Hi {{firstName|Futurist}},
Six weeks. That's how long it's been since the last Digital Dips landed in your inbox. Sorry. Won't happen again. The truth is that the past month disappeared into executive course preparations, work commitments, and everything else that comes with life outside of a newsletter. Before I knew it, six weeks had passed. Which, in AI terms, is about one year. In that time, there are two Omni models released. Anthropic launched two new models. And Donald Trump may have personally increased China's chances of winning the AI race.
That's also exactly why I'm looking at ways to make sure these updates keep coming, even when life gets busy. The pace of progress is only increasing. Missing a week of AI news used to be fine. Missing five weeks now means catching up on a completely different landscape. So, let's do exactly that.
So grab your favorite snack, settle in, and let's dip into what's cooking. 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.
AI / SPORT: from AI hype to AI value
On 25 June, I'll be speaking at AI / SPORT in Rotterdam. A compact seminar for sports leaders who are looking beyond the latest AI tools and asking a more important question: what should we actually do with AI now?
We'll cover the biggest AI developments for the second half of 2026, where AI is already creating value inside sports organizations, and how to build an AI strategy that includes people, processes and data.
Together with other experts like Thomas van Schaik, Hans Westerbeek and Peter Sprenger, we'll connect strategy, sports innovation and real-world AI practice. The sessions will be held in Dutch.
📅 25 June 2026
🕚 11:00 – 13:30
📍 Rotterdam
🎟️ Digital Dips readers can attend for €75 with discount code SPORTAI20
If you're responsible for the future of your sports organization, this session is for you. Hopefully I'll see some of you there.
🍢 Finger food for thought
One topic that deserves more than a bite. Not too long. Just enough to chew on.
Headstory: The gap is getting bigger
A few weeks ago, I was part of an executive program with C-level leaders from the sports industry. The topic was simple: what does AI mean for the future of sport? The discussions were fascinating. Not because people lacked interest. Quite the opposite. Everyone understood that AI matters.
What struck me was something else. There is a growing gap between what AI models are already capable of today and how organizations are actually using them. And that gap is getting bigger.
In the past two years, many executives could comfort themselves with the idea that AI was mostly hype. Interesting technology, promising demos, impressive headlines. Useful for writing emails or summarizing documents. But not something that fundamentally changed how organizations operated. That argument is becoming increasingly difficult to defend.
Recent evaluations of Anthropic's Mythos (Fable 5) model paint a remarkable picture. Frontