The AI Act changed yesterday. Is your business ready?

Hi Futurist,

By the time you read this, there’s a good chance OpenAI already dropped GPT-5. Sam Altman wasn’t lying when he said we’ve crossed the event horizon. Liftoff has started. His quote I used in the last Dip. The last two weeks proves it. In fact, it’s starting to look like OpenAI has hit their own definition of AGI, especially now that Microsoft is in talks to secure exclusive access to their tech. And if you think that’s already newsworthy: Mark Zuckerberg just revealed that Meta now has self-learning AI systems, and ASI (Artificial Super Intelligence) is already on their radar.

Meanwhile, OpenAI is building its first AI factory in Norway as part of their Stargate Project. At the same time, Altman confirmed that OpenAI is now officially chasing ASI. That backs up what I said earlier: OpenAI has reached its AGI definition, AI systems that can perform most economically valuable tasks.

Back to GPT-5. After the Agent Mode release, I didn’t expect it before late August. But the rumors are persistent. GPT-5 will launch before you receive the next edition of Digital Dips. GPT-5 is expected to bring a 1M token input window, MCP support, fact-checking itself before answering and advanced reasoning. The world moves fast, and the possibilities for everyone are rising to wild new heights. The past two weeks, a lot has happened again. 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: The AI Act just got real. From this week, every business in Europe using AI, yes, even ChatGPT, must comply. Meanwhile, Runway redefines video editing, a ten-person startup beats the giants, Tencent makes 3D worlds from a single image, Google upgrades Imagen and Gemini, and China drops a coding beast.

• 🤖 Digital Toolbox: MagicPath turns plain talk into pixel-perfect designs. Atria brings clarity to ad chaos. And Context AI is like having a smart intern who never sleeps.

• 🧐 In Case You Missed It: Google drops features like it’s on fire: Gemini, Docs, Notebook, LangExtract, you name it. Replit stacks your tasks. PayPal goes crypto. Midjourney loops. Firecrawl watches the web. Magnific and Morphic make your visuals pop. AI agents land everywhere, from your CRM to your Apple Watch. Yes, even Norton built a browser. And many more!

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.

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: The AI Act changed yesterday. Is your business ready?

In Europe, there’s an attempt to tame the AI war with regulations. So far, it mostly results in AI labs delaying the release of their newest features in Europe, but that’s a side note. The law has been active since 2 February 2025, but from 2 August, it really deserves your full attention. Every company developing or using AI in their business must now comply with new regulations. Yes, even if your only AI usage is staff members chatting with ChatGPT.

From that date, companies across every sector must comply with a new layer of governance. Not just the AI labs. Not just the tech giants. Not just the startups. Every business using AI, even passively, is legally bound to act. That includes you. Non-compliance? It’s not just a slap on the wrist. It could cost you €35 million or 7% of global revenue. But the real risk? Falling behind, while your competitors turn regulation into a competitive edge. Because here’s what most companies miss: compliance isn’t the end goal. It’s the starting line for cultural transformation.

Why does this matter so much? Because this isn’t GDPR 2.0. It’s bigger. The Act doesn’t just regulate data. It regulates intelligence. Decision-making, autonomy and judgment. It classifies every AI system into four risk levels, from minimal to high, and dictates the guardrails for each. If you’re touching HR, finance, education, healthcare, infrastructure, or surveillance, you’re likely in the high-risk zone. And that means documentation, transparency, oversight and governance. Not someday in the future. Now.

So what do you do next? Start with a comprehensive AI inventory, you'll be surprised what you discover lurking in your shadow IT. Classify your risks properly against the Act's specific definitions. Build governance from the ground up with clear roles and escalation paths. Most critically, invest heavily in AI literacy training that goes beyond minimum requirements, because when your entire organization understands AI's capabilities and limitations, you don't just avoid fines; you make better decisions, win more deals, and capture the "compliance premium" that early movers are already experiencing in competitive bidding situations.

