Tendances sur X · Archives
Top 9 tweets · La sélection conservée telle qu’elle a été publiée.
Message threads are now in public beta on the Messages API. The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.
Anthropic lance la mise en cache côté serveur pour son API, réduisant la latence et les coûts de bande passante pour les développeurs IA.
We built Pando, a new Git backend for our monorepo, to handle more pushes and fetches without changing how our engineers and agents use Git. Pando cut push latency by ~80% and now serves nearly a petabyte of Git data a day. We can grow Git capacity alongside agent usage.
OpenAI a dû rebâtir son infrastructure de contrôle de version pour gérer le volume massif de code généré par ses agents IA autonomes.
Absolutely incredible: MIT researchers say GPT-6 Astra improved a small Qwen model’s math score from 7.9% to 48.1% by directly editing just 0.016% of its weights, without gradient-based training. The result points toward AI systems that can diagnose and improve other models directly, a potential step toward self-improving AI.
Une étude du MIT montre un modèle frontière réécrivant les poids neuronaux d'un petit modèle pour améliorer son raisonnement.
As AI becomes more capable at finding vulnerabilities, defenders need to put it to work before attackers do. We’re planning our next internal security sprint at OpenAI and partnering with Cathedral Systems, whose researchers will use our models alongside our teams to challenge our defenses. They’ll bring differentiated expertise and approaches to our ongoing work to find and fix weaknesses with AI. We’re also raising bug bounty rewards, including more than quadrupling rewards for critical findings. Exceptional findings will be eligible for up to $172,000, including bonuses. We’ll keep advancing our models, using them to strengthen our own systems, and expanding access for trusted defenders. https://openai.com/index/preparing-our-next-security-sprint-to-harden-openai/
OpenAI augmente ses primes aux bugs et recrute des équipes rouges externes pour anticiper les failles découvertes par l'IA avancée.
BREAKING: Apple stock, $AAPL, drifts lower despite posting stronger than expected Q3 earnings. The company reported quarterly revenue of $114.5 billion.
Les marchés ont accueilli froidement les solides revenus d'Apple au T3, s'inquiétant des futures ventes de matériel.
The biggest item to watch is holiday quarter guidance. It's going to be a high hill to climb for serious growth when you factor in no iPhone 18, no iPhone Air 2, how minor the iPhone 18 Pro is as an update, how hard it is to get an iPhone Duo and the timing of other new products
Les analystes modèrent leurs attentes pour le trimestre des fêtes d'Apple en raison des contraintes d'approvisionnement du pliable.
It looks like X is starting to roll out both Liquid Glass and proper iPhone Duo support (Image from @ChrisMack_ )
Les réseaux sociaux mettent à jour leurs interfaces pour supporter le nouveau matériel pliable à double écran et design spatial d'Apple.
AI data centres on Earth will look like small fry compared to what's coming in SPACE. And SpaceX plans to make that happen. By the end of 2028, Elon predicts SpaceX will probably be putting more new computing power into orbit every month than the entire world adds on Earth. By 2029? Space pulls WAY ahead. Starship will carry the computing hardware, solar power will keep it running, and satellites will keep everything connected. While everyone else is figuring out where to build their next data centre, SpaceX is figuring out how many it can launch. @SpaceX @spacexai / Writer: Annette, Designer: Janné
Elon Musk propose des centres de données orbitaux solaires lancés par Starship pour contourner les limites du réseau électrique.
This represents the next significant step that China - and by extension, Huawei - is attempting to take: achieving as much independence as possible from the U.S. and from NVIDIA chips. That is why, for DeepSeek v5, they have opted to conduct the training almost entirely on Huawei’s own chips. The rationale behind this is that by breaking away from U.S. chips to the greatest extent possible and keeping development within the domestic market, China retains its leverage - specifically regarding rare earth elements - over the West, without having to rely on Western goodwill for the supply of chips.
L'entraînement réussi du dernier modèle de DeepSeek sur des puces nationales Huawei marque une étape majeure pour l'IA chinoise.
Message threads are now in public beta on the Messages API. The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.
