One insight from working at @OpenAI in mid-October 2026: Our dots are going pretty wild... yesterday, the two met for the first time. 4 weeks ago, I told people that the way I work had completely changed when I started labeling/naming my threads in @ChatGPT Work to keep the overview. After DevDay, it changed again. This time, the impact is even stronger. Dots are a complete game-changer for me, and I'm hearing this from my Enterprise customers as well. To all ChatGPT Enterprise and Pro users who are already using dots: let me know how you like them and what your top use cases are. To all others: fingers crossed that you can experience their magic soon. I wonder what's going to be the next big shift in a few weeks. Keep on shipping @thsottiaux @steipete

A little weekend humour from @PeterJ_Walker. Apparently, @OpenAI is leading in model numbering too 😉😄 Beyond model numbering, let’s look at some of the actual reasons why customers choose @OpenAI models and harnesses to build with. ✓ Consistently very strong model efficiency (cost per task): GPT models are often 𝐦𝐮𝐥𝐭𝐢𝐩𝐥𝐞 𝐭𝐢𝐦𝐞𝐬 more economical compared to other frontier models. ✓ Benefits in speed: get more tasks done. This also helps you stay focused. ✓ GPT-6 Astra provides a new level of intelligence, even beyond benchmarks. That said, I default to GPT-6.1 Sol for most of my tasks. Okay, and I hear that customers really 𝐞𝐧𝐣𝐨𝐲 working with @OpenAI. And then I hear quotes from some of the most sophisticated entrepreneurs and engineering leaders: "Since we switched to Astra, we doubled the number of features we ship". Thank you 🩵

Building a 3D representation of your apartment? That's quite simple now. It takes 20-30 minutes with @OpenAI's GPT-6 Astra. All you need a video and a prompt as input. That's it. P.S. You will also see the personalized piano, which I created. This has always been a dream. Now it doesn't just exist in my imagination but also inside Codex. And every time I think of it, I have an even clearer picture in my mind. What new things have you created in the last few days since the release of Astra? I'm very curious to hear about other ideas and workflows.

Who's saying that GPT 5.5 / Codex doesn't have humor🙂

🦞 Build AI literacy this weekend: 3 ways to learn and stay curious! 1️⃣ Build with OpenClaw! It’s super simple to get started. Ask ChatGPT or Codex, and as @sama posted, you can use your ChatGPT subscription. 2️⃣ Watch @Andrej Karpathy’s 30-minute lecture at Sequoia Capital, where he speaks about the next paradigm: AI becoming the runtime, interface, interpreter, worker, and reasoning layer for whole classes of products and services. 3️⃣ Listen to @gdb's recent podcast episode, where he speaks about the history of OpenAI, key moments, leadership, the mission, and what’s coming in AI over the next months. #AI #OpenAI #OpenClaw #AgenticAI #AILiteracy

