Presented by HubSpot<hr/>INBOUND, HubSpot&#x27;s annual conference for marketing and sales professionals, took place in San Francisco this year, with three days of insights and events across marketing, sales, CX, and AI innovation. It was a mix of the new, like the Creators Corner and the Tech Stack Showcase Stage, and the familiar, like HubSpot HQ, the Spotlight Product Demo Stage, HubSpot Academy Labs for hands-on product learning, and attendee favorite, Braindates. The opening HubSpot Spotlight on Wednesday, by HubSpot co-founder and CTO, Dharmesh Shah, dug into the nitty gritty of AI, what transformation really looks like, why generative AI is more than just glorified autocomplete, and why it&#x27;s good a thing that it&#x27;s not going anywhere, anytime soon.<blockquote>&quot;Is it an exponential opportunity? Or is it an existential threat? I asked AI this question. It thought deeply about it for a while and came up with its answer. Yes.&quot;</blockquote>&quot;Not everyone is necessarily sold on the merits of AI. It’s the most consequential, and also the most controversial, technology of our time,&quot; Shah said. &quot;Is it an exponential opportunity? Or is it an existential threat? I asked AI this question. It thought deeply about it for a while and came up with its answer. Yes.&quot;<h3>How do you compete with AI?</h3>An informal poll of 6K respondents asked the question, &quot;How do you compete with AI?” Shah said a third of the respondents read that as, how do I compete against it? But, he added, that frames it as a zero-sum game which isn&#x27;t useful to anyone. &quot;We should think about it as a positive-sum collaboration,&quot; he explained. &quot;The goal isn’t to battle the machine. The goal is to build with the machine.&quot;This is critical because while AI capability is growing on an exponential curve, the learning curve for AI is much more linear, and so is the value seen from embracing the technology that&#x27;s learned everything it knows from Shakespeare&#x27;s full body of work, Reddit threads debating whether a hot dog is a sandwich, and Stephen Hawking&#x27;s academic talks. &quot;At the end of it all, this machine is capable of predicting, with such accuracy, a really good next word, that what it’s doing is indistinguishable from thinking. That’s why it feels so magical,&quot; Shah said. &quot;It’s autocomplete with a PhD in everything. It’s like having 1,000 PhDs in your pocket. Suddenly you can write poetry. You can write prose. You can write computer programs. It’s like Neo in the Matrix, except it’s not &#x27;I know kung fu.&#x27; It’s &#x27;I know everything.&#x27; You have the world’s knowledge sitting there in your pocket.”There are some drawbacks. LLMs are limited to the training data they were given. Sometimes they hallucinate. They&#x27;re frozen in time, and stateless — but they do learn. User interactions go into their long-term memories, including chats, resources they&#x27;re asked to interact with, such as PDFs or images, and tools like the internet, databases, APIs with third-party systems and more. <h3>How to better use AI in our daily lives </h3><blockquote>&quot;My pro tip: every time you sit down at a computer to do something, try it first with AI. See if it can help, don’t overthink it.&quot;</blockquote>So if AI is the real deal, and the transformation is important, how should you use AI? &quot;My pro tip: every time you sit down at a computer to do something, try it first with AI. See if it can help, don’t overthink it,&quot; Shah said. &quot;You’ll be surprised. When it doesn’t work, don’t think, oh, well, this doesn’t work. Think, this doesn’t work yet. If it’s an important use case for you, leave yourself a calendar reminder to try it again in three months. Remember, AI is on an exponential curve. There’s a chance, three months or six months from now, that same thing that doesn’t work now will work.&quot;The quality of the results you get is based on the quality of the model you use, the quality of the prompt you write, and the quality of the context you put in the context window, otherwise known as prompt engineering. The quality of the model. There&#x27;s been a Cambrian explosion in the world of large language models, but Shah recommends any of the three top-tier frontier models: OpenAI&#x27;s GPT-5, Claude from Anthropic, or Google Gemini. But don&#x27;t overthink it, he adds. Pick one you like or that your company is using.The quality of the prompt. Humans use considerably less than 10% of AI&#x27;s potential, Shah said. About 95% of the time, users repeat one of the handful of prompts they&#x27;ve found that have worked for the majority of their use cases. Instead, he recommends that 60% of the time, you use the prompts that work, carving out 30% of the time for iteration, or taking the prompts you&#x27;re using right now and seeing if you can improve them for better results, spending 10% of your time on experimentation, or using AI for things that you’ve never tried before, and you&#x27;re not sure will work. Metaprompting can be useful here — actually asking AI to make the prompt better<h3>Beyond prompt engineering</h3>Context engineering is just what it sounds like — adding the right context to a request so AI offers better results. Improving the quality of the context includes things like adding custom instructions, which every major LLM includes. That&#x27;s basically telling the AI what perspective it&#x27;s working from — the personality, how it should behave, how it should answer and so on. Once you provide custom instructions, it will use them for every answer, and your results improve.The other improvement is a recent change — MCP, or model context protocol, a relatively new way to supply tools to the LLM. Any AI application that supports MCP can immediately and directly connect with the thousands of applications that also support MCP, like HubSpot, to bring that context directly into the LLM. <h3>Taking action with AI agents </h3>&quot;Last year, at Inbound 24, I predicted that this year would be the year of AI agents. I was wrong,&quot; Shah said. &quot;This is not the year of AI agents. This is the decade of AI agents. We are just getting started on this massive transformational wave in AI.&quot; Last year Shah launched <a href="http://agent.ai/">Agent.ai</a>, a place to discover, use, and build your own AI agents. Over 2 million people are now using the platform. Around 26,000 users have already built agents on Agenti.ai, including Shah, who took all of the information from his keynote prep and put it into an AI agent anyone can chat with, at <a href="http://you.ai/">You.ai</a>. <blockquote>The TEAM strategy: triage, experiment, automate, and measure. </blockquote><h3>The next steps</h3>How do you take your newfound AI enthusiasm and bring it to your team? Shah offered a gameplan, the TEAM strategy: triage, experiment, automate, and measure. &quot;The goal behind that strategy is to go from AI being led by individual heroics, where you have the super ambitious, super creative person, but translate those individual heroics into team habits,&quot; he explained. &quot;Take those things and apply TEAM to go through that process again and again. We are big believers that tomorrow’s teams will be hybrid. This is a time to get started. Even in a small way, start building out your hybrid team.&quot;But the most important thing to remember is that as smart as AI is, humans win on EQ, or emotional quotient. Marrying lived experience with an AI tool will improve your life, and improve your experiences, he added. &quot;But my challenge to you is, don’t stop there,&quot; he said. &quot;The future does not belong to artificial intelligence. It belongs to you, with augmented intelligence. AI isn’t here to replace us. It’s here to replace the parts of our work that don’t bring us joy. To handle the repetitive so we can focus on the remarkable. The better AI gets, the more it allows us to be human.&quot;<hr/>Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact <a href="mailto:sales@venturebeat.com">sales@venturebeat.com</a>.

في مؤتمر INBOUND لهذا العام في سان فرانسيسكو، أكد دارميش شاه من HubSpot على أهمية إتقان الذكاء الاصطناعي من خلال استخدام العبارات الفعالة والسياق والتجريب. جمع هذا الحدث محترفي التسويق والمبيعات لاستكشاف أفكار وأدوات مبتكرة، مع عرض ميزات جديدة مثل Creators Corner وعناصر محبوبة مثل مختبرات أكاديمية HubSpot. الأفكار التي تم مشاركتها مهمة للمحترفين الذين يسعون للاستفادة من الذكاء الاصطناعي في استراتيجياتهم، مما يجعل هذا المؤتمر فرصة كبيرة للنمو والتعلم.

En la conferencia INBOUND de este año en San Francisco, Dharmesh Shah de HubSpot destacó la importancia de dominar la IA a través de prompts efectivos, contexto y experimentación. Este evento reunió a profesionales de marketing y ventas para explorar ideas y herramientas innovadoras, mostrando tanto nuevas características como el Creators Corner y elementos queridos como los HubSpot Academy Labs. Las ideas compartidas son cruciales para los profesionales que buscan aprovechar la IA en sus estrategias, convirtiendo esta conferencia en una oportunidad significativa para el crecimiento y el aprendizaje.

Lors de la conférence INBOUND de cette année à San Francisco, Dharmesh Shah de HubSpot a souligné l'importance de maîtriser l'IA grâce à des invites efficaces, un bon contexte et l'expérimentation. Cet événement a rassemblé des professionnels du marketing et des ventes pour explorer des idées et des outils innovants, mettant en avant à la fois de nouvelles fonctionnalités comme le Creators Corner et des éléments appréciés comme les HubSpot Academy Labs. Les idées partagées sont cruciales pour les professionnels cherchant à tirer parti de l'IA dans leurs stratégies, faisant de cette conférence une opportunité significative de croissance et d'apprentissage.

At this year's INBOUND conference in San Francisco, HubSpot's Dharmesh Shah emphasized the importance of mastering AI through effective prompts, context, and experimentation. This event brought together marketing and sales professionals to explore innovative ideas and tools, showcasing both new features like the Creators Corner and beloved staples like HubSpot Academy Labs. The insights shared are crucial for professionals looking to leverage AI in their strategies, making this conference a significant opportunity for growth and learning.

HubSpot’s Dharmesh Shah on AI mastery: Why prompts, context, and experimentation matter most

&nbsp;As US educators embrace AI in the classroom, firms are selling software to flag mentions of self-harm, raising concerns over privacy and control.

مع تزايد استخدام تقنية الذكاء الاصطناعي في الفصول الدراسية في الولايات المتحدة، بدأ المعلمون في استخدام الدردشة الآلية لمراقبة رفاهية الطلاب. بينما يمكن أن تساعد هذه الأدوات في تحديد علامات السلوكيات الذاتية الضارة، فإنها تثير أيضًا مخاوف كبيرة بشأن الخصوصية ومدى المراقبة في البيئات التعليمية. قد يؤثر هذا التحول نحو المراقبة المستمرة على ثقة الطلاب وحريتهم، مما يجعله قضية حاسمة للآباء والمعلمين على حد سواء.

A medida que la tecnología de IA se vuelve más común en las aulas de EE. UU., los educadores están utilizando cada vez más chatbots para monitorear el bienestar de los estudiantes. Si bien estas herramientas pueden ayudar a identificar signos de autolesionismo, también plantean preocupaciones significativas sobre la privacidad y el alcance de la vigilancia en los entornos educativos. Este cambio hacia la vigilancia constante podría afectar la confianza y la libertad de los estudiantes, convirtiéndolo en un tema crítico para padres y educadores.

Alors que la technologie de l'IA devient de plus en plus courante dans les salles de classe aux États-Unis, les éducateurs utilisent de plus en plus des chatbots pour surveiller le bien-être des élèves. Bien que ces outils puissent aider à identifier des signes de comportements autodestructeurs, ils soulèvent également d'importantes préoccupations concernant la vie privée et l'étendue de la surveillance dans les établissements éducatifs. Ce passage vers une surveillance constante pourrait affecter la confiance et la liberté des élèves, ce qui en fait un enjeu crucial pour les parents et les éducateurs.

As AI technology becomes more prevalent in classrooms across the US, educators are increasingly using chatbots to monitor student well-being. While these tools can help identify signs of self-harm, they also raise significant concerns about privacy and the extent of surveillance in educational settings. This shift towards constant monitoring could impact students' trust and freedom, making it a critical issue for parents and educators alike.

Chatbots Are Sparking a New Era of Student Surveillance

Olympus Corp.’s new Chief Executive Officer Bob White is shaking up the Tokyo-based medical devices maker after a difficult period that included the ouster of his predecessor in a drug scandal.

تخضع شركة أوليمبوس كورب. لتغييرات كبيرة تحت قيادة الرئيس التنفيذي الجديد بوب وايت، الذي ينفذ خطة إعادة هيكلة تهدف إلى إحياء الشركة بعد فترة صعبة تميزت بفضيحة تتعلق بالمخدرات وإقالة سلفه. هذه الخطوة مهمة لأنها تشير إلى بداية جديدة لأوليومبوس، مما قد يعيد ثقة المستثمرين ويحسن الكفاءة التشغيلية في سوق الأجهزة الطبية التنافسية.

Olympus Corp. está experimentando cambios significativos bajo el liderazgo del nuevo CEO Bob White, quien está implementando un plan de reestructuración destinado a revitalizar la empresa tras un período difícil marcado por un escándalo de drogas y la destitución de su predecesor. Este movimiento es crucial, ya que señala un nuevo comienzo para Olympus, lo que podría restaurar la confianza de los inversores y mejorar la eficiencia operativa en el competitivo mercado de dispositivos médicos.

Olympus Corp. connaît des changements importants sous la direction du nouveau PDG Bob White, qui met en œuvre un plan de restructuration visant à revitaliser l'entreprise après une période difficile marquée par un scandale lié aux médicaments et le renvoi de son prédécesseur. Ce mouvement est crucial car il signale un nouveau départ pour Olympus, ce qui pourrait restaurer la confiance des investisseurs et améliorer l'efficacité opérationnelle sur le marché concurrentiel des dispositifs médicaux.

Olympus Corp. is undergoing significant changes under the leadership of new CEO Bob White, who is implementing a restructuring plan aimed at revitalizing the company after a challenging period marked by a drug scandal and the ousting of his predecessor. This move is crucial as it signals a fresh start for Olympus, potentially restoring investor confidence and improving operational efficiency in the competitive medical devices market.

Olympus Cuts Jobs, Streamlines Operations Under New Leadership

I have no problem recommending the Hoto 12-in-1 electric screwdriver - a reliable tool that's been in my kit for over a year now.

لقد أثبت مفك البراغي الكهربائي Hoto 12 في 1 أنه أداة موثوقة في مجموعة أدواتي لأكثر من عام، والآن أصبح أكثر تكلفة. هذه التخفيضات في الأسعار تجعله خيارًا ممتازًا لأي شخص يتطلع إلى تحسين مشاريعه اليدوية دون إنفاق الكثير.

El destornillador eléctrico Hoto 12 en 1 ha demostrado ser una herramienta confiable en mi kit durante más de un año, y ahora es aún más asequible. Esta reducción de precio lo convierte en una excelente opción para cualquiera que busque mejorar sus proyectos de bricolaje sin gastar demasiado.

Le tournevis électrique Hoto 12-en-1 s'est révélé être un outil fiable dans ma boîte à outils depuis plus d'un an, et maintenant il est encore plus abordable. Cette baisse de prix en fait un excellent choix pour quiconque souhaite améliorer ses projets de bricolage sans se ruiner.

The Hoto 12-in-1 electric screwdriver has proven to be a dependable tool in my toolkit for over a year, and now it's even more affordable. This price drop makes it an excellent choice for anyone looking to enhance their DIY projects without breaking the bank.

This electric screwdriver is a reliable addition to my toolkit - and it just got even cheaper

<a href="https://news.ycombinator.com/item?id=45845772">Comments</a>

تكشف التوقعات الأخيرة لشركة ميتا أن نسبة مذهلة تبلغ 10% من إيراداتها المتوقعة لعام 2024 قد تأتي من الاحتيالات. تسلط هذه الإحصائية المقلقة الضوء على المشكلة المتزايدة للأنشطة الاحتيالية على المنصة، مما يثير القلق بشأن سلامة المستخدمين وثقتهم. بينما تتعامل ميتا مع هذه التحديات، قد تكون العواقب على سمعتها وصحتها المالية كبيرة، مما يجعلها موضوعًا حاسمًا للمستخدمين والمستثمرين على حد سواء.

Las recientes proyecciones de Meta revelan que un asombroso 10% de sus ingresos anticipados para 2024 podrían provenir de estafas. Esta estadística alarmante destaca el creciente problema de las actividades fraudulentas en la plataforma, lo que genera preocupaciones sobre la seguridad y la confianza de los usuarios. A medida que Meta navega por estos desafíos, las implicaciones para su reputación y salud financiera podrían ser significativas, convirtiéndolo en un tema crítico para usuarios e inversores.

Les récentes projections de Meta révèlent qu'un impressionnant 10 % de ses revenus prévus pour 2024 pourraient provenir d'escroqueries. Cette statistique alarmante met en lumière le problème croissant des activités frauduleuses sur la plateforme, suscitant des inquiétudes quant à la sécurité et à la confiance des utilisateurs. Alors que Meta fait face à ces défis, les implications pour sa réputation et sa santé financière pourraient être significatives, ce qui en fait un sujet crucial pour les utilisateurs et les investisseurs.

Meta's recent projections reveal that a staggering 10% of its anticipated revenue for 2024 could stem from scams. This alarming statistic highlights the growing issue of fraudulent activities on the platform, raising concerns about user safety and trust. As Meta navigates these challenges, the implications for its reputation and financial health could be significant, making it a critical topic for users and investors alike.

Meta projected 10% of 2024 revenue came from scams

Garmin could be working on an annual summary to help you understand your health and fitness.

تستعد شركة جارمين لإطلاق ميزة جديدة في عام 2025 ستوفر للمستخدمين ملخصًا سنويًا مخصصًا لأبرز إنجازاتهم في اللياقة البدنية، مشابهًا لتجربة سبوتيفاي وراپيد. تهدف هذه المبادرة إلى مساعدة الأفراد على فهم رحلتهم الصحية واللياقية بشكل أفضل، مما يسهل تتبع التقدم وتحديد الأهداف المستقبلية. إنه تطور مثير لعشاق اللياقة البدنية الذين يرغبون في البقاء متحمسين ومطلعين على إنجازاتهم.

Garmin se está preparando para lanzar una nueva función en 2025 que proporcionará a los usuarios un resumen anual personalizado de sus momentos destacados en fitness, similar a Spotify Wrapped. Esta iniciativa tiene como objetivo ayudar a las personas a comprender mejor su viaje de salud y fitness, facilitando el seguimiento del progreso y la fijación de objetivos futuros. Es un desarrollo emocionante para los entusiastas del fitness que desean mantenerse motivados e informados sobre sus logros.

Garmin se prépare à lancer une nouvelle fonctionnalité en 2025 qui fournira aux utilisateurs un résumé annuel personnalisé de leurs moments forts en matière de fitness, similaire à Spotify Wrapped. Cette initiative vise à aider les individus à mieux comprendre leur parcours de santé et de fitness, facilitant ainsi le suivi des progrès et la définition des objectifs futurs. C'est un développement passionnant pour les amateurs de fitness qui souhaitent rester motivés et informés de leurs réalisations.

Garmin is gearing up to launch a new feature in 2025 that will provide users with a personalized annual summary of their fitness highlights, similar to Spotify Wrapped. This initiative aims to help individuals better understand their health and fitness journey, making it easier to track progress and set future goals. It's an exciting development for fitness enthusiasts who want to stay motivated and informed about their achievements.

Garmin is planning a Spotify Wrapped-style review of your fitness highlights in 2025

Tune out (or rock out) with our favorite over-ears and earbuds.

في عام 2025، تتصاعد المنافسة على أفضل سماعات الرأس المزودة بتقنية إلغاء الضوضاء، مع تصدر علامات تجارية مثل سوني وبوز. لا تساعدك هذه السماعات فقط على عزل نفسك عن المشتتات، بل تعزز أيضًا تجربتك السمعية، مما يجعلها ضرورية لعشاق الموسيقى والمحترفين على حد سواء. مع التقدم التكنولوجي، تقدم الطرازات الأحدث جودة صوت وراحة فائقة، مما يضمن لك الاستمتاع بأغانيك أو بودكاستاتك المفضلة دون انقطاع.

En 2025, la competencia por los mejores auriculares con cancelación de ruido se intensifica, con marcas líderes como Sony y Bose a la cabeza. Estos auriculares no solo te ayudan a desconectarte de las distracciones, sino que también mejoran tu experiencia auditiva, haciéndolos esenciales para los amantes de la música y los profesionales. Con los avances tecnológicos, los últimos modelos ofrecen una calidad de sonido y comodidad superiores, asegurando que puedas disfrutar de tus canciones o podcasts favoritos sin interrupciones.

En 2025, la compétition pour les meilleurs écouteurs à réduction de bruit s'intensifie, avec des marques de premier plan comme Sony et Bose en tête. Ces écouteurs vous aident non seulement à vous isoler des distractions, mais améliorent également votre expérience d'écoute, ce qui les rend essentiels pour les amateurs de musique et les professionnels. Avec les avancées technologiques, les derniers modèles offrent une qualité sonore et un confort supérieurs, garantissant que vous pouvez profiter de vos morceaux ou podcasts préférés sans interruption.

In 2025, the competition for the best noise-canceling headphones is heating up, with top brands like Sony and Bose leading the charge. These headphones not only help you tune out distractions but also enhance your listening experience, making them essential for music lovers and professionals alike. With advancements in technology, the latest models offer superior sound quality and comfort, ensuring that you can enjoy your favorite tunes or podcasts without interruption.