AI literacy training is not an option. The AI Act requires companies to ensure AI literacy across their workforce (article 4). Not just developers. Everyone who touches, operates, or is impacted by AI. That means sales, HR, customer service, leadership. Don’t see this as just a mandate. Look at it as an invitation. A hidden opportunity to level up. For years, upskilling has been pushed aside: too expensive, too time-consuming, too hard to prove ROI. But now? It’s a legal requirement. So don’t waste the moment. Go beyond the minimum. Don’t just teach compliance. Teach competence. Teach creativity. Teach your teams how to wield AI as a tool for performance, insight, and innovation. To build an organization that understands AI.

This raises the real questions for leaders:

• Do we even know how AI is used across our organization, including shadow AI usage by individual employees?

• Are our systems categorized under the AI Act risk tiers and documented?

• Are we training our people not just in AI safety, but in practical, performance-driving AI use?

• Have we embedded AI governance into our risk frameworks?

• Is our leadership engaged, or is compliance buried in IT?

• Are we prepared for regulators? For customers? For partners who demand proof?

Because your investors will ask. Your clients will ask. And your employees will ask.

But the real question here is not ‘Can you comply?’ It’s ‘Will you lead?’. This is where regulation meets transformation. How? By turning a legal deadline into a cultural shift. Let’s make the AI Act your next strategic move.

Runway Aleph defines a new era in video creation, production, and editing

TL;DR

Runway Aleph is a cutting-edge video editing model that revolutionizes multi-task visual generation. It allows users to perform extensive edits on video inputs, such as adding or removing objects, generating any angle of a scene, and manipulating style and lighting. It brings an unprecedented level of control and creativity to video production.

Read it yourself?

Sentiment

The online community is exploring the capabilities of Runway Aleph with great enthusiasm. Users rave about the model's potential and reliability, especially how it effortlessly integrates new camera perspectives with sequential shot generation. This groundbreaking feature allows creators to construct a cohesive story from a single video frame, which has sparked widespread excitement.

My thoughts

It's no surprise that Netflix utilizes Runway to finalize some of their series. Runway is becoming indispensable for cinematic creators and storytellers. This democratizes the creation of high-quality, Hollywood-standard content, empowering more people to produce breathtaking videos, movies or commercials. You don't need to be an expert cameraman, editor, or lighting technician. Runway handles it all. The only requirement is to be a great storyteller, and you can achieve this with just your mobile phone.

A tiny team just crushed the giants

TL;DR

A small AI lab has reached 92.4% accuracy on GPQA Diamond and 96.1% on SimpleQA, outperforming every major model from OpenAI, Google, Anthropic, and xAI. It’s the first time a system excels at both complex reasoning and basic factual understanding. Even more surprising? This breakthrough came not from a tech giant, but from a startup with just ten people.

Read it yourself?

Sentiment

Online reactions were a mix of disbelief and admiration. While Silicon Valley pours billions into massive AI ventures, this modest team quietly built a system that outperforms them all. People are wondering how such a small crew managed to do what the biggest players couldn’t. Some are even joking that Meta might start showing up with acquisition offers.

My thoughts

Most AI today works like a very confident intern, fluent, fast, but often wrong. That’s fine for writing emails or summaries. But when it comes to science, medicine, or anything where truth matters, you need something smarter. This new model marks a turning point. It knows when it’s right, and it admits when it’s not. That kind of honesty is rare in AI. What we’re seeing isn’t just better performance, it’s a shift from persuasive AI to precise, self-aware AI. And that’s the leap we’ve been waiting for.

More byte-sized breakthroughs:

• Tencent’s AI builds explorable 3D worlds from a single image
  Hunyuan has announced the release and open-sourcing of the groundbreaking Hunyuan3D World Model 1.0. This innovative model allows users to generate immersive 3D worlds from a single sentence or image, revolutionizing game development, VR, and digital content creation, with full editability and simulation capabilities. This is Photoshop + GPT for 3D worlds. It transforms how we create, share, and interact with experiences.

• Google just launched Imagen 4 and Gemini 2.5 with Deep Think
  Google has released Imagen 4, the latest iteration of its text-to-image model. This new release enhances the quality of text rendering within images, a leap forward from their previous models. With Deep Think, Gemini 2.5 is able to intelligently extend its "thinking time" so it can generate multiple, parallel streams of thought simultaneously. It’s the same model that achieved the gold-medal status at IMO. Big news!

• China's breakthrough in open source agentic code models

  Qwen has released Qwen3-Coder, setting new benchmarks in the world of open-source AI coding. The underlying model, Qwen3-Coder-480B-A35B-Instruct, has an impressive 256K token context, which is expandable to 1M tokens. This new model outperforms counterparts like Anthropic’s Opus 4 in agentic coding and browser-use, marking a significant milestone for Chinese-developed AI.

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

MagicPath - Imagine designing a site by talking to it

MagicPath lets you describe an user interface in plain English and instantly turns it into web page designs on an infinite canvas. Then you refine, generate variants, and export clean HTML/CSS or React code, all without a single line of manual layout. It understands natural language so you can say “create a hero section” and see it appear in seconds. The canvas stretches endlessly: build multiple screens, move components and test flows. When ready, export usable code for developers.

Atria – Boost your ad creativity

Atria lets you search through millions of top‑performing TikTok and Meta ads with filters that actually make sense. Save your favourites, break down what works, and generate your own concepts, scripts, and visuals, no more writer’s block. Performance data sits right next to creative insights, so you know what hits and why. Then export and collaborate directly in Canva or your preferred workflow. It’s your shortcut to smarter, faster, better ads.

Context AI – Your AI office teammate

Context.ai gathers all your chats, docs, spreadsheets and data sources into one secure workspace and then turns them into polished presentations, reports and sheets. Ask it to do research across Slack, Salesforce or Excel and it gives you data-grounded outputs tailored to your style. Generate slide decks, written reports or insight-ready spreadsheets in moments. Canvas is infinite, collaboration is seamless, and export-ready work is just a chat away.

A roundup of updates that are too cheesy to ignore.

• Pipedream releases MCP Chat, an open-source tool to connect with 2700+ APIs and 10k+ software tools.

• Replit Queue lets you stack, edit, and reorder tasks, optimizing your workflow with Replit Agent.

• Replit ramps up security with Checkpoints, Smarter Docs, and upcoming Chat Mode for safer vibe coding.

• FLUX releases Krea-1, an open-source image model bringing striking realism and aesthetics.

• Magnific unveils its new Creative Upscaler for photographers, enhancing resolution without altering your original vision.

• PayPal World connects global payment systems and digital wallets for billions worldwide.

• PayPal unveils Pay with Crypto, streamlining global purchases and money transfers via blockchain.

• Eigent lands on desktops with customizable, parallel, AI agents & MCP’s for tackling complex tasks.

• OpenAI’s ChatGPT Study Mode navigates students through problems step-by-step for deeper learning.

• Google’s LangExtract transforms text into structured data with source tracking and instant visuals.

• Google’s Notebook Studio UI debuts with multi-output creation and audio overviews in different languages for a streamlined workflow.

• Google’s Notebook unveils Video Overviews, turning notebooks into slide shows for a visual flair.

• Google’s AI Mode unveils Search Live for real-time concept help, Canvas for project organization and helps to solve tough questions as Study Mode.

• Google Gemini 2.5 introduces conversational image segmentation to understand complex visual queries effortlessly.

• Google's Gemini 2.5 Flash-Lite is now stable, offering unmatched speed and cost-efficiency with support for a 1 million-token context window.

• Google supercharges Docs, Slides, and Vids with Imagen 4’s multilingual image generation.

• Google’s Flow lets you draw your video prompt, turning sketches into scenes instantly.

• Google’s Opal lets you create and share AI mini-apps with natural language, no coding needed, now in US public beta.

• YouTube Shorts unveils Veo 2 to transform photos into videos and adds AI-powered effects to spice up your creations.