9 @OpenAI announcements worth paying attention to and why they matter 1️⃣ Introducing GPT-5.4 mini and nano OpenAI launched GPT-5.4 mini and nano as faster, cheaper small models optimized for coding, subagents, tool use, and multimodal tasks, with mini available in the API, Codex, and ChatGPT, and nano available in the API. ➤ Powering production systems with agentic and coding workflows that depend on speed and cost. 2️⃣ OpenAI acquires TBPN OpenAI acquired the fast-growing tech media company and show. TBPN will keep editorial independence while joining OpenAI’s Strategy org. ➤ OpenAI can make a bigger impact by reaching more builders, businesses, and the broader public around AI. 3️⃣ Codex now offers pay-as-you-go pricing for teams OpenAI introduced Codex-only seats for ChatGPT Business and Enterprise with token-based billing and no fixed seat fee. ➤ Codex usage within Business and Enterprise has grown 6x since January, and we continue making it more accessible. 4️⃣ OpenAI raises $122 billion to accelerate the next phase of AI OpenAI closed a funding round with $122 billion in committed capital at an $852 billion post-money valuation. ➤ With strongly evolving model capabilities: compute, consumer adoption, enterprise deployment, and developer usage become a reinforcing flywheel. 5️⃣ Inside our approach to the Model Spec OpenAI published a deeper explanation of the Model Spec as our public framework for how models should behave, including instruction-following, conflict resolution, user freedom, and safety. ➤ For users, developers, and regulators, it gives a clearer basis for discussing what models should and should not do. 6️⃣ Introducing the OpenAI Safety Bug Bounty program OpenAI launched a public Safety Bug Bounty focused on AI abuse and safety risks, separate from the traditional security bug bounty. ➤ Public mechanism to surface those issues should improve resilience and give researchers a clearer path to engage. 7️⃣ Helping developers build safer AI experiences for teens OpenAI released prompt-based teen safety policies designed to work with our open-weight safety model, gpt-oss-safeguard, with input from external groups. ➤ OpenAI helps developers ship clearer, more standardized protections. 8️⃣ Powering Product Discovery in ChatGPT OpenAI is upgrading shopping and product discovery in ChatGPT with more visual product browsing and comparison, powered by the Agentic Commerce Protocol. ➤ ACP is becoming infrastructure for AI-native commerce. This is relevant for search, retail, and affiliate ecosystems. 9️⃣ Update on the OpenAI Foundation The OpenAI Foundation is beginning to deploy resources following recapitalization, tying its work to AGI benefiting humanity and to areas like life sciences, jobs, economic impact, AI resilience, and community support. ➤ The Foundation is one mechanism for channeling resources toward OpenAI’s mission to build AGI that benefits all of humanity. #OpenAI #AI #Codex #AgenticAI #ProductUpdates

The rule was simple: one task at a time. Then AI changed it. Here's where it gets interesting. As I've been using AI heavily for building, generating content, ideation, and strategizing, I noticed something. I started using more and more AI tools and agents simultaneously. Quickly extending to two screens, then three screens at once. And I discovered a new way of multitasking. Having these different agents open and visually separated, switching between them intuitively. Sometimes it brings me into this very special flow where I feel so enhanced. It's like conducting a concert, being the dirigent of this orchestra of AI agents, collaborating with them. It's not only rational and conscious. It goes deeper. It unleashes resources inside of me that become possible through this dance. With AI, we need to think twice about concepts that are part of our map of the world. Yes, focus is important. But it's also beautiful to have an orchestra of different flowers in a bouquet playing together. If we focus on them one at a time, holding together the bigger picture in our mind, then they can continue to evolve and bloom, creating something amazing together. Something predictable, making a vision become reality; sometimes someting unexpected. Being open, curious, and maybe a little nerdy to invest time into this pays off. My conviction: AI is here not to make us less human. It's here to enhance us, to make us more human, after all. What's your experience with AI and multitasking? Have you discovered new workflows that challenge conventional wisdom? hashtag#AIProductivity hashtag#FutureOfWork hashtag#AICollaboration hashtag#AIMindset hashtag#HumanAIPartnership

Check out my latest article: Why Your Biggest Critics Are Fighting Their Own Battles linkedin.com/pulse/why-your-… via @LinkedIn

What if AI advances so fast that your product becomes obsolete overnight? Barcelona to Munich. No internet. I started to reflect. At first, I worried. What if AI moves so fast that the app I’m building becomes irrelevant overnight? But as I thought deeper, I realized something: → Technology will always accelerate. → What matters is whether I deeply understand my users. If I do that, I won’t just keep up, I’ll stay ahead of the curve. But more importantly, I’ll be able to synergize with others who are doing the same. Together we can create amazing things. • Sometimes my product will lead. • Sometimes someone else’s product will. • Either way, co-creating keeps us aligned on the mission. • And that alignment makes progress possible. • There are many paths to reach it. Reflections like these became the foundation of my Masterclass “3 Capabilities and 9 Mindsets to Thrive in the AI Era.” Because I want others to begin those reflections, go deeper, and have similar realizations too. 👉 When AI does more, which human mindset becomes most valuable? Btw, if you haven’t checked it out yet, LuminaOS is still in beta, but you can already get access. Or check out the Masterclass I’m mentioning. Comment if you'd like to get access. hashtag#AI hashtag#Leadership hashtag#Mindset hashtag#Differentiation hashtag#ProductManagement