Best Noise-Canceling Headphones (2025): Sony, Bose, and More

Even as <a href="https://www.tomshardware.com/tech-industry/openai-walks-back-statement-it-wants-a-government-backstop-for-its-massive-loans-company-says-government-playing-its-part-critical-for-industrial-ai-capacity-increases">concern and skepticism</a> grows over U.S. AI startup OpenAI&#x27;s buildout strategy and high spending commitments, Chinese open source AI providers are escalating their competition and one has even caught up to OpenAI&#x27;s flagship, paid proprietary model GPT-5 in key third-party performance benchmarks with a new, free model. The Chinese AI startup <a href="https://moonshotai.github.io/Kimi-K2/thinking.html">Moonshot AI’s new Kimi K2 Thinking model</a>, released today, has vaulted past both proprietary and open-weight competitors to claim the top position in reasoning, coding, and agentic-tool benchmarks. Despite being fully open-source, the model now outperforms OpenAI’s GPT-5, Anthropic’s Claude Sonnet 4.5 (Thinking mode), and xAI&#x27;s Grok-4 on several standard evaluations — an inflection point for the competitiveness of open AI systems.Developers can access the model via <a href="https://platform.moonshot.ai">platform.moonshot.ai</a> and <a href="https://kimi.com">kimi.com</a>; weights and code are hosted on <a href="https://huggingface.co/moonshotai/Kimi-K2-Thinking/tree/main">Hugging Face</a>. The open release includes APIs for chat, reasoning, and multi-tool workflows.Users can try out Kimi K2 Thinking directly through its own <a href="https://www.kimi.com/">ChatGPT-like website competitor</a> and on <a href="https://huggingface.co/spaces/moonshotai/Kimi-VL-A3B-Thinking">a Hugging Face space as well. </a><h3>Modified Standard Open Source License</h3>Moonshot AI has formally released Kimi K2 Thinking under a <a href="https://huggingface.co/moonshotai/Kimi-K2-Thinking/blob/main/LICENSE">Modified MIT License</a> on Hugging Face.The license grants full commercial and derivative rights — meaning individual researchers and developers working on behalf of enterprise clients can access it freely and use it in commercial applications — but adds one restriction:&quot;If the software or any derivative product serves over 100 million monthly active users or generates over $20 million USD per month in revenue, the deployer must prominently display &#x27;Kimi K2&#x27; on the product’s user interface.&quot;For most research and enterprise applications, this clause functions as a light-touch attribution requirement while preserving the freedoms of standard MIT licensing. It makes K2 Thinking one of the most permissively licensed frontier-class models currently available.<h3>A New Benchmark Leader</h3>Kimi K2 Thinking is a Mixture-of-Experts (MoE) model built around one trillion parameters, of which 32 billion activate per inference. It combines long-horizon reasoning with structured tool use, executing up to 200–300 sequential tool calls without human intervention.According to Moonshot’s published test results, K2 Thinking achieved:<ul><li>44.9 % on Humanity’s Last Exam (HLE), a state-of-the-art score;</li><li>60.2 % on BrowseComp, an agentic web-search and reasoning test;</li><li>71.3 % on SWE-Bench Verified and 83.1 % on LiveCodeBench v6, key coding evaluations;</li><li>56.3 % on Seal-0, a benchmark for real-world information retrieval.</li></ul>Across these tasks, K2 Thinking consistently outperforms GPT-5’s corresponding scores and <a href="https://venturebeat.com/ai/minimax-m2-is-the-new-king-of-open-source-llms-especially-for-agentic-tool">surpasses the previous open-weight leader MiniMax-M2</a>—released just weeks earlier by Chinese rival MiniMax AI.<h3>Open Model Outperforms Proprietary Systems</h3>GPT-5 and Claude Sonnet 4.5 Thinking remain the leading proprietary “thinking” models. Yet in the same benchmark suite, K2 Thinking’s agentic reasoning scores exceed both: for instance, on BrowseComp the open model’s 60.2 % decisively leads GPT-5’s 54.9 % and Claude 4.5’s 24.1 %.K2 Thinking also edges GPT-5 in GPQA Diamond (85.7 % vs 84.5 %) and matches it on mathematical reasoning tasks such as AIME 2025 and HMMT 2025. Only in certain heavy-mode configurations—where GPT-5 aggregates multiple trajectories—does the proprietary model regain parity.That Moonshot’s fully open-weight release can meet or exceed GPT-5’s scores marks a turning point. The gap between closed frontier systems and publicly available models has effectively collapsed for high-end reasoning and coding.<h3>Surpassing MiniMax-M2: The Previous Open-Source Benchmark</h3>When <a href="https://venturebeat.com/ai/minimax-m2-is-the-new-king-of-open-source-llms-especially-for-agentic-tool">VentureBeat profiled MiniMax-M2</a> just a week and a half ago, it was hailed as the “new king of open-source LLMs,” achieving top scores among open-weight systems:<ul><li>τ²-Bench 77.2</li><li>BrowseComp 44.0</li><li>FinSearchComp-global 65.5</li><li>SWE-Bench Verified 69.4</li></ul>Those results placed MiniMax-M2 near GPT-5-level capability in agentic tool use. Yet Kimi K2 Thinking now eclipses them by wide margins.Its BrowseComp result of 60.2 % exceeds M2’s 44.0 %, and its SWE-Bench Verified 71.3 % edges out M2’s 69.4 %. Even on financial-reasoning tasks such as FinSearchComp-T3 (47.4 %), K2 Thinking performs comparably while maintaining superior general-purpose reasoning.Technically, both models adopt sparse Mixture-of-Experts architectures for compute efficiency, but Moonshot’s network activates more experts and deploys advanced quantization-aware training (INT4 QAT). This design doubles inference speed relative to standard precision without degrading accuracy—critical for long “thinking-token” sessions reaching 256 k context windows.<h3>Agentic Reasoning and Tool Use</h3>K2 Thinking’s defining capability lies in its explicit reasoning trace. The model outputs an auxiliary field, reasoning_content, revealing intermediate logic before each final response. This transparency preserves coherence across long multi-turn tasks and multi-step tool calls.A reference implementation published by Moonshot demonstrates how the model autonomously conducts a “daily news report” workflow: invoking date and web-search tools, analyzing retrieved content, and composing structured output—all while maintaining internal reasoning state.This end-to-end autonomy enables the model to plan, search, execute, and synthesize evidence across hundreds of steps, mirroring the emerging class of “agentic AI” systems that operate with minimal supervision.