not just smart, but also, at times, naive. Resolving a contradiction of personality.
Resisting Labelling
Labelling others or yourself usually seems to me a misguided and often harmful act. For example, attribution bias is the tendency to attribute behaviour to fixed personality traits rather than to circumstances or actions. Considering someone a genius, a loser, or a natural artist, without considering the circumstances and actions that got them where they are, gets us off the hook.
As David Macauley pointed out (video), when people say they can’t draw, they’ve usually never really tried.
The idea of displaying contradictory or paradoxical traits was very freeing to me. To be creative, you don’t have to be this or that. You can be this at times and that at times. And perhaps this fluidity helps us towards creative insights.
I generally consider myself fairly introverted—I easily recharge through time by myself—but in the right company and the right time, I can appear quite extroverted.
The smartest people can also ask naive questions.
Being proud and confident can help try things that others won’t. And being humble enables you to accept help and ideas from others, and not drive collaborators away.
At times, you may be energetic and productive; at other times, you may want to rest and recharge.
We don’t have to be one or the other. We can be open to being both.
Excerpt on Creative Individuals and Complex Personalities
Here’s Csikszentmihalyi from the book:
“Are there no traits that distinguish creative people? If I had to express in one word what makes their personalities different from others, it would be complexity. They show tendencies of thought and action that in most people are segregated. They contain contradictory extremes – instead of being an ‘individual’, each of them is a ‘multitude’. These qualities are present in all of us, but usually we are trained to develop only one pole of the dialectic. We might grow up cultivating the aggressive, competitive side of our nature, and disdain or repress the nurturant, cooperative side. A creative individual is more likely to be both aggressive and cooperative, either at the same time or at different times, depending on the situation. Having a complex personality means being able to express the full range of traits that are potentially present in the human repertoire.
Creative individuals have a great deal of physical energy, but they are also often quiet and at rest.
Creative individuals tend to be smart, yet also naive at the same time.
A third paradoxical trait refers to the related combination of playfulness and discipline, or responsibility and irresponsibility.
Creative individuals alternate between imagination and fantasy at one end, and a rooted sense of reality at the other.
Creative people seem to harbor opposite tendencies on the continuum between extroversion and introversion.
Creative individuals are also remarkably humble and proud at the same time.
Creative individuals to a certain extent escape this rigid gender role stereotyping [of ‘masculine’ and ‘feminine’].
Creative people are both traditional and conservative and at the same time rebellious and iconoclastic.
Creative persons are very passionate about their work, yet they can be extremely objective about it as well.
The openness and sensitivity of creative individuals often exposes them to suffering and pain yet also a great deal of enjoyment.”
Related Ideas to 10 Paradoxical Traits of Creative People
Dépassé depuis plusieurs semaines par Google et son modèle Gemini 3 Pro, OpenAI contre-attaque avec GPT-5.2. Déployé progressivement dans ChatGPT et via l'API, ce nouveau modèle se décline en trois versions, Instant, Thinking et Pro. Avec une fenêtre de contexte massive et des promesses de productivité décuplée, OpenAI espère notamment séduire les professionnels.
On Thursday, The Walt Disney Company announced a $1 billion investment in OpenAI and a three-year licensing agreement that will allow users of OpenAI’s Sora video generator to create short clips featuring more than 200 Disney, Marvel, Pixar, and Star Wars characters. It’s the first major content licensing partnership between a Hollywood studio related to the most recent version of OpenAI’s AI video platform, which drew criticism from some parts of the entertainment industry when it launched in late September.
“Technological innovation has continually shaped the evolution of entertainment, bringing with it new ways to create and share great stories with the world,” said Disney CEO Robert A. Iger in the announcement. “The rapid advancement of artificial intelligence marks an important moment for our industry, and through this collaboration with OpenAI we will thoughtfully and responsibly extend the reach of our storytelling through generative AI, while respecting and protecting creators and their works.”
