Welcome to the Age of Agency with GPT-5

Hi Futurist,

What happened in the last two weeks might not feel like a revolution to you. But it is. History rarely announces itself. It whispers. It blends into your inbox, your meetings, your tools. Paradigm shifts don’t arrive with drumrolls, they show up as product updates. You don’t feel them. You scroll past them. That’s exactly what happened. We’re living through something our grandkids will study in history class, and to us it feels like just another Thursday. You argue in Slack about prompt formats while the nature of work is quietly rewritten beneath your feet. The coffee still burns your tongue. GPT-5, Genie 3 or Shopify’s Commerce Agents show that AI doesn’t just help you work, it changes what you’re capable of doing. Therefore, I am sending you insights, inspiration, and innovation straight to your inbox. Let’s dive into the depths of the digital future together and discover the waves of change shaping our industry.

💡 In this post, we're dipping in:

• 📣 Byte-Sized Breakthroughs: GPT‑5 doesn’t just help. It thinks. A true extension of human cognition, reshaping how we work, lead and live. Claude 4.1 sharpens its code skills, Genie 3 builds worlds from words, and OpenAI drops open models for everyone. Lindy 3.0 makes agents useful at last, and Shopify turns shopping into chatting.

• 🎙️ MarTech Maestros: OpenAI’s Sam Altman talks superintelligence with Cleo Abram. Why build something that might outsmart us? A candid look at GPT‑5, power, and the future of work, truth and society.

• 🧐 In Case You Missed It: Apple builds its own answer engine, Grok goes global, and Tencent floods the field with tiny LLMs. From Claude to Cursor and from agents to avatars, everyone wants a bite of the AI pie.

Do you have tips, feedback or ideas? Or just want to give your opinion? Feel free to share it at the bottom of this dip. That's what I'm looking for.

One quick favor 🙏
I’m thinking of tweaking the rhythm of Digital Dips. Just a small change, but I’d love your input. Mind clicking one of the options in the poll below?

No time to read? Listen to this episode of Digital Dips on Spotify and stay updated while you’re on the move. The link to the podcast is only available to subscribers. If you haven’t subscribe already, I recommend to do so.

Quick highlights of the latest technological developments.

Headstory: Welcome to the Age of Agency with GPT-5

A few weeks back I posted a comment on the LinkedIn post of my former teacher from high school. He writes opinion columns for a newspaper in our region. His bottom line was simple: “AI makes us less human.” I argued the opposite. Because the real question isn’t whether AI makes us less human. It’s whether we’ll use the time it gives us back to become more human.

Let me explain this first. Humans don’t run from tools. We run through them. We’ve always evolved through tools. Fire. Steam. Electricity. And now AI. Not because we planned it, but because it changed what we could do and what we believed we were meant to do. We used to hunt. Then we farmed. Then we worked. Why? Because the tools we invented and used made it possible. And somewhere along the way, we started measuring life in hours sold and meetings held. But AI cracks that open. Tools extend us. AI is just the next extension.

You don’t really feel paradigm shifts while you’re in them. You don’t notice history being written when you’re just trying to get your inbox to zero. We’re living through a moment future generations will call historic. But to us, it feels like just another Thursday. Another rollout. Another tab in the browser. GPT-5 writes strategy docs better than your team’s smartest consultant and we complain it forgot a bullet point. Midjourney generates campaign visuals in seconds, and we nitpick the shade of blue. We scroll past breakthroughs while waiting for our Uber.

To me, GPT-5 is the first widely used example of AI as an extension. Just like fire redefined survival, and electricity redefined society, GPT-5 redefines cognition. It extends thinking at scale. I’ll explain.

First things first. The GPT-5 rollout didn’t land with fireworks. It landed with complaints. “Colder than 4o.” “Worse at feelings.” “Smarter but less fun.” And yet, somehow, by the end of the week, everyone was using it. Because under the hood, it quietly redefined the game.