<h3>Efficiency and Access</h3>Despite its trillion-parameter scale, K2 Thinking’s runtime cost remains modest. Moonshot lists usage at:<ul><li>$0.15 / 1 M tokens (cache hit)</li><li>$0.60 / 1 M tokens (cache miss)</li><li>$2.50 / 1 M tokens output</li></ul>These rates are competitive even against MiniMax-M2’s $0.30 input / $1.20 output pricing—and an order of magnitude below GPT-5 ($1.25 input / $10 output).<h3>Comparative Context: Open-Weight Acceleration</h3>The rapid succession of M2 and K2 Thinking illustrates how quickly open-source research is catching frontier systems. MiniMax-M2 demonstrated that open models could approach GPT-5-class agentic capability at a fraction of the compute cost. Moonshot has now advanced that frontier further, pushing open weights beyond parity into outright leadership.Both models rely on sparse activation for efficiency, but K2 Thinking’s higher activation count (32 B vs 10 B active parameters) yields stronger reasoning fidelity across domains. Its test-time scaling—expanding “thinking tokens” and tool-calling turns—provides measurable performance gains without retraining, a feature not yet observed in MiniMax-M2.<h3>Technical Outlook</h3>Moonshot reports that K2 Thinking supports native INT4 inference and 256 k-token contexts with minimal performance degradation. Its architecture integrates quantization, parallel trajectory aggregation (“heavy mode”), and Mixture-of-Experts routing tuned for reasoning tasks.In practice, these optimizations allow K2 Thinking to sustain complex planning loops—code compile–test–fix, search–analyze–summarize—over hundreds of tool calls. This capability underpins its superior results on BrowseComp and SWE-Bench, where reasoning continuity is decisive.<h3>Enormous Implications for the AI Ecosystem</h3>The convergence of open and closed models at the high end signals a structural shift in the AI landscape. Enterprises that once relied exclusively on proprietary APIs can now deploy open alternatives matching GPT-5-level reasoning while retaining full control of weights, data, and compliance.Moonshot’s open publication strategy follows the precedent set by DeepSeek R1, Qwen3, GLM-4.6 and MiniMax-M2 but extends it to full agentic reasoning. For academic and enterprise developers, K2 Thinking provides both transparency and interoperability—the ability to inspect reasoning traces and fine-tune performance for domain-specific agents.The arrival of K2 Thinking signals that Moonshot — a <a href="https://techcrunch.com/2024/02/21/moonshot-ai-funding-china/">young startup founded in 2023</a> with investment from some of China&#x27;s biggest apps and tech companies — is here to play in an intensifying competition, and comes amid growing scrutiny of the financial sustainability of AI’s largest players. Just a day ago, OpenAI CFO Sarah Friar sparked controversy after <a href="https://www.wsj.com/tech/ai/openai-isnt-yet-working-toward-an-ipo-cfo-says-58037472?gaa_at=eafs&amp;gaa_n=AWEtsqfPSJ91IQIG678OHyyBAwXqBLDfUb01-yZGSGU-uiYZq5Oudu3hFs6HuIdk-2A%3D&amp;gaa_ts=690cdeaf&amp;gaa_sig=vcfYpYv-Y0bP8WOkElUbVMCF-MAssgO9v2J34nzp5c7Mr0hs2x6WddR6gK12PW2xhfZiNCQrUUQCrRizxbFkMA%3D%3D">suggesting at WSJ Tech Live</a> event that the U.S. government might eventually need to provide a “backstop” for the company’s more than $1.4 trillion in compute and data-center commitments — a comment widely interpreted as a call for taxpayer-backed loan guarantees.Although <a href="https://www.cnbc.com/2025/11/06/openai-cfo-sarah-friar-says-company-is-not-seeking-government-backstop.html">Friar later clarified that OpenAI</a> was not seeking direct federal support, the episode reignited debate about the scale and concentration of AI capital spending. With OpenAI, Microsoft, Meta, and Google all racing to secure long-term chip supply, critics warn of an unsustainable investment bubble and “AI arms race” driven more by strategic fear than commercial returns — one that could &quot;blow up&quot; and take down the entire global economy with it if there is hesitation or market uncertainty, as so many trades and valuations have now been made in anticipation of continued hefty AI investment and massive returns. Against that backdrop, Moonshot AI’s and MiniMax’s open-weight releases put more pressure on U.S. proprietary AI firms and their backers to justify the size of the investments and paths to profitability. If an enterprise customer can just as easily get comparable or better performance from a free, open source Chinese AI model than they do with paid, proprietary AI solutions like OpenAI&#x27;s GPT-5, Anthropic&#x27;s Claude Sonnet 4.5, or Google&#x27;s Gemini 2.5 Pro — why would they continue paying to access the proprietary models? Already, Silicon Valley stalwarts like Airbnb have raised eyebrows for admitting to heavily <a href="https://www.scmp.com/tech/tech-trends/article/3329921/airbnb-picks-alibabas-qwen-over-chatgpt-win-chinese-open-source-ai">using Chinese open source alternatives like Alibaba&#x27;s Qwen over OpenAI&#x27;s proprietary offerings</a>. For investors and enterprises, these developments suggest that high-end AI capability is no longer synonymous with high-end capital expenditure. The most advanced reasoning systems may now come not from companies building gigascale data centers, but from research groups optimizing architectures and quantization for efficiency.In that sense, K2 Thinking’s benchmark dominance is not just a technical milestone—it’s a strategic one, arriving at a moment when the AI market’s biggest question has shifted from how powerful models can become to who can afford to sustain them.<h3>What It Means for Enterprises Going Forward</h3>Within weeks of MiniMax-M2’s ascent, Kimi K2 Thinking has overtaken it—along with GPT-5 and Claude 4.5—across nearly every reasoning and agentic benchmark. The model demonstrates that open-weight systems can now meet or surpass proprietary frontier models in both capability and efficiency.For the AI research community, K2 Thinking represents more than another open model: it is evidence that the frontier has become collaborative. The best-performing reasoning model available today is not a closed commercial product but an open-source system accessible to anyone.