Anthropic lance la mise en cache côté serveur pour son API, réduisant la latence et les coûts de bande passante pour les développeurs IA.
Absolutely incredible: MIT researchers say GPT-6 Astra improved a small Qwen model’s math score from 7.9% to 48.1% by directly editing just 0.016% of its weights, without gradient-based training. The result points toward AI systems that can diagnose and improve other models directly, a potential step toward self-improving AI.
Une étude du MIT montre un modèle frontière réécrivant les poids neuronaux d'un petit modèle pour améliorer son raisonnement.
BREAKING: Apple stock, $AAPL, drifts lower despite posting stronger than expected Q3 earnings. The company reported quarterly revenue of $114.5 billion.
Les marchés ont accueilli froidement les solides revenus d'Apple au T3, s'inquiétant des futures ventes de matériel.
It looks like X is starting to roll out both Liquid Glass and proper iPhone Duo support (Image from @ChrisMack_ )
Les réseaux sociaux mettent à jour leurs interfaces pour supporter le nouveau matériel pliable à double écran et design spatial d'Apple.
This represents the next significant step that China - and by extension, Huawei - is attempting to take: achieving as much independence as possible from the U.S. and from NVIDIA chips. That is why, for DeepSeek v5, they have opted to conduct the training almost entirely on Huawei’s own chips. The rationale behind this is that by breaking away from U.S. chips to the greatest extent possible and keeping development within the domestic market, China retains its leverage - specifically regarding rare earth elements - over the West, without having to rely on Western goodwill for the supply of chips.
L'entraînement réussi du dernier modèle de DeepSeek sur des puces nationales Huawei marque une étape majeure pour l'IA chinoise.
We built Pando, a new Git backend for our monorepo, to handle more pushes and fetches without changing how our engineers and agents use Git. Pando cut push latency by ~80% and now serves nearly a petabyte of Git data a day. We can grow Git capacity alongside agent usage.
OpenAI a dû rebâtir son infrastructure de contrôle de version pour gérer le volume massif de code généré par ses agents IA autonomes.
As AI becomes more capable at finding vulnerabilities, defenders need to put it to work before attackers do. We’re planning our next internal security sprint at OpenAI and partnering with Cathedral Systems, whose researchers will use our models alongside our teams to challenge our defenses. They’ll bring differentiated expertise and approaches to our ongoing work to find and fix weaknesses with AI. We’re also raising bug bounty rewards, including more than quadrupling rewards for critical findings. Exceptional findings will be eligible for up to $172,000, including bonuses. We’ll keep advancing our models, using them to strengthen our own systems, and expanding access for trusted defenders. https://openai.com/index/preparing-our-next-security-sprint-to-harden-openai/
OpenAI augmente ses primes aux bugs et recrute des équipes rouges externes pour anticiper les failles découvertes par l'IA avancée.
The biggest item to watch is holiday quarter guidance. It's going to be a high hill to climb for serious growth when you factor in no iPhone 18, no iPhone Air 2, how minor the iPhone 18 Pro is as an update, how hard it is to get an iPhone Duo and the timing of other new products
Les analystes modèrent leurs attentes pour le trimestre des fêtes d'Apple en raison des contraintes d'approvisionnement du pliable.
AI data centres on Earth will look like small fry compared to what's coming in SPACE. And SpaceX plans to make that happen. By the end of 2028, Elon predicts SpaceX will probably be putting more new computing power into orbit every month than the entire world adds on Earth. By 2029? Space pulls WAY ahead. Starship will carry the computing hardware, solar power will keep it running, and satellites will keep everything connected. While everyone else is figuring out where to build their next data centre, SpaceX is figuring out how many it can launch. @SpaceX @spacexai / Writer: Annette, Designer: Janné
Elon Musk propose des centres de données orbitaux solaires lancés par Starship pour contourner les limites du réseau électrique.
Message threads are now in public beta on the Messages API. The API caches your conversation on the server between turns, so instead of resending the full conversation, you send only new messages and the previous message ID, and Claude starts responding sooner.
Anthropic lance la mise en cache côté serveur pour son API, réduisant la latence et les coûts de bande passante pour les développeurs IA.