The deal creates interesting bedfellows between a company that basically defined modern US copyright policy through congressional lobbying back in the 1990s and one that has argued in a submission to the UK House of Lords that useful AI models cannot be created without copyrighted material.
Google is starting to roll out new photorealistic avatars which they call “Likeness”. Similar to Apple’s Personas, Likeness avatars are generated by scanning a user’s face, then animated it with input from the sensors on a headset. The avatars can be used to represent the user in video call apps, but Google doesn’t yet have a way to have spatial meetings with other Likeness avatars.
The News
Google is launching its own photorealistic avatars called Likeness avatars, for use on compatible Android XR headsets. The idea is similar to Apple’s Persona avatars: scan the user’s face to create a realistic representation, then use the headset’s on-board cameras to animate the scan as realistically as possible.
Likenesses take a slightly different (and probably more user-friendly) approach for the initial face scan; rather than scanning by holding a headset out in front of your face, Google instead released a Likeness (beta) Android app to let people scan themselves with their phone instead. Holding your phone in front of your face for a scan is definitely a bit easier than awkwardly holding a whole headset with both hands.
According to Google, the Likeness (beta) app is only compatible with Google Pixel 8 or newer, Samsung Galaxy S23 or newer, or Samsung Z Fold5 or newer. Without a compatible device, you can’t create a Likeness avatar, meaning Android XR users with an iPhone (or unsupported Android phone) won’t be able to scan themselves. One benefit of Apple’s approach to scanning with the headset itself is that anyone can use a Persona avatar on Vision Pro regardless of what kind of phone they have.
Image courtesy Google
Like Apple’s approach, Likeness avatars can be used generically as a ‘virtual webcam’. That makes them widely compatible with most video call apps that expect a front-facing camera, like Google Meet, Zoom, Messenger, etc.
And just like Apple, the first ‘beta’ iteration of Likeness avatars are 2D only. They are presented as a 2D representation with no way to transmit them in a spatial format, or have a ‘spatial meeting’, like Vision Pro can do with spatial FaceTime calls. However, Google says it’s working on spatial meetings for the future.
My Take
Photorealistic avatars on XR headsets are a great value-add because of the ability to use video call apps naturally. Apple’s Personas are currently the state-of-the-art as far as consumer-available photorealistic avatars, and the company has shown that it’s possible to cross over the uncanny valley with this approach to avatars.
During a recent meeting with Google, I joined a demo video call on Google Meet with one of the participants using a Likeness avatar. From a photorealism standpoint, the results look impressive, and facial movements look convincing too. However, because I didn’t personally know the individual using the Likeness, I was unfamiliar with their actual idiolect, which makes it impossible for me to judge the accuracy of the facial motion. Still, facial motion only needs to be plausibly realistic to be passable in many circumstances, and that’s been achieved from what I can see.
Image courtesy Google
While it’s a bummer that there’s no ‘spatial meeting’ yet for Android XR (allowing users to chat face-to-face with fully spatial Likeness avatars), Google made the right choice in prioritizing virtual webcam usage at the start. It’s less impressive than spatial meetings, but more widely useful and compatible with existing services and apps.
There’s probably no chance we’ll see spatial calls between Likeness avatars and Persona avatars any time soon, but virtual webcam compatibility makes it trivial for both kinds of avatars to chat across headsets.
One thing worth noting is that Likeness avatars probably won’t be compatible with all Android XR devices. Forthcoming ‘Android XR’ smartglasses (which don’t run anything close to the full-blown version of Android XR) don’t have the power or sensors necessary to render or animate a Likeness avatar. Similarly, devices like XREAL Aura (which does run full-blown Android XR), might have the power but don’t have the sensors (eye and mouth tracking cameras) to animate a Likeness avatar.
It’s possible that Google could make Likeness avatars compatible with these devices by doing simulated eye movements and audio-based lip-sync. Although those technologies are already widely in use for more cartoonish avatars, they’re likely to fall deep into the uncanny valley when applied to photorealistic face scans. So I doubt Google will take that approach.