So, what’s actually new? GPT-5 comes with a few smart upgrades under the hood. First, there’s a router, it quickly decides how hard a question is and picks the best way to answer. Then there’s a verifier, which is like a built-in fact-checker that double-checks the answer before showing it to you. And instead of just one big model doing everything, GPT-5 has different versions for different tasks. Some are fast and simple, others think harder and slower for tricky stuff. We’ve gone from autocomplete to autonomous reasoning. From chat to cognition.

After using GPT-5 for a week straight, I’ll tell you what I’ve noticed. GPT-5 remembers everything. Sometimes too much. I’ve noticed it recalling parts of past chats even when I start a new one, as if I’m continuing a conversation I didn’t ask to continue. Maybe that’s helpful for some. But in my case, it’s more annoying than smart. That being said, I find the answers of GPT-5 better. Period. It follows instructions more accurately. It hallucinates less. It gets what you mean, faster and better. But here’s the twist: My old prompts stopped working as well. I was still using the same prompts from the GPT-4o days. The answers were weaker. Less on point. Less helpful. Why? Because GPT-5 doesn’t just sound smarter, it thinks differently. It reasons differently. It’s more independent. More active. And way more sensitive to vague instructions. It’s not a chatbot anymore. It’s a digital colleague.

And like any good colleague, it doesn’t thrive on wishy-washy directions. You don’t ask it to “think step-by-step.” You show it the steps. You define the output. You steer its behavior. That’s the real skill shift. GPT-5 isn’t about asking better questions, it’s about giving better directions. Structure matters more than style. Precision beats creativity. You have to lead it like you lead a team. Do that, and GPT-5 gives you more than any model before it. Don’t, and it becomes your biggest time-waster. To help you started, use my Custom GPT to write the perfect prompts for GPT-5. You can find it here.

Now, let’s zoom out for a second. OpenAI is clearly building toward a dedicated hardware device, a personal AI supercomputer that fits in your pocket. Always with you. Always thinking. Always on. No wonder OpenAI teamed up with Jony Ive, the mind behind the iPhone. You might remember that from a previous Dip. I believe GPT-5 is the bridge to that future. With its routers, verifiers, and fast-switch logic, GPT-5 is the operating system for it. They’re building an AI that wraps itself around you. Around what you like. What you care about. How you work. What your day looks like. A system that remembers everything you’ve ever interacted with, and reasons across all of it with intelligence that far exceeds the average user.

GPT-5 doesn’t just answer questions faster or better. It takes work off your plate entirely. Writing, coding, planning, analyzing, summarizing. Things that used to take hours now happen in seconds. Whether you're a teacher in Rotterdam, a startup founder in Nairobi, or a community builder in Medellín, GPT-5 meets you where you are, and helps you move forward. And when the basics are handled by a system smarter than most people you know, what’s left is time. Time to think. To create. To connect. Time to be human. That’s the shift. It doesn’t just boost productivity, it creates space. And what we choose to do with that space will define what kind of humans we become. That’s what I meant when I said: the real question isn’t whether AI makes us less human. It’s whether we use the time it gives us back to become more human.

For businesses, leaders need to stop asking, “What can AI do?” and start asking, “What will humans still need to do?” When a $20 model knows more than your best analyst, when reasoning becomes a commodity, when speed and scale are no longer your advantage; how do you compete? What jobs become irrelevant? What workflows become real-time? What happens when customers expect answers not in hours, but seconds?

We’ve officially moved beyond assistance. We’ve entered the age of agency. The cost of intelligence just collapsed. The value of human insight just skyrocketed. And the companies that win will be the ones who figure out how to combine the two. If your strategy hasn’t changed in the past 90 days, you don’t have one.

GPT-5 is not just a better model. It’s a mirror. It shows us what’s possible. And asks us what’s next.

Claude Opus 4.1 gets sharper on code and agents

TL;DR

Anthropic ships Claude Opus 4.1 with upgrades in agentic tasks, real-world coding, and reasoning. It posts 74.5% on SWE-bench Verified, and shows steadier detail tracking in research and data work. GitHub and Rakuten report that this version performs more reliably and precisely across large codebases. Opus 4.1 is now available to paid users, and via API and cloud platforms.