Moonshot's Kimi K2 Thinking has emerged as a leading open source AI, surpassing established models like GPT-5 and Claude Sonnet 4.5 on key benchmarks. This development is significant as it highlights the growing competition in the AI sector, particularly from Chinese providers, and raises questions about the sustainability of OpenAI's high spending strategy. As the landscape evolves, the advancements in open source AI could democratize access to powerful technologies, benefiting a wider range of users and industries.

Moonshot's Kimi K2 Thinking emerges as leading open source AI, outperforming GPT-5, Claude Sonnet 4.5 on key benchmarks

<a href="https://cloud.google.com/?hl=en">Google Cloud</a> is introducing what it calls its most powerful artificial intelligence infrastructure to date, unveiling a seventh-generation <a href="https://cloud.google.com/tpu?hl=en">Tensor Processing Unit</a> and expanded <a href="https://cloud.google.com/discover/what-are-arm-based-processors?hl=en">Arm-based computing options</a> designed to meet surging demand for AI model deployment — what the company characterizes as a fundamental industry shift from training models to serving them to billions of users.The announcement, made Thursday, centers on <a href="https://cloud.google.com/resources/ironwood-tpu-interest?utm_source=cgc-blog&amp;utm_medium=blog&amp;utm_campaign=FY25-Q2-global-ENT33820-website-cs-ironwood-tpu-interest&amp;utm_content=ironwood_announcement_blog&amp;utm_term=ironwood&amp;hl=en">Ironwood</a>, Google&#x27;s latest custom AI accelerator chip, which will become generally available in the coming weeks. In a striking validation of the technology, <a href="https://www.anthropic.com/">Anthropic</a>, the AI safety company behind the Claude family of models, disclosed plans to access up to <a href="https://www.googlecloudpresscorner.com/2025-10-23-Anthropic-to-Expand-Use-of-Google-Cloud-TPUs-and-Services">one million of these TPU chips</a> — a commitment worth tens of billions of dollars and among the largest known AI infrastructure deals to date.The move underscores an intensifying competition among cloud providers to control the infrastructure layer powering artificial intelligence, even as questions mount about whether the industry can sustain its current pace of capital expenditure. Google&#x27;s approach — building custom silicon rather than relying solely on <a href="https://www.reuters.com/business/nvidia-poised-record-5-trillion-market-valuation-2025-10-29/">Nvidia&#x27;s dominant GPU chips</a> — amounts to a long-term bet that vertical integration from chip design through software will deliver superior economics and performance.<h3>Why companies are racing to serve AI models, not just train them</h3>Google executives framed the announcements around what they call &quot;the age of inference&quot; — a transition point where companies shift resources from training frontier AI models to deploying them in production applications serving millions or billions of requests daily.&quot;Today&#x27;s frontier models, including Google&#x27;s Gemini, Veo, and Imagen and Anthropic&#x27;s Claude train and serve on Tensor Processing Units,&quot; said Amin Vahdat, vice president and general manager of AI and Infrastructure at Google Cloud. &quot;For many organizations, the focus is shifting from training these models to powering useful, responsive interactions with them.&quot;This transition has profound implications for infrastructure requirements. Where training workloads can often tolerate batch processing and longer completion times, inference — the process of actually running a trained model to generate responses — demands consistently low latency, high throughput, and unwavering reliability. A chatbot that takes 30 seconds to respond, or a coding assistant that frequently times out, becomes unusable regardless of the underlying model&#x27;s capabilities.Agentic workflows — where AI systems take autonomous actions rather than simply responding to prompts — create particularly complex infrastructure challenges, requiring tight coordination between specialized AI accelerators and general-purpose computing.<h3>Inside Ironwood&#x27;s architecture: 9,216 chips working as one supercomputer</h3><a href="https://cloud.google.com/resources/ironwood-tpu-interest?utm_source=cgc-blog&amp;utm_medium=blog&amp;utm_campaign=FY25-Q2-global-ENT33820-website-cs-ironwood-tpu-interest&amp;utm_content=ironwood_announcement_blog&amp;utm_term=ironwood&amp;hl=en">Ironwood</a> is more than incremental improvement over Google&#x27;s sixth-generation TPUs. According to technical specifications shared by the company, it delivers more than four times better performance for both training and inference workloads compared to its predecessor — gains that Google attributes to a system-level co-design approach rather than simply increasing transistor counts.The architecture&#x27;s most striking feature is its scale. A single Ironwood &quot;pod&quot; — a tightly integrated unit of TPU chips functioning as one supercomputer — can connect up to 9,216 individual chips through Google&#x27;s proprietary <a href="https://blog.google/products/google-cloud/ironwood-tpu-age-of-inference/">Inter-Chip Interconnect network</a> operating at 9.6 terabits per second. To put that bandwidth in perspective, it&#x27;s roughly equivalent to downloading the entire Library of Congress in under two seconds.This massive interconnect fabric allows the 9,216 chips to share access to 1.77 petabytes of <a href="https://blog.google/products/google-cloud/ironwood-tpu-age-of-inference/">High Bandwidth Memory</a> — memory fast enough to keep pace with the chips&#x27; processing speeds. That&#x27;s approximately 40,000 high-definition Blu-ray movies&#x27; worth of working memory, instantly accessible by thousands of processors simultaneously. &quot;For context, that means Ironwood Pods can deliver 118x more FP8 ExaFLOPS versus the next closest competitor,&quot; Google stated in technical documentation.The system employs <a href="https://www.opencompute.org/projects/optical-circuit-switching">Optical Circuit Switching</a> technology that acts as a &quot;dynamic, reconfigurable fabric.&quot; When individual components fail or require maintenance — inevitable at this scale — the OCS technology automatically reroutes data traffic around the interruption within milliseconds, allowing workloads to continue running without user-visible disruption.This reliability focus reflects lessons learned from deploying five previous TPU generations. Google reported that its fleet-wide uptime for liquid-cooled systems has maintained approximately 99.999% availability since 2020 — equivalent to less than six minutes of downtime per year.<h3>Anthropic&#x27;s billion-dollar bet validates Google&#x27;s custom silicon strategy</h3>Perhaps the most significant external validation of Ironwood&#x27;s capabilities comes from <a href="https://www.googlecloudpresscorner.com/2025-10-23-Anthropic-to-Expand-Use-of-Google-Cloud-TPUs-and-Services">Anthropic&#x27;s commitment to access up to one million TPU chips</a> — a staggering figure in an industry where even clusters of 10,000 to 50,000 accelerators are considered massive.&quot;Anthropic and Google have a longstanding partnership and this latest expansion will help us continue to grow the compute we need to define the frontier of AI,&quot; said Krishna Rao, Anthropic&#x27;s chief financial officer, in the official partnership agreement. &quot;Our customers — from Fortune 500 companies to AI-native startups — depend on Claude for their most important work, and this expanded capacity ensures we can meet our exponentially growing demand.&quot;According to a separate statement, Anthropic will have access to &quot;well over a gigawatt of capacity coming online in 2026&quot; — enough electricity to power a small city. The company specifically cited TPUs&#x27; &quot;price-performance and efficiency&quot; as key factors in the decision, along with &quot;existing experience in training and serving its models with TPUs.&quot;Industry analysts estimate that a commitment to access one million TPU chips, with associated infrastructure, networking, power, and cooling, likely represents a <a href="https://www.reuters.com/technology/anthropic-expand-use-google-clouds-tpu-chips-2025-10-23/">multi-year contract worth tens of billions of dollars</a> — among the largest known cloud infrastructure commitments in history.James Bradbury, Anthropic&#x27;s head of compute, elaborated on the inference focus: &quot;Ironwood&#x27;s improvements in both inference performance and training scalability will help us scale efficiently while maintaining the speed and reliability our customers expect.&quot;<h3>Google&#x27;s Axion processors target the computing workloads that make AI possible</h3>Alongside <a href="https://cloud.google.com/resources/ironwood-tpu-interest?utm_source=cgc-blog&amp;utm_medium=blog&amp;utm_campaign=FY25-Q2-global-ENT33820-website-cs-ironwood-tpu-interest&amp;utm_content=ironwood_announcement_blog&amp;utm_term=ironwood&amp;hl=en">Ironwood</a>, Google introduced expanded options for its <a href="https://cloud.google.com/products/axion?hl=en">Axion processor family</a> — custom Arm-based CPUs designed for general-purpose workloads that support AI applications but don&#x27;t require specialized accelerators.The <a href="http://forms.gle/HYY5FWRKewYuDMB27">N4A instance type</a>, now entering preview, targets what Google describes as &quot;microservices, containerized applications, open-source databases, batch, data analytics, development environments, experimentation, data preparation and web serving jobs that make AI applications possible.