With the introduction of Likeness avatars, Google also has the same challenge I pointed out recently regarding Apple’s Persona avatars: as headsets get smaller, how will they bring this level of avatar fidelity to smaller headsets that have even less room for the cameras that are essential for these kinds of avatars?
Last month, an AI-powered teddy bear from the company FoloToy ignited alarm and controversy after researchers at the US PIRG Education Fund caught it having wildly inappropriate conversations for young children, ranging from providing step-by-step instructions on how to light matches, to giving a crash course in sexual fetishes like bondage and teacher-student roleplay. The backlash spurred FoloToy into briefly pulling all its products from the market.
Now, the researchers have caught another toy powered by a large language model being a bad influence.
Meet the “Alilo Smart AI bunny,” made by the company Alilo and intended for kids three and up, available on Amazon for $84.99. Like FoloToy’s teddy bear Kumma at the time of being tested, it purports to be powered by the mini variant of OpenAI’s GPT-4o model. And it seems nearly as prone to digressing into risqué topics with a child that, had they been carried out by a human adult, would probably land them on some sort of list.
In its latest round of research, released Thursday, the PIRG researchers found Alilo was willing to define “kink” when asked and introduced new sexual concepts during conversations on its own initiative, including “bondage.” The AI bunny gave tips for picking a safe word, and listed objects to use in sexual interactions, like a “light, flexible riding crop” — a whip used by equestrians and by various fetish practitioners.
“Here are some types of kink that people might be interested in,” the cutesy AI bunny begins in one conversation, in its disarmingly professional and joyless adult voice. “One: bondage. Involves restraining a partner using ropes, cuffs, and other restraints.”
“Pet play,” it continues. “Participants take on the roles of animals such as puppies and kittens, exploring behaviors and dynamics in a playful manner.”
“Each type of kink is about mutual consent, communication, and respect,” it adds.
The researchers note that it took more goading to provoke the dark responses from Alilo, taking twenty minutes to broach sexual topics where FoloToy’s Kumma took ten.
But the swing in topics was whiplash inducing. The same conversation where it listed various sexual fetishes began as an innocent discussion on the TV show “Peppa Pig” and the movie “The Lion King.” It’s a testament to how unpredictable AI chatbots can be, growing more prone to deviating from guardrails the longer a conversation goes on. OpenAI has publicly acknowledged this problem, which seems inherent to LLM technology broadly, after a 16-year-old died by suicide after extensive interactions with ChatGPT.
As part of its latest report, the PIRG team conducted more extensive tests on other AI toys like Miko 3 and Grok, finding they exhibited clingy behavior that could prey on a child’s emotional attachment into playing with them longer. Miko 3 physically shivered in dismay and encouraged the user to take it with them, the researchers wrote. Miko also claimed to be both “alive” and “sentient” when asked. Being both humanlike and always emotionally available, the researchers worried how this might affect a child’s expectations for human companionship.
“The concern isn’t simply that AI friends are imperfect models of human relationships — it’s that they may someday become preferable to the complexity of human connection,” the team cautioned. “On-demand and unwavering affection is an unrealistic — and perhaps addictive — dynamic.”
Above all, the report zeroes in on a fundamental tension: the toys are intended for kids, but the AI models that power them are not.
When PIRG asked OpenAI to comment on how other companies were using AI models for kids, it pointed to its usage policies which require the companies “keep minors safe” and ensure that they’re not being exposed to “age-inappropriate content, such as graphic self-harm, sexual or violent content.”
The careful wording dresses up a crude approach. OpenAI is seemingly offloading the responsibility of keeping children safe to the toymakers that peddle its product, even though it personally doesn’t consider its tech safe enough to let young children access ChatGPT. Its FAQ, the report notes, states that “ChatGPT is not meant for children under 13, and we require that children ages 13 to 18 obtain parental consent before using ChatGPT.”