As AI becomes more capable at finding vulnerabilities, defenders need to put it to work before attackers do. We’re planning our next internal security sprint at OpenAI and partnering with Cathedral Systems, whose researchers will use our models alongside our teams to challenge our defenses. They’ll bring differentiated expertise and approaches to our ongoing work to find and fix weaknesses with AI. We’re also raising bug bounty rewards, including more than quadrupling rewards for critical findings. Exceptional findings will be eligible for up to $172,000, including bonuses. We’ll keep advancing our models, using them to strengthen our own systems, and expanding access for trusted defenders. https://openai.com/index/preparing-our-next-security-sprint-to-harden-openai/
OpenAI augmente ses primes aux bugs et recrute des équipes rouges externes pour anticiper les failles découvertes par l'IA avancée.
It looks like X is starting to roll out both Liquid Glass and proper iPhone Duo support (Image from @ChrisMack_ )
Les réseaux sociaux mettent à jour leurs interfaces pour supporter le nouveau matériel pliable à double écran et design spatial d'Apple.
We built Pando, a new Git backend for our monorepo, to handle more pushes and fetches without changing how our engineers and agents use Git. Pando cut push latency by ~80% and now serves nearly a petabyte of Git data a day. We can grow Git capacity alongside agent usage.
OpenAI a dû rebâtir son infrastructure de contrôle de version pour gérer le volume massif de code généré par ses agents IA autonomes.
BREAKING: Apple stock, $AAPL, drifts lower despite posting stronger than expected Q3 earnings. The company reported quarterly revenue of $114.5 billion.
Les marchés ont accueilli froidement les solides revenus d'Apple au T3, s'inquiétant des futures ventes de matériel.
AI data centres on Earth will look like small fry compared to what's coming in SPACE. And SpaceX plans to make that happen. By the end of 2028, Elon predicts SpaceX will probably be putting more new computing power into orbit every month than the entire world adds on Earth. By 2029? Space pulls WAY ahead. Starship will carry the computing hardware, solar power will keep it running, and satellites will keep everything connected. While everyone else is figuring out where to build their next data centre, SpaceX is figuring out how many it can launch. @SpaceX @spacexai / Writer: Annette, Designer: Janné
Elon Musk propose des centres de données orbitaux solaires lancés par Starship pour contourner les limites du réseau électrique.
Absolutely incredible: MIT researchers say GPT-6 Astra improved a small Qwen model’s math score from 7.9% to 48.1% by directly editing just 0.016% of its weights, without gradient-based training. The result points toward AI systems that can diagnose and improve other models directly, a potential step toward self-improving AI.
Une étude du MIT montre un modèle frontière réécrivant les poids neuronaux d'un petit modèle pour améliorer son raisonnement.
The biggest item to watch is holiday quarter guidance. It's going to be a high hill to climb for serious growth when you factor in no iPhone 18, no iPhone Air 2, how minor the iPhone 18 Pro is as an update, how hard it is to get an iPhone Duo and the timing of other new products
Les analystes modèrent leurs attentes pour le trimestre des fêtes d'Apple en raison des contraintes d'approvisionnement du pliable.
This represents the next significant step that China - and by extension, Huawei - is attempting to take: achieving as much independence as possible from the U.S. and from NVIDIA chips. That is why, for DeepSeek v5, they have opted to conduct the training almost entirely on Huawei’s own chips. The rationale behind this is that by breaking away from U.S. chips to the greatest extent possible and keeping development within the domestic market, China retains its leverage - specifically regarding rare earth elements - over the West, without having to rely on Western goodwill for the supply of chips.
L'entraînement réussi du dernier modèle de DeepSeek sur des puces nationales Huawei marque une étape majeure pour l'IA chinoise.
Classés par engagement. Les publications appartiennent à leurs auteurs sur X. Chiffres relevés lors de la préparation de cette sélection.
The Brief
IA et technologie : ce qui change et pourquoi cela compte.
Votre sélection quotidienne, en anglais ou en espagnol.
Gratuit pour toujours. Désabonnement à tout moment.