Read it yourself?

Sentiment

Some users swear by Claude and are excited about the (seemingly) “small” improvements of Claude Opus 4.1 over version 4. Others laugh at it. Especially with the release of GPT-5, which outperforms Claude in most benchmarks. Still, some online users report that GPT-5’s Thinking mode often needed multiple tries to solve a problem, while Claude Opus 4.1 got it right in one shot. That said, GPT-5 is cheaper, so there’s more room for error.

My thoughts

To me, it’s still a bit of a mystery why Anthropic decided to launch Opus 4.1 in the same week as GPT-5. And yes, a 2% jump on the SWE benchmark might look minor. But it’s not. That 2% could mean the difference between working code or hours of debugging. And that matters. Even though Anthropic probably knew GPT-5 would come in cheaper, Claude Opus 4.1 remains a favorite among developers. ChatGPT is mostly for people. Claude is often the engine behind AI-driven tools where code, agentic planning and longer context really matter. In that light, these updates make perfect sense.

Google’s Genie 3 creates worlds in real-time

TL;DR

Google DeepMind’s Genie 3 can generate rich, interactive environments based on text prompts at 24 fps, in 720p. It simulates realistic physics, supports long interactions, and stays visually consistent. This makes it possible to walk through a volcanic wasteland, a serene zen garden, or a portal to a desert. All in real-time, and all AI-generated.

Read it yourself?

Sentiment

People online compare it to game engines, Ready Player One, and the metaverse. They’re impressed by how much more realistic Genie 3 looks compared to Genie 2, and that only launched six months ago. Many recognize real world models as a key piece in the AGI puzzle, because they allow AI to “live” in simulated versions of our world. Elon Musk talked about this on X too, saying Tesla uses similar models to train self-driving cars. Simulated environments = infinite training data.

My thoughts

Interactive worlds this rich and responsive are more than eye candy, they’re the perfect training ground for AI. It gives agents a way to explore, act, and learn about the real world. They now can learn from experiences. Just like we do. Think of entire worlds with functioning economies, characters that remember you and rules that stay consistent. For humans, it offers fully immersive environments, built in real time, straight from a text prompt. The line between simulation and reality starts to blur. In the future, some of these places might feel more real than the world around us. Some people might even prefer the world they just prompted into over the one they live in. Some may even struggle with that. But world models like Genie 3 bring us closer to AGI. Because it gives AI a sandbox to learn in. Just like a child learns by moving through the world, interacting with objects, making mistakes, and figuring things out, AI need something similar. They can’t really understand the world just by reading text. And because these worlds are consistent, persistent, and interactive, with memory and complex systems, it becomes possible to train AI in situations that mimic real life at scale. That’s exactly the kind of learning AGI needs. So yes, this is a serious step in that direction and also, just very cool.

More byte-sized breakthroughs:

• OpenAI releases powerful open models
  OpenAI just dropped two open-weight models: gpt-oss-120b and 20b. They run on everyday hardware, follow instructions well, and match the o3-mini and o4-mini on most tasks. They’re built for real-world use, with tool support, local inference, and solid safety features baked in. You can fine-tune, deploy and even run them on consumer hardware. But while it’s a big step for open models in the West, Chinese labs are still ahead.

• Lindy 3.0 makes AI agents actually useful
  Lindy’s Agent Builder lets anyone “vibe code” their own AI worker in minutes. No technical know-how. Just a prompt. Then, Autopilot gives these agents their own cloud computer, so they can open tabs, click buttons, fill in forms, and actually get work done. Just like a person would. And with Team Accounts, you can roll them out across your company in a few clicks. Lindy 3.0 is what AI agents should’ve been all along.