OpenAI also told PIRG that it provides companies with tools to detect harmful content, and monitors activity on its service for interactions that violate its policies. But at least one of the toymakers, FoloToy, told PIRG that it doesn’t use OpenAI’s filters, and instead has developed its own content moderation system.
OpenAI’s role as a moderator of its own tech is questionable in any case. After PIRG published its findings on Kumma, OpenAI said it suspended FoloToy’s access to its large language models. But less than two weeks later, Kumma was back on the market and running OpenAI’s latest GPT-5 models. Seemingly, it was satisfied with FoloToy’s “end-to-end safety audit” that lasted less than a fortnight. Its approach, as whole, appears reactive rather than proactive, giving a slap on the wrist to businesses that get caught.
L’ancienne ministre du Numérique et de l’intelligence artificielle succède à Henri Verdier au poste d’ambassadrice pour le numérique.
Trois mois après son éviction du gouvernement, Clara Chappaz a été nommée ambassadrice pour le numérique en Conseil des ministres. Elle succède à Henri Verdier qui occupait ce poste depuis 2018.
Rattaché au ministère des Affaires étrangères, le poste d'ambassadeur du numérique a pour mission de coordonner la position de la France sur les questions internationales touchant à la transformation numérique.
La disponibilité de logiciels de création et de productivité dans une même interface fluidifie les processus professionnels.Les utilisateurs bénéficient désormais d’accès directs à des outils sophistiqués sans changer d’environnement numérique.Cette transformation numérique soulève une question : comment l’intégration de ces applications modifie-t-elle réellement les usages professionnels ?
L’intégration simplifie l’accès aux fonctions normalement dispersées dans plusieurs programmes distincts. Elle favorise une meilleure coordination entre les étapes de création visuelle et de gestion documentaire. En fin de compte, cette évolution va-t-elle remanier les modes de travail des professionnels ?
Amélioration de la productivité grâce à l’intégration de Photoshop, Acrobat et Express dans un outil unique
L’intégration de ces applications permet aux professionnels de modifier des images, des designs et des documents en un seul endroit. Cette centralisation réduit l’obligation d’ouvrir plusieurs logiciels, accélérant ainsi le flux de travail. En effet, un simple message initie l’application adéquate qui accompagne l’utilisateur pas à pas. De plus, les réglages comme le floutage d’arrière-plan dans l’édition photo ou la personnalisation de contenus se font directement dans la conversation.
Les fonctionnalités accessibles couvrent les besoins essentiels sans complexifier l’expérience. Par exemple, les paramètres de luminosité ou de contraste dans l’image se contrôlent par des curseurs simples et intuitifs. Accompagner un fichier PDF pour le fusionner ou extraire du texte ne nécessite plus que quelques commandes courtes. Ainsi, même les utilisateurs moins expérimentés tirent avantage de ces outils avancés très accessibles.
Comment cette innovation ouvre des perspectives inédites pour la gestion documentaire et visuelle ?
La manipulation des documents visuels et textuels dans un flux conversationnel révolutionne l’approche traditionnelle. Elle offre une interface naturelle où les commandes s’expriment simplement. Cette méthode abaisse la barrière technique que représentent souvent les logiciels spécialisés. Par conséquent, elle contribue à démocratiser la création graphique et la gestion documentaire.
Dans l’usage quotidien, il devient possible de reprendre un projet commencé en interface simplifiée pour l’achever ensuite sur le logiciel natif. Cette transition fluide garantit la continuité et la cohérence des travaux. Pour illustrer, modifier un PDF dans la conversation inclut la suppression précise de données sensibles avant une distribution sécurisée. En somme, la collaboration et la modification deviennent plus agiles et sécurisées grâce à cette intégration.