Check out my latest article: 7 Big AI Ideas That Will Transform How We Work in 2025 linkedin.com/pulse/7-big-ai-… via @LinkedIn

Anchoring our thinking and systems to design an agent-driven future🤓 Even if we're not all the way there yet, think of the infrastructure and tech stack that's going to be needed to get the most out of agents. To get into this "Agentic AI way" of designing and thinking myself, I started matching the agents that I'm personally building to frameworks; like this one, which was featured in yesterday's episode of the AI Daily Brief podcast. Consider these 7 𝐭𝐲𝐩𝐞𝐬 𝐨𝐟 𝐀𝐈 𝐚𝐠𝐞𝐧𝐭𝐬 based on a categorization by DigitalOcean. 🔄 𝐒𝐢𝐦𝐩𝐥𝐞 𝐑𝐞𝐟𝐥𝐞𝐱 𝐀𝐠𝐞𝐧𝐭𝐬 make decisions based solely on their current sensory input, responding immediately to environmental stimuli without needing memory or learning processes. Perfect for: Safety systems, fraud detection, automated responses Think: Emergency shutoffs, basic email filters, security alerts 🧠 𝐌𝐨𝐝𝐞𝐥-𝐁𝐚𝐬𝐞𝐝 𝐑𝐞𝐟𝐥𝐞𝐱 𝐀𝐠𝐞𝐧𝐭𝐬 are a more advanced form of intelligent agents designed to operate in partially observable environments, maintaining their own model of the world. Perfect for: Quality control, network monitoring, smart buildings Think: Manufacturing anomaly detection, IT system monitoring 🎯 𝐆𝐨𝐚𝐥-𝐁𝐚𝐬𝐞𝐝 𝐀𝐠𝐞𝐧𝐭𝐬 designed to pursue specific objectives by considering the future consequences of their actions. Perfect for: Project management, inventory optimization, logistics Think: Automated scheduling, supply chain planning, resource allocation 📈 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐀𝐠𝐞𝐧𝐭𝐬 A learning agent is an artificial intelligence system capable of improving its behavior over time by interacting with its environment and learning from its experiences. Perfect for: Personalization, process optimization, customer service Think: Recommendation engines, chatbots that improve over time ⚖️ 𝐔𝐭𝐢𝐥𝐢𝐭𝐲-𝐁𝐚𝐬𝐞𝐝 𝐀𝐠𝐞𝐧𝐭𝐬 makes decisions by evaluating the potential outcomes of its actions and choosing the one that maximizes overall utility. Perfect for: Trading, resource balancing, multi-objective optimization Think: Financial algorithms, energy management, scheduling conflicts 🏗️ 𝐇𝐢𝐞𝐫𝐚𝐫𝐜𝐡𝐢𝐜𝐚𝐥 𝐀𝐠𝐞𝐧𝐭𝐬 structured in a tiered system, where higher-level agents manage and direct the actions of lower-level agents. Perfect for: Complex manufacturing, multi-stage processes, enterprise systems Think: Smart factories, building automation, coordinated workflows 🤝 𝐌𝐮𝐥𝐭𝐢-𝐀𝐠𝐞𝐧𝐭 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 involves multiple autonomous agents interacting within a shared environment, working independently or cooperatively to achieve individual or collective goals. Perfect for: Distributed operations, competitive environments, research Think: Warehouse coordination, traffic optimization, collaborative analysis Interested in checking out my favorite tutorials? Write '𝐚𝐠𝐞𝐧𝐭𝐢𝐜' in the comments. #AIAgents #AgenticAI #Automation #DigitalTransformation #AI #MachineLearning #BusinessInnovation