شهدت الأسبوع الماضي تقدمًا ملحوظًا في مجال الذكاء الاصطناعي مع إصدار Opus 4.5 وDeepSeek Math وFLUX.2، مما يبرز الابتكار المستمر في نماذج الذكاء الاصطناعي. هذه التطورات تسلط الضوء على التطور السريع لتقنيات الذكاء الاصطناعي وتطبيقاتها عبر مختلف القطاعات.

La semana pasada se produjeron avances significativos en inteligencia artificial con el lanzamiento de Opus 4.5, DeepSeek Math y FLUX.2, lo que demuestra la continua innovación en los modelos de IA. Estos desarrollos destacan la rápida evolución de las tecnologías de IA y sus aplicaciones en diversos sectores.

La semaine dernière a marqué des avancées significatives dans le domaine de l'intelligence artificielle avec la sortie d'Opus 4.5, DeepSeek Math et FLUX.2, mettant en avant l'innovation continue dans les modèles d'IA. Ces développements soulignent l'évolution rapide des technologies d'IA et leurs applications dans divers secteurs.

Last week marked significant advancements in artificial intelligence with the release of Opus 4.5, DeepSeek Math, and FLUX.2, showcasing the ongoing innovation in AI models. These developments highlight the rapid evolution of AI technologies and their applications across various sectors.

The Sequence Radar #763: Last Week AI Trifecta: Opus 4.5, DeepSeek Math, and FLUX.2

Google launches Gemini 3 & Nano Banana Pro, Anthropic releases Opus 4.5, and more!

أطلقت جوجل نموذجها الجديد للذكاء الاصطناعي، جمنّي 3، إلى جانب أداة جديدة لتوليد الصور، نانو بانانا برو، التي تستخدم قدرات جمنّي 3 لإنتاج صور مولدة بواسطة الذكاء الاصطناعي أكثر واقعية. يمثل هذا الإطلاق تقدمًا كبيرًا في تكنولوجيا الذكاء الاصطناعي لجوجل، مما يعزز جودة ونية توليد الصور للمستخدمين في جميع أنحاء العالم.

Google ha lanzado su último modelo de IA, Gemini 3, junto con la nueva herramienta de generación de imágenes, Nano Banana Pro, que utiliza las capacidades de Gemini 3 para producir imágenes generadas por IA más realistas. Este lanzamiento marca un avance significativo en la tecnología de IA de Google, mejorando la calidad y la intencionalidad de la generación de imágenes para usuarios de todo el mundo.

Google a lancé son dernier modèle d'IA, Gemini 3, accompagné du nouvel outil de génération d'images, Nano Banana Pro, qui utilise les capacités de Gemini 3 pour produire des images générées par IA plus réalistes. Ce lancement marque une avancée significative dans la technologie IA de Google, améliorant la qualité et l'intentionnalité de la génération d'images pour les utilisateurs du monde entier.

Google has launched its latest AI model, Gemini 3, alongside the new image generation tool, Nano Banana Pro, which utilizes Gemini 3's capabilities to produce more realistic AI-generated images. This launch marks a significant advancement in Google's AI technology, enhancing the quality and intentionality of image generation for users worldwide.