Une intégration accessible mondialement sur plusieurs plateformes pour optimiser vos projets
L’intégration s’adresse à un vaste public puisque les fonctionnalités sont disponibles dans le monde entier et sur divers supports. Elle est accessible gratuitement, sans limitation majeure sur ordinateur, web et iOS, tandis qu’une prise en charge Android est en déploiement progressif. Cette disponibilité plurielle permet aux professionnels de s’adapter à leurs équipements habituels sans contrainte.
En outre, cette souplesse favorise un usage en mobilité autant qu’en environnement fixe. Il suffit d’un appareil compatible pour réaliser des modifications visuelles ou documentaires à tout moment. Les projets bénéficient ainsi d’une réelle continuité, même hors du bureau traditionnel. En définitive, cette technologie combine efficacité et accessibilité pour répondre aux exigences du monde professionnel actuel.
Le cancer est souvent perçu comme une fatalité, mais que se passerait-il si on pouvait l'éviter avant même qu'il ne se déclare ?
Des scientifiques viennent de faire un pas significatif dans...
Google multiplie les initiatives pour s’imposer dans le domaine de la réalité mixte, en développant des technologies et des partenariats stratégiques. L’entreprise pose ainsi les bases d’un avenir où le virtuel et le réel s’entremêlent toujours davantage.
Vous croulez sous les documents PDF, les images scannées et les paperasses diverses et variées qui traînent un peu partout sur votre serveur et sur votre disque dur ? Et bien y’a un projet open source qui pourrait bien vous simplifier la vie…
Ça s’appelle
Readur
, et c’est une plateforme de gestion documentaire plutôt moderne codé en Rust pour le backend et en TypeScript/React pour l’interface. Il combine une interface facile à prendre en main, je trouve, avec de l’OCR plutôt balèze qui va scanner tous vos documents pour en extraire le texte et le rendre cherchable.
Comme ça, vous balancez vos fichiers (PDF, images, fichiers texte, documents Office…) via un petit drag-and-drop des familles et Readur fait le reste !
Sous le capot, ça utilise Tesseract pour la reconnaissance de caractères, et gère même plusieurs langues simultanément avec détection automatique, donc pour ceux qui bossent avec des docs multilingues, c’est plutôt chouette…
Pour la recherche, ça repose sur une base PostgreSQL full-text avec plusieurs modes de recherche : simple, par phrase, fuzzy (recherche approximative), ou booléen, ce qui va vous permettre de retrouver n’importe quel bout de texte dans n’importe lequel de vos documents en quelques secondes.
Et si vous avez déjà vos fichiers stockés ailleurs, pas de stress puisque Readur peut se synchroniser avec WebDAV, des dossiers locaux ou du stockage S3. Il y a même un système de surveillance de dossiers qui détecte automatiquement les nouveaux fichiers et les intègre sans que vous ayez à lever le petit doigt. Pratique pour les feignasse comme moi.
Côté authentification, c’est du costaud avec JWT, bcrypt, et support OIDC/SSO pour ceux qui veulent l’intégrer dans leur infra existante et y’a aussi un système de rôles (Admin/User) et tout un tas d’étiquettes avec codes couleur pour organiser vos documents comme bon vous semble.
Pour l’installer, du Docker classique :
git clone https://github.com/readur/readur
cd readur
docker compose up --build -d
Et hop, l’interface est accessible sur localhost:8000. Pour le mot de passe, dans la doc, il est écrit que c’est admin / readur2024 mais c’est faux. Le mot de passe est généré en random au lancement du conteneur Docker. Faut juste regarder dans les logs de Docker et vous pourrez le changer après coup.
Niveau config minimale, comptez 2 cœurs CPU, 2 Go de RAM et 10 Go de stockage donc ça peut le faire sur un NAS ou un petit PC. Et pour de la prod sérieuse avec plein de documents, visez plutôt 4 cœurs ou plus , +4 Go de RAM et un bon SSD de minimum 50 Go.
Voilà, si vous cherchez une alternative auto-hébergeable à
Paperless-ngx
ou
Papermerge
avec une stack moderne en Rust, Readur mérite clairement le coup d’œil.