• Shopify turns AI agents into personal shoppers

  Shopify's new tools make it easy to build AI agents that can help people shop, directly inside a chat. Checkout Kit adds instant buying to any conversation. Catalog gives global access to millions of products. And Universal Cart lets people buy from any store, all in one place. Just commerce that fits where people already are. Shopify’s making buying feel like talking to a friend who knows what you want.

A must-see webinar, podcast, or article that’s too good to miss.

The race to build superintelligence

Cleo Abram sits down with OpenAI’s CEO Sam Altman to unpack the big questions behind GPT-5. Why build something that might one day outsmart us? What does it mean for truth, science and our jobs? In this wide-ranging conversation, Altman reveals what it can do, why the stakes are so high, and how far we might go. Billions of dollars, global competition, and the fate of work, science, and society are all in play. It’s part forecast, part philosophy and an attempt to understand the mind behind the machine that could reshape everything.

A roundup of updates that are too cheesy to ignore.

• Apple forms a new team to develop a ChatGPT-style AI 'answer engine' for its ecosystem.

• Grok 4 is now free globally, offering expansive access with Auto and Expert modes.

• Grok launches Imagine: an AI Vine reviving your favorite snippets for sharing.

• Grok Imagine by xAI lets you craft NSFW AI-generated images and videos, xAI positions Grok as an unfiltered, boundary-pushing AI.

• Cohere unveils North, an AI platform empowering enterprises to deploy secure, scalable AI agents and automations.

• XBai o4 debuts, surpassing OpenAI−o3−mini with enhanced open-source scaling.

• Hailuo's new Agent Templates empower creativity with Editable Storyboards and speedy Image-to-Video tools.

• Tencent Hunyuan launches four compact open-source LLMs for low-power AI across devices.

• Tencent Hunyuan-GameCraft debuts open-source video generation, turning images into game scenes with user actions.

• Tencent rolls out Yan: a cutting-edge tool for generating interactive videos.

• LTX Studio introduces Sessions for seamless asset organization in Gen Space.

• LTX Studio debuts multi-reference, letting you seamlessly blend elements from varied images with precision.

• n8n simplifies multi-agent workflows with its new AI Agent Tool node, optimizing costs and debugging.

• Qwen-Image launches as an open-source powerhouse for text-to-image magic, excelling in bilingual and complex designs.

• Descript debuts Control Room, letting Producers and Co-Hosts manage live sessions incognito.

• Replit Agent introduces object storage, effortlessly saving all your media files on demand.

• Google's Gemma 3 270M delivers hyper-efficient AI with strong task-specific capabilities for on-device and research roles.

• Google’s Gemini lets you craft personalized storybooks with illustrations and read-aloud features.

• Google’s Gemini App turbocharges study with AI-driven Guided Learning, visuals, and free access to Google AI Pro.

• Google’s Gemini CLI GitHub Actions debuts as a free AI partner, streamlining coding tasks and reviews.

• Google’s Gemini CLI enhances VS Code with smart suggestions and seamless in-editor diffing.

• Google Gemini now learns from your past chats for tailored conversations, starting with 2.5 Pro users.

• Google is ready to roll out AI Mode ads by giving ad agencies and brands more information about how the new channel differs from traditional search

• Elevenlabs’ Music debuts an AI model giving you full control over multilingual songs and styles.

• ElevenLabs Studio debuts Video-to-Music flow, crafting custom soundtracks with a single click.

• Higgsfield AI UPSCALE, powered by Topaz Labs, transforms your blurry photos and videos into PRO-grade 4K.

• Higgsfield AI Draw-to-Video transforms your sketches into cinematic masterpieces using top video models.

• Higgsfield AI releases Product-to-Video for effortless, prompt-free perfect product imagery.

• LeonardoAI’s Lucid Origin launches as an all-in-one creative tool, offering Full HD, diverse styles, and precise text rendering.

• Ideogram Character unveils face swap templates, letting you slip into any scene or meme with Magic Fill.

• Ideogram unleashes character consistency in the API, no costly training required.

• JetBrain’s launches an AI platform for no-code web apps, perfect for creative project makers called Kineto.