• Firecrawl Observer debuts as an open-source web monitor for instant page updates.

• Higgsfield Steal reinvents the web with photo recreation and personalizes with Soul ID.

• Higgsfield introduces Multi-Reference, offering up to 4 images for perfect character consistency.

• ElevenLabs integrates WebRTC into its AI, enhancing echo cancellation and noise removal.

• ElevenLabs’ AgentCLI lets you manage conversational agents like code, bringing version control and deep workflow integration.

• Genspark AI debuts Multi-Agent Orchestration: create 10 slide decks simultaneously with coordinated agents.

• GenSpark’s AI Meeting Notes debuts on Apple Watch, keeping your summaries handy on the go.

• Genspark AI Slides 2.0 offers free templates, deep thinking, and PowerPoint-style edits, leaving Canva behind.

• Manus unveils Wide Research for Pro users, revolutionizing complex multi-tasking research at scale.

• Midjourney now supports creating videos with specific start and end frames. This also means you can create looping videos.

• Notion's upgraded MCP lets AI tools transform chats into pages and access tasks seamlessly.

• Anthrophic’s Claude integrates with Notion and Linear to manage projects and update docs seamlessly.

• Anthrophic’s Claude app lets you draft and send emails, messages, and invites right from your mobile.

• Anthropic’s Claude allows you to upload PDFs, images, and code files to its AI-driven apps for seamless data integration.

• HubSpot integrates MCP access, transforming AI into your CRM's personal sales assistant.

• Synthesia unveils perfect lip-sync tech AI Dubbing for seamless video translation.

• Synthesia's Express‑2 avatars bring scripts to life with natural gestures and pixel-perfect lip sync.

• Norton Neo debuts as the first AI-native browser, offering smarter search and seamless chat without trackers.

• Hedra launches Live Avatars for video streaming with ultra-low latency at $0.05/min.

• CompositeAI launches an agent to automate tasks directly from your browser.

• Morphic Camera Motion now adds smooth, cinematic flair to your video creations.

• Morphic introduces 3D motion. It transforms static images into dynamic 3D motion videos.

• GitHub Spark launches for Copilot Pro+ subscribers, turning ideas into apps instantly with no setup.

• Grok introduces App Connections for automated summaries of your apps, from Gmail to Slack.

• Kling Lab debuts as a workspace enhancing creativity, efficiency, and collaboration.

• Alibaba’s Wan2.2 debuts as the first open-source MoE video model, offering cinematic control and enhanced motion generation.

• Alibaba’s Qwen3-30B-A3B models medium-sized brilliance with a 256K-token context, excelling in reasoning and tool use.

• Microsoft Edge’s debuts Copilot Mode, analyzing tabs and introducing task delegation for AI-driven browsing.

• Microsoft Copilot transforms your interests into podcast episodes with just a request.

• Julius debuts Data Connectors for seamless insights from Postgres, Google Drive, and SharePoint.

• Zhipu debuts GLM 4.5 with hybrid reasoning, boosting tool calling and coding benchmarks.

• Ideogram Character lets you craft consistent character visuals from just a single image, now free for all users.

• Shortcut debuts as the first superhuman Excel agent, outperforming McKinsey analysts 89.1% of the time.

• LumaLabs’ Dream Machine unveils Modify with Instructions, streamlining VFX and design workflows with natural language commands.

• Firecrawl makes migrating website content to platforms like WordPress and Shopify a breeze.

• NEO revolutionizes ML engineering with autonomous, full-stack capabilities, scoring medals in 34% of Kaggle competitions.

• Mistral's Codestral 25.08 delivers boosted code completion and search for enterprises.

This was it. Our thirty eight 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 organization to treat this Act not as a burden, but as a blueprint for resilience, capability, and competitive advantage, then I can help. From AI system audits to workforce training strategies, from governance frameworks to real-world implementation. Just reply to this email or hit me up on LinkedIn.

Looking forward to what tomorrow brings! ▽

-Wesley