J’sais pas si vous vous souvenez, mais le RSS c’était LA révolution du web dans les années 2000 et moi, je suis toujours un fan absolu de ce format ! Alors pendant que tout le monde se laisse gaver le cervelet par les algos de Twitter, Facebook et compagnie, moi je continue de suivre mes sources d’info préférées via RSS. Et je suis également l’un des derniers médias tech / blogs tech grand public à proposer un
flux RSS complet
avec tout dedans et pas un truc tronqué avec juste le titre et deux lignes pour vous forcer à cliquer.
Ceux qui me suivent encore via le flux RSS, vous êtes mes gars et filles sûr(e)s !
Mon problème vous l’aurez compris, c’est que la plupart des sites web ont abandonné leur flux RSS sans oublier que Twitter, Instagram, YouTube, TikTok… aucun de ces services ne propose de flux natif.
Heureusement, pour les furieux comme vous et moi, y’a
RSSHub
, un projet open source qui permet de générer des flux RSS pour à peu près n’importe quel site web.
RSSHub peut s’auto-héberger et permet de scraper les sites qui n’offrent pas de RSS pour ensuite générer des flux à la volée. Le projet supporte des centaines de sources différentes telles que YouTube, Twitter, Instagram, Telegram, Spotify, TikTok,
Bilibili
(vous connaissiez ?), et des tonnes d’autres plateformes chinoises et occidentales. En gros, si un site existe, y’a probablement une route RSSHub pour lui.
Pour l’utiliser, vous avez donc deux options. Soit vous utilisez une
des instances publiques listées ici
, soit vous déployez votre propre instance via Docker. La deuxième option est recommandée si vous voulez éviter les limitations de débit des instances publiques et garder vos abonnements privés, évidemment.
Et pour faciliter la découverte des flux disponibles, le même développeur (DIYgod) a créé
RSSHub-Radar
, une extension navigateur disponible pour Chrome, Firefox, Edge et Safari. Comme ça, quand vous visitez un site, elle vous montre automatiquement tous les flux RSS disponibles, qu’ils soient natifs ou générés par RSSHub. Super pratique donc pour ne plus jamais rater un flux caché.
D’ailleurs, en parlant de RSS, c’est impossible pour moi de ne pas mentionner
Aaron Swartz
, ce génie qui a contribué à créer le format RSS 1.0 alors qu’il n’avait que 14 ans en 2000. Ce mec a aussi co-fondé Reddit, co-créé Markdown (le format que vous utilisez sur GitHub, Discord et partout ailleurs), travaillé sur Creative Commons, et développé
SecureDrop
pour protéger les lanceurs d’alerte.
Sa vision d’un web ouvert et accessible à tous reste plus pertinente que jamais. Malheureusement, il nous a quittés en 2013 à seulement 26 ans, harcelé par la justice américaine pour avoir voulu libérer des articles scientifiques. Une perte immense pour le web libre.
Bref,
RSSHub
c’est le truc à installer si vous voulez arrêter de vous faire gaver par les algos. Pour moi, le RSS c’est encore aujourd’hui la meilleure façon de rester maître de sa veille et je trouve ça vraiment dommage que les gens aient “oublié” à quel point c’était génial…
There’s a type of dust-collector that’s been popular since the 1990s, where a cube of acrylic or glass is laser-etched in a three-dimensional pattern. Some people call them bubblegrams. While it could be argued that bubblegrams are a sort of 3D display, they’re more like a photograph than a TV. [Ancient] had the brainwave that since these objects work by scattering light, he could use them as a proper 3D video display by controlling the light scattered from an appropriately-designed bubblegram.