&quot; The company claims N4A delivers up to 2X better price-performance than comparable current-generation x86-based virtual machines.Google is also <a href="https://docs.google.com/forms/d/e/1FAIpQLSd14sMYz79SeRI665dM7lnUbsAg7zilVPdDfK2_6u1vBmiUfg/viewform?usp=send_form">previewing C4A metal</a>, its first bare-metal Arm instance, which provides dedicated physical servers for specialized workloads such as Android development, automotive systems, and software with strict licensing requirements.The Axion strategy reflects a growing conviction that the future of computing infrastructure requires both specialized AI accelerators and highly efficient general-purpose processors. While a TPU handles the computationally intensive task of running an AI model, Axion-class processors manage data ingestion, preprocessing, application logic, API serving, and countless other tasks in a modern AI application stack.Early customer results suggest the approach delivers measurable economic benefits. Vimeo reported observing &quot;a 30% improvement in performance for our core transcoding workload compared to comparable x86 VMs&quot; in initial N4A tests. ZoomInfo measured &quot;a 60% improvement in price-performance&quot; for data processing pipelines running on Java services, according to Sergei Koren, the company&#x27;s chief infrastructure architect.<h3>Software tools turn raw silicon performance into developer productivity</h3>Hardware performance means little if developers cannot easily harness it. Google emphasized that <a href="https://cloud.google.com/resources/ironwood-tpu-interest?utm_source=cgc-blog&amp;utm_medium=blog&amp;utm_campaign=FY25-Q2-global-ENT33820-website-cs-ironwood-tpu-interest&amp;utm_content=ironwood_announcement_blog&amp;utm_term=ironwood&amp;hl=en">Ironwood</a> and <a href="https://cloud.google.com/products/axion?hl=en">Axion</a> are integrated into what it calls <a href="https://cloud.google.com/solutions/ai-hypercomputer">AI Hypercomputer</a> — &quot;an integrated supercomputing system that brings together compute, networking, storage, and software to improve system-level performance and efficiency.&quot;According to an October 2025 IDC Business Value Snapshot study, AI Hypercomputer customers achieved on average 353% three-year return on investment, 28% lower IT costs, and 55% more efficient IT teams.Google disclosed several software enhancements designed to maximize Ironwood utilization. <a href="https://cloud.google.com/kubernetes-engine?utm_source=google&amp;utm_medium=cpc&amp;utm_campaign=na-US-all-en-dr-bkws-all-all-trial-e-dr-1710134&amp;utm_content=text-ad-none-any-DEV_c-CRE_772251321321-ADGP_Hybrid+%7C+BKWS+-+EXA+%7C+Txt-AppMod-GKE-Kubernetes+Engine-KWID_369526655975-kwd-369526655975&amp;utm_term=KW_google+kubernetes+engine-ST_google+kubernetes+engine&amp;gclsrc=aw.ds&amp;gad_source=1&amp;gad_campaignid=23052915519&amp;gclid=Cj0KCQiAiKzIBhCOARIsAKpKLAMMFmNaZgmWnQ3CYrziXElfmMXmQphYqoSvICvf6jfUjLqR9XqFt3oaArkYEALw_wcB&amp;hl=en">Google Kubernetes Engine</a> now offers advanced maintenance and topology awareness for TPU clusters, enabling intelligent scheduling and highly resilient deployments. The company&#x27;s <a href="https://github.com/AI-Hypercomputer/maxtext">open-source MaxText framework</a> now supports advanced training techniques including Supervised Fine-Tuning and Generative Reinforcement Policy Optimization.Perhaps most significant for production deployments, Google&#x27;s <a href="https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-gke-inference-gateway">Inference Gateway</a> intelligently load-balances requests across model servers to optimize critical metrics. According to Google, it can reduce time-to-first-token latency by 96% and serving costs by up to 30% through techniques like prefix-cache-aware routing.The <a href="https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-gke-inference-gateway">Inference Gateway</a> monitors key metrics including KV cache hits, GPU or TPU utilization, and request queue length, then routes incoming requests to the optimal replica. For conversational AI applications where multiple requests might share context, routing requests with shared prefixes to the same server instance can dramatically reduce redundant computation.<h3>The hidden challenge: powering and cooling one-megawatt server racks</h3>Behind these announcements lies a massive physical infrastructure challenge that Google addressed at the recent <a href="https://www.opencompute.org/summit/emea-summit">Open Compute Project EMEA Summit</a>. The company disclosed that it&#x27;s implementing +/-400 volt direct current power delivery capable of supporting up to one megawatt per rack — a tenfold increase from typical deployments.&quot;The AI era requires even greater power delivery capabilities,&quot; explained Madhusudan Iyengar and Amber Huffman, Google principal engineers, in an <a href="https://cloud.google.com/blog/topics/systems/enabling-1-mw-it-racks-and-liquid-cooling-at-ocp-emea-summit">April 2025 blog post</a>. &quot;ML will require more than 500 kW per IT rack before 2030.&quot;Google is collaborating with Meta and Microsoft to standardize electrical and mechanical interfaces for high-voltage DC distribution. The company selected <a href="https://www.opencompute.org/files/OCP18-400VDC-Efficiency-02.pdf">400 VDC</a> specifically to leverage the supply chain established by electric vehicles, &quot;for greater economies of scale, more efficient manufacturing, and improved quality and scale.&quot;On cooling, Google revealed it will contribute its fifth-generation cooling distribution unit design to the Open Compute Project. The company has deployed liquid cooling &quot;at GigaWatt scale across more than 2,000 TPU Pods in the past seven years&quot; with fleet-wide availability of approximately 99.999%.Water can transport approximately 4,000 times more heat per unit volume than air for a given temperature change — critical as individual AI accelerator chips increasingly dissipate 1,000 watts or more.<h3>Custom silicon gambit challenges Nvidia&#x27;s AI accelerator dominance</h3>Google&#x27;s announcements come as the AI infrastructure market reaches an inflection point. While Nvidia maintains overwhelming dominance in AI accelerators — holding an estimated 80-95% market share — cloud providers are increasingly investing in custom silicon to differentiate their offerings and improve unit economics.Amazon Web Services pioneered this approach with <a href="https://aws.amazon.com/ec2/graviton/">Graviton Arm-based CPUs</a> and <a href="https://aws.amazon.com/ai/machine-learning/inferentia/">Inferentia</a> / <a href="https://aws.amazon.com/ai/machine-learning/trainium/">Trainium</a> AI chips. Microsoft has developed <a href="https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/cobalt-overview">Cobalt processors</a> and is reportedly working on AI accelerators. Google now offers the most comprehensive custom silicon portfolio among major cloud providers.The strategy faces inherent challenges. Custom chip development requires enormous upfront investment — often billions of dollars. The software ecosystem for specialized accelerators lags behind Nvidia&#x27;s <a href="https://developer.nvidia.com/about-cuda">CUDA platform</a>, which benefits from 15+ years of developer tools. And rapid AI model architecture evolution creates risk that custom silicon optimized for today&#x27;s models becomes less relevant as new techniques emerge.Yet Google argues its approach delivers unique advantages. &quot;This is how we built the first TPU ten years ago, which in turn unlocked the invention of the Transformer eight years ago — the very architecture that powers most of modern AI,&quot; the company noted, referring to the seminal <a href="https://arxiv.org/abs/1706.03762">&quot;Attention Is All You Need&quot; paper</a> from Google researchers in 2017.The argument is that tight integration — &quot;model research, software, and hardware development under one roof&quot; — enables optimizations impossible with off-the-shelf components.Beyond Anthropic, several other customers provided early feedback. Lightricks, which develops creative AI tools, reported that early Ironwood testing &quot;makes us highly enthusiastic&quot; about creating &quot;more nuanced, precise, and higher-fidelity image and video generation for our millions of global customers,&quot; said Yoav HaCohen, the company&#x27;s research director.Google&#x27;s announcements raise questions that will play out over coming quarters. Can the industry sustain current infrastructure spending, with major AI companies collectively committing hundreds of billions of dollars? Will custom silicon prove economically superior to Nvidia GPUs? How will model architectures evolve?For now, Google appears committed to a strategy that has defined the company for decades: building custom infrastructure to enable applications impossible on commodity hardware, then making that infrastructure available to customers who want similar capabilities without the capital investment.As the AI industry transitions from research labs to production deployments serving billions of users, that infrastructure layer — the silicon, software, networking, power, and cooling that make it all run — may prove as important as the models themselves.And if Anthropic&#x27;s willingness to commit to accessing up to one million chips is any indication, Google&#x27;s bet on custom silicon designed specifically for the age of inference may be paying off just as demand reaches its inflection point.