LWiAI Podcast #226 - Gemini 3, Claude Opus 4.5, Nano Banana Pro, LeJEPA

Agent memory remains a problem that enterprises want to fix, as agents forget some instructions or conversations the longer they run. <a href="https://www.anthropic.com/">Anthropic</a> believes it has solved this issue for its <a href="https://venturebeat.com/ai/anthropics-new-claude-can-code-for-30-hours-think-of-it-as-your-ai-coworker">Claude Agent SDK</a>, developing a two-fold solution that allows an agent to work across different context windows.“The core challenge of long-running agents is that they must work in discrete sessions, and each new session begins with no memory of what came before,” Anthropic wrote in <a href="https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents">a blog post</a>. “Because context windows are limited, and because most complex projects cannot be completed within a single window, agents need a way to bridge the gap between coding sessions.”Anthropic engineers proposed a two-fold approach for its Agent SDK: An initializer agent to set up the environment, and a coding agent to make incremental progress in each session and leave artifacts for the next.  <h2>The agent memory problem</h2>Since agents are built on foundation models, they remain constrained by the limited, although continually growing, context windows. For long-running agents, this could create a larger problem, leading the agent to forget instructions and behave abnormally while performing a task. <a href="https://venturebeat.com/ai/enhancing-ai-agents-with-long-term-memory-insights-into-langmem-sdk-memobase-and-the-a-mem-framework">Enhancing agent memory</a> becomes essential for consistent, business-safe performance. Several methods emerged over the past year, all attempting to bridge the gap between context windows and agent memory. <a href="https://www.langchain.com/">LangChain</a>’s LangMem SDK, <a href="https://www.memobase.io/">Memobase</a> and <a href="https://openai.com/">OpenAI</a>’s Swarm are examples of companies offering memory solutions. Research on agentic memory has also exploded recently, with proposed <a href="https://venturebeat.com/ai/how-procedural-memory-can-cut-the-cost-and-complexity-of-ai-agents">frameworks like Memp</a> and the <a href="https://venturebeat.com/ai/googles-nested-learning-paradigm-could-solve-ais-memory-and-continual">Nested Learning Paradigm</a> from <a href="https://www.google.com/">Google</a> offering new alternatives to enhance memory. Many of the current memory frameworks are open source and can ideally adapt to different large language models (LLMs) powering agents. Anthropic’s approach improves its Claude Agent SDK. <h2>How it works</h2>Anthropic identified that even though the Claude Agent SDK had context management capabilities and “should be possible for an agent to continue to do useful work for an arbitrarily long time,” it was not sufficient. The company said in its blog post that a model <a href="https://venturebeat.com/ai/anthropics-claude-opus-4-5-is-here-cheaper-ai-infinite-chats-and-coding">like Opus 4.5</a> running the Claude Agent SDK can “fall short of building a production-quality web app if it’s only given a high-level prompt, such as &#x27;build a clone of claude.ai.&#x27;” The failures manifested in two patterns, Anthropic said. First, the agent tried to do too much, causing the model to run out of context in the middle. The agent then has to guess what happened and cannot pass clear instructions to the next agent. The second failure occurs later on, after some features have already been built. The agent sees progress has been made and just declares the job done. Anthropic researchers broke down the solution: Setting up an initial environment to lay the foundation for features and prompting each agent to make incremental progress towards a goal, while still leaving a clean slate at the end. This is where the two-part solution of Anthropic&#x27;s agent comes in. The initializer agent sets up the environment, logging what agents have done and which files have been added. The coding agent will then ask models to make incremental progress and leave structured updates. “Inspiration for these practices came from knowing what effective software engineers do every day,” Anthropic said. The researchers said they added testing tools to the coding agent, improving its ability to identify and fix bugs that weren’t obvious from the code alone. <h2>Future research</h2>Anthropic noted that its approach is “one possible set of solutions in a long-running agent harness.” However, this is just the beginning stage of what could become a wider research area for many in the AI space. The company said its experiments to boost long-term memory for agents haven’t shown whether a single general-purpose coding agent works best across contexts or a multi-agent structure. Its demo also focused on full-stack web app development, so other experiments should focus on generalizing the results across different tasks.“It’s likely that some or all of these lessons can be applied to the types of long-running agentic tasks required in, for example, scientific research or financial modeling,” Anthropic said.

أعلنت شركة أنثروبيك عن إطلاق مجموعة أدوات وكيل كلود، التي تعالج مشكلة ذاكرة الوكلاء في أنظمة الذكاء الاصطناعي. تتيح هذه القدرة الجديدة على الجلسات المتعددة للوكلاء الاحتفاظ بالسياق عبر جلسات مختلفة، مما يعزز من وظيفتهم وفائدتهم في المهام المعقدة.

Anthropic ha anunciado el lanzamiento del Claude Agent SDK, que aborda el problema de la memoria de los agentes en los sistemas de IA. Esta nueva capacidad de múltiples sesiones permite a los agentes retener el contexto a través de diferentes sesiones, mejorando su funcionalidad y utilidad en tareas complejas.

Anthropic a annoncé le lancement du Claude Agent SDK, qui résout le problème de mémoire des agents dans les systèmes d'IA. Cette nouvelle capacité multi-session permet aux agents de conserver le contexte entre différentes sessions, améliorant ainsi leur fonctionnalité et leur utilité dans des tâches complexes.

Anthropic has announced the release of the Claude Agent SDK, which addresses the long-standing issue of agent memory in AI systems. This new multi-session capability allows agents to retain context across different sessions, enhancing their functionality and usability in complex tasks.

The Sequence Radar #763: Last Week AI Trifecta: Opus 4.5, DeepSeek Math, and FLUX.2

Was this article worth reading? Share it

Metaflow AI

Alpha FX

Dynamiq