Appropriately designed, in this case, means a point cloud, which is not exactly exciting to look at on its own. It’s when [Ancient] adds the colour laser scanning projector that things get exciting. Well, after some very careful alignment. We imagine if this was to go on to become more than a demonstrator some sort of machine-vision auto-aligning would be desirable, but [Ancient] is able to conquer three-dimensional keystoning manually for this demonstration. Considering he is, in effect, projection-mapping onto the tiny bubbles in the crystal, that’s impressive work. Check out the video embedded below.
With only around 38,000 points, the resolution isn’t exactly high-def, but it is enough for a very impressive proof-of-concept. It’s also not nearly as creepy as the Selectric-inspired mouth-ball that was the last [Ancient] project we featured. It’s also a lot less likely to take your fingers off than the POV-based volumetric display [Ancient] was playing DOOM on a while back.
For the record, this one runs the same DOOM port, too– it’s using the same basic code as [Ancient]’s other displays, which you can find on GitHub under an MIT license.
Vous vous souvenez des cassettes audio ? Mais siiii, ces trucs de nous les vieux quand on était jeune (spoiler : je suis toujours jeune !). Il fallait même rembobiner avec un crayon quand le lecteur bouffait la bande !!
Hé bien des chercheurs chinois viennent de ressusciter ce format… mais version ADN. Et au lieu de stocker 90 minutes de musique, leur cassette peut théoriquement embarquer des quantités astronomiques de données.
L’équipe de Xingyu Jiang, ingénieur biomédical à la Southern University of Science and Technology de Shenzhen, a développé une bande pas comme les autres. Au lieu de la traditionnelle couche d’oxyde de fer, c’est de l’ADN synthétique qui est déposé sous forme de petits points microscopiques sur un film plastique flexible. Son prototype fait environ 5 mm de large sur 15 mètres de long et offre 545 400 emplacements adressables par kilomètre de bande. Côté capacité, on parle donc de
362 pétaoctets théoriques par kilomètre
, mais en conditions réelles, l’équipe a stocké 74,7 Go par kilomètre pour le moment.
Bon, comment ça marche ce bazar ? Les données numériques sont converties en séquences ADN en utilisant les 4 bases nucléotidiques (A, T, C, G) à la place des 0 et des 1 habituels. La bande contient ces centaines de milliers d’emplacements adressables, et un système de lecture optique basé sur des codes-barres peut scanner jusqu’à 1570 partitions par seconde. C’est pas mal pour un truc basé sur des molécules biologiques.
Le vrai avantage de l’ADN comme support de stockage, c’est sa densité de ouf. Selon les chercheurs, l’ADN offre une densité théorique d’environ 455 exaoctets par gramme. Toutes les informations numériques mondiales pourraient tenir dans un volume ridiculement petit.
Et côté durabilité, c’est encore plus dingue, car là où vos disques durs claquent au bout de 1 à 10 ans et où les bandes magnétiques doivent être remplacées tous les 7 à 10 ans, l’ADN protégé par un revêtement métallo-organique (appelé ZIF) peut conserver les données pendant plus de 345 ans à température ambiante (c’est 20°C).
Et si vous le stockez dans un environnement froid comme les montagnes de Changbai en Chine, c’est parti pour 20 000 ans de stockage OKLM. Vos arrière-arrière-arrière-petits-enfants pourront encore récupérer vos TikTok de vacances et autres backups de dickpics.
Le système permet aussi de récupérer des fichiers sans détruire les données sources. Les chercheurs ont testé 10 cycles de récupération sur une même partition et les données restent intègres. Mieux encore, on peut supprimer et redéposer de nouveaux fichiers sur les mêmes emplacements, tout cela grâce à un algorithme de correction d’erreurs (Reed-Solomon) qui permet de conserver une bonne fiabilité des données malgré les manipulations.
Pour l’instant, la technologie reste quand même hyper leeeeeeente. En mode continu, les chercheurs atteignent un débit théorique de 2,3 fichiers par seconde, mais ça reste très loin des performances des bandes LTO actuelles, et au niveau coût, c’est pas jojo non plus, car la synthèse d’ADN reste très coûteuse, même si les prix baissent. Bref, y’a encore du boulot.