Google Cloud has just launched its most advanced AI infrastructure yet, featuring a seventh-generation Tensor Processing Unit that promises a fourfold performance boost. This innovation comes at a crucial time as demand for AI model deployment skyrockets. Additionally, Google has secured a significant deal with Anthropic worth billions, further solidifying its position in the competitive AI landscape. This development is important not only for Google but also for the broader tech industry, as it highlights the rapid advancements in AI technology and the increasing reliance on powerful computing solutions.

Google debuts AI chips with 4X performance boost, secures Anthropic megadeal worth billions

Presented by Salesforce <hr/>Vibe coding — the fast-growing trend of using generative AI to spin up code from plain-language prompts — is quick, creative, and great for instant prototypes. But many argue that it&#x27;s not cut out for building production-ready business apps with the security, governance, and trusted infrastructure that enterprises require. In other words, a few saved hours in development can mean a future full of security vulnerabilities, endless maintenance, and scalability headaches, says Mohith Shrivastava, principal developer advocate at Salesforce.&quot;For rapid experimentation, building minimum viable products, and tackling creative challenges, vibe coding is a game-changer,&quot; Shrivastava says. &quot;However, that same speed and improvisational nature are exactly what makes its application in a professional, enterprise setting a topic of intense debate. And the skepticism from the developer community is 100% justified.&quot;<h3>Risks and rewards of vibe coding </h3>The excitement is all about speed: going from a rough idea to a working prototype in hours, not weeks, is a massive advantage. But as Shrivastava shared, developers have been vocal about the potential downsides.&quot;When you apply vibe coding indiscriminately to an entire application stack, you’re not just moving fast; you’re accumulating risk at an unprecedented rate,&quot; Shrivastava explains. &quot;The cons are significant.&quot; That includes potential security nightmares, as AI models don&#x27;t typically take into consideration the company&#x27;s specific security policies. They can easily introduce vulnerabilities like hardcoded secrets or use insecure, hallucinated packages. Then there’s the issue of what Shrivastava calls &quot;spaghetti code on steroids,&quot; or verbose code that lacks a coherent architectural pattern, creating a mountain of technical debt.Equally concerning is the illusion of progress: vibe coding may complete 80% of a feature in record time, but the remaining 20% — the edge cases, performance tuning, and compliance work — becomes exponentially harder.But does this mean vibe coding has no place in the enterprise?&quot;The idea that you can just vibe your way to a complex, secure, and maintainable enterprise application is a dangerous fantasy,&quot; Shrivastava says. &quot;But — the pros are undeniable if it&#x27;s used correctly. The key is not to avoid vibe coding, but to apply it intelligently in your enterprise.&quot;<h3>Red and green zones: Enterprise-grade vibe coding</h3>You can&#x27;t, and you absolutely should not, vibe code your entire enterprise stack with just any generic tool, Shrivastava warns. But when paired with no-, low-, or pro-code tools that are built for the enterprise, many of the gaps can be addressed. An enterprise-grade vibe coding solution, for example, can automatically scan for security issues, flag performance bottlenecks, and provide a safety net. It’s also critical to understand which parts of an application suit this approach — and which demand a higher level of trust and control. Shrivastava divides the stack into red and green zones to illustrate.The green zone is the presentation layer, or the UI and UX. It’s ideal for vibe coding, where developers can move fast and iterate quickly without much risk. In contrast is the red zone, which covers the foundational pillars of an application, including business logic and data layers.<h3>Empowering developers in the green zone</h3>Developer expertise remains the foundation for effective and safe vibe coding. But developers can be amplified by AI tools and emerging agents that are grounded in business context, connected to real applications, integrations, and data flows.&quot;A generic AI agent can&#x27;t grasp your company&#x27;s unique processes, but a context-aware tool can act as a powerful pair programmer, helping a developer draft complex logic or model data with greater speed and accuracy,&quot; Shrivastava says. &quot;It’s about making the expert developer more efficient, not trying to do their job for them.&quot;Some areas will always be high risk for ungoverned AI — especially infrastructure and security. Letting a generic AI agent configure firewalls or Identity and Access Management [IAM] policies without oversight, Shrivastava warns, is a recipe for disaster. The solution isn’t to avoid the red zone entirely, but to approach it with the right tools — ones that embed governance, security, and context from the ground up.&quot;The winning strategy is clear: Vibe code the green zone for agility, approach the red zone by augmenting your developers with powerful, context-aware tools, and never, ever DIY your core infrastructure with AI,&quot; he says.<h3>Embracing enterprise vibe coding</h3>To harness the power of enterprise vibe coding, Salesforce developed <a href="https://developer.salesforce.com/blogs/2025/10/unleash-your-innovation-with-agentforce-vibes-vibe-coding-for-the-enterprise">Agentforce Vibes</a>. This new vibe coding offering for the enterprise includes Agentforce, an autonomous AI agent built to collaborate like a pair programmer on the Salesforce Platform. It’s designed precisely to provide developers with the right tools for the job, covering both the green and red zones. For the green zone, it offers the speed and agility to rapidly build UIs and prototypes. But its true power lies in how it augments developers in the red zone.&quot;Enterprise vibe coding like Agentforce lets organizations take AI-assisted development to the organizational level, accelerating coding, testing, and deployment, while ensuring consistency, security, and performance,&quot; says Dan Fernandez, VP of product, developer services at Salesforce. &quot;It&#x27;s not about throwing away governance for speed; it’s about integrating AI into every stage of the application lifecycle to work smarter.&quot;Because Agentforce Vibes’ tooling is deeply integrated with your business context on the platform, it can safely assist with business logic and data modeling. Most importantly, it operates on a trusted platform. Instead of a DIY approach — jury-rigging a generic AI agent to handle your networking — developers build on a foundation that has security and governance built in, so they can innovate safely, knowing the most critical layers of the stack are secure and compliant.<h3>Major enterprises are putting vibe coding to work </h3>Agentforce Vibes users are now tapping the tool to build around 20 to 25% of their new code base, according to Salesforce data, and users are accepting around 1.2 million lines of agentic code per month. That includes companies like Coinbase, CGI, Grupo Globo, and one of the top five banks in the U.S., which is using Agentforce Vibes capabilities to develop production-ready apps faster. Agentforce Vibes is part of a suite of tools in Agentforce 360 that span from no-code and low-code to pro-code development. These tools are together helping customers develop and deploy at speeds previously unheard of.With the low-code Agent Builder in Agentforce, the <a href="https://www.salesforce.com/customer-stories/secret-escapes/">Secret Escapes</a> team was able to build, test, and launch their agent to support customer service in just two weeks, compared to the six months it had previously taken the company to build and train a bot. With Agentforce, <a href="https://www.salesforce.com/customer-stories/1800-accountant/">1-800Accountant</a> autonomously resolved 70% of customer chat engagements during tax week in 2025, without writing a line of code, using Salesforce’s low-code tools and AI assistance. Meanwhile, media company <a href="https://www.salesforce.com/news/stories/agentforce-data-cloud-customer-use-cases/#:~:text=Media%20company%20Grupo%20Globo%20deployed%20agents%20to%20identify%20subscribers%20at%20risk%20of%20lapsing%2C%20offer%20personalized%20upgrades%2C%20cross%2Dsell%2C%20and%20convert%20non%2Dsubscribers.%20As%20a%20result%2C%20Agentforce%20boosted%20Globo%E2%80%99s%20retention%20rates%20by%2022%25%20in%20less%20than%20three%20months.">Grupo Globo</a> deployed agents to identify subscribers at risk of lapsing, offer personalized upgrades, cross-sell, and convert non-subscribers. As a result, Agentforce boosted Globo’s retention rates by 22% in less than three months.<h3>Innovation meets discipline </h3>Enterprise tools show that disciplined engineering and creative experimentation can coexist — and that balance, Shrivastava says, is the key to lasting innovation.&quot;Vibe coding is not a fad, but it&#x27;s also not a silver bullet that will replace disciplined software engineering,&quot; Shrivastava says. &quot;The smart path forward is a hybrid approach where human software skills are augmented with agentic intelligence. This balanced approach is how you get the best of both worlds: radical innovation at the edge and unwavering stability at the core.&quot;<hr/>Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact <a href="mailto:sales@venturebeat.com">sales@venturebeat.com</a>.

تعتبر برمجة Vibe، وهي اتجاه يستخدم الذكاء الاصطناعي التوليدي لإنشاء كود من تعليمات بسيطة، تكتسب شعبية بسبب سرعتها وإبداعها في إنشاء النماذج الأولية. ومع ذلك، يحذر الخبراء من أنها قد لا تلبي بعد المعايير الصارمة اللازمة للتطبيقات على مستوى المؤسسات، خاصة فيما يتعلق بالأمان والحوكمة. هذه المناقشة مهمة حيث توازن الشركات بين فوائد التطوير السريع والمخاطر المحتملة للثغرات والتحديات في الصيانة في برامجها.

El codificación Vibe, una tendencia que utiliza IA generativa para crear código a partir de simples indicaciones, está ganando popularidad por su rapidez y creatividad en la creación de prototipos. Sin embargo, los expertos advierten que aún puede no cumplir con los rigurosos estándares necesarios para aplicaciones a nivel empresarial, especialmente en lo que respecta a la seguridad y la gobernanza. Esta discusión es crucial a medida que las empresas sopesan los beneficios del desarrollo rápido frente a los riesgos potenciales de vulnerabilidades y desafíos de mantenimiento en su software.

Le codage Vibe, une tendance qui utilise l'IA générative pour créer du code à partir de simples instructions, gagne en popularité grâce à sa rapidité et sa créativité dans le prototypage. Cependant, des experts mettent en garde contre le fait qu'il ne répond pas encore aux normes rigoureuses nécessaires pour les applications au niveau des entreprises, en particulier en ce qui concerne la sécurité et la gouvernance. Cette discussion est cruciale alors que les entreprises pèsent les avantages d'un développement rapide contre les risques potentiels de vulnérabilités et de défis de maintenance dans leurs logiciels.

Vibe coding, a trend leveraging generative AI to create code from simple prompts, is gaining traction for its speed and creativity in prototyping. However, experts caution that it may not yet meet the rigorous standards needed for enterprise-level applications, particularly concerning security and governance. This discussion is crucial as businesses weigh the benefits of rapid development against the potential risks of vulnerabilities and maintenance challenges in their software.

HubSpot’s Dharmesh Shah on AI mastery: Why prompts, context, and experimentation matter most

Was this article worth reading? Share it