• Morphic now lets you create 360° character turnarounds for dynamic visual presentations.

• OpenAI's reasoning system clinches gold at the 2025 IOI, leading AI participants.

• OpenAI partners with the US government to deploy ChatGPT Enterprise across the federal executive branch for just $1 per agency.

• Gamma let you turn any Notion page into a sleek presentation, directly into Gamma through their "Import from URL" flow.

• Notion AI steps up, managing complex tasks across your workspace like a pro by editing multiple pages or updating entire databases.

• Anthropic’s Claude Code automates security reviews with new /security-review command and GitHub Actions integration.

• Anthropic’s Claude Code introduces /output-style for personalized communication flair.

• Anthropic’s Claude now recalls past chats, letting you seamlessly resume conversations.

• Anthropic’s Claude Sonnet 4 expands Anthropic API context to 1 million tokens for code and document processing.

• Midjourney unveils HD Video mode for Pro and Mega users, delivering quadruple the pixels of SD.

• Payy Card launches a visa card for stablecoins, with built-in privacy. Your onchain activity and balance is shielded with ZK proofs.

• Arc launches a Layer-1 blockchain tailored for stablecoins, featuring USDC gas and high-speed transactions.

• Gitbook launches an AI Assistant where you can create tailored docs experiences for every user using external sources via MCP.

• Cursor lands in your terminal. Seamlessly switch between CLI and editor with early beta access.

• Cursor CLI unveils new MCPs, Review Mode, and smart UX features for streamlined coding.

• GLM-4.5V sets new standards in open-source visual reasoning with its 106B-parameter powerhouse.

• Suno Studio announces their newest updates with multi-track creation and MIDI export features.

• Pika Labs unveils a lightning-fast AI video model, delivering HD content in 6 seconds flat.

• Perplexity unveils video generation on web and mobile, letting Pro users craft 5 videos and Max users 15 monthly.

• Perplexity makes a bold $34.5 billion bid to acquire Google's Chrome browser.

• Jan-v1 launches as an open-source web search powerhouse, outshining Perplexity Pro in QA accuracy.

• Groq unveils Code CLI, a customizable template for developers to craft their perfect command line interface.

• Mistral Medium 3.1 powers up Le Chat with performance boosts, tone refinements, and smarter searches.

• Skywork AI Matrix-Game 2.0 rolls out real-time interactive world models, now fully open-source at 25FPS.

• Skywork Matrix-3D converts images or text prompts into expansive, explorable 3D worlds.

• FireCrawl launches an open-source AI web app builder called ‘Open Lovable’ to clone and edit websites instantly.

• Spielwerk launches TikTok-style app for endless mini games, compete with friends and remix favorites.

• Inworld Runtime debuts as the auto-scaling AI solution for consumer apps, boosting MLOps and experiments.

• Vivo’s Vision mixed reality headset launches on August 21, enhancing immersive tech experiences.

• HTC unveils VIVE Eagle smart glasses with AI assistant for snapping photos and managing your day.

• Mule Run opens the first AI Agent marketplace, offering tools that game, code, and earn for you.

• Genspark AI launches AI Developer tool for coding novices with Claude Sonnet 4 and GPT-5.

• Orchids Editor debuts as the ultimate visual tool for crafting AI-powered apps and websites.

• Bolt, Netlify and Supabase join forces to launch a robust platform for building and scale to million users vibe coding.

• ByteDance releases a free, open-source AI agent that automates desktop tasks using local vision models.

• SkyReels A3 introduces the next generation of interactive talking avatars.

This was it. Our thirty-ninth digital dip together. It might seem like a lot, but remember; this wasn't even everything that happened in the past few weeks. This was just a fraction.

If you want your leadership team to turn AI from a shiny tool into a true strategic lever, I can help. From redefining workflows to sharpening decision frameworks, from building AI literacy in your teams to embedding human advantage at scale. Just reply to this email or connect with me on LinkedIn.

Looking forward to what tomorrow brings! ▽

-Wesley