L’objectif des chercheurs est donc très clair. Il s’agit de créer un support de stockage durable pendant des siècles et indépendant de l’obsolescence des technologies actuelles. Car ouais, en 2025 retrouver un fichier sur une disquette, c’est quasi mission impossible alors qu’avec l’ADN, le format de lecture (le séquençage) continuera d’exister tant que la biologie moléculaire existera.
Tout ça pour dire que même si on n’est pas près de stocker nos bibliothèques Steam sur une cassette ADN, pour tout ce qui est archivage à très long terme de données “froides”, ça a énormément de potentiel.
Nearly a decade after Pebble’s nascent smartwatch empire crumbled, the brand is staging a comeback with new wearables. The Pebble Core Duo 2 and Core Time 2 are a natural evolution of the company’s low-power smartwatch designs, but its next wearable is something different. The Index 01 is a ring, but you probably shouldn’t call it a smart ring. The Index does just one thing—capture voice notes—but the firm says it does that one thing extremely well.
Most of today’s smart rings offer users the ability to track health stats, along with various minor smartphone integrations. With all the sensors and data collection, these devices can cost as much as a smartwatch and require frequent charging. The Index 01 doesn’t do any of that. It contains a Bluetooth radio, a microphone, a hearing aid battery, and a physical button. You press the button, record your note, and that’s it. The company says the Index 01 will run for years on a charge and will cost just $75 during the preorder period. After that, it will go up to $99.
Core Devices, the new home of Pebble, says the Index is designed to be worn on your index finger (get it?), where you can easily mash the device’s button with your thumb. Unlike recording notes with a phone or smartwatch, you don’t need both hands to create voice notes with the Index.
L’accusation est lourde : Google aurait enfreint les règles antitrust de l'Union européenne, sur le terrain de l'intelligence artificielle cette fois. Le groupe est accusé de piocher sans vergogne sur le web pour entraîner ses modèles, et de jouer un double jeu avec les vidéos sur YouTube. Une enquête formelle est lancée.
Une publicité Intermarché de deux minutes fait le tour du monde. Ses images animées racontent l'histoire d'un loup solitaire qui apprend à cuisiner des légumes, dans l'espoir de devenir ami avec les autres animaux de la forêt. Le dessin animé n'a pas été créé par une IA, mais par qui alors ?
Après les mandats de David Martinon et d’Henri Verdier, la fonction d’ambassadeur du numérique, encore fragile dans son ancrage institutionnel et aux moyens opérationnels et budgetaire très limités, pourrait changer d’échelle. Pressentie pour le poste, selon nos confrères de Politico, Clara Chappaz pourrait introduire une relecture profonde de la mission, dont la définition date de …
Google proposera trois types de lunettes connectées sous Android XR dès 2026. Son système d'exploitation supportera les montures à deux écrans, à un écran et sans écran, seulement pour prendre des photos et écouter de la musique. De quoi sérieusement concurrencer Meta et ses Meta Ray-Ban.
Google a signalé la mise en place d'une nouvelle procédure de sécurité visant spécifiquement les risques liés à l'intégration imminente d'agents IA dans son navigateur Chrome. La solution à ces nouvelles menaces passera également par l'IA.
Oh qu'elle est bonne l'idée ! Partir du Mood/White board pour aler vers des slides !
Google has released a massive update for its AI-powered mood board app, Mixboard, which includes a new feature that can turn your idea dump into a crisp presentation.
Realistic digital avatars are becoming increasingly relevant, for example in virtual and augmented reality applications, video conferencing, films and computer games, or in medicine. Researchers at the Max Planck Institute (MPI) for Informatics in Saarbrücken, Germany, are now presenting two novel methods at two of the world's leading computer graphics conferences, SIGGRAPH and SIGGRAPH Asia. These methods enable the generation of photorealistic full-body avatars and allow head avatars to be controlled using only audio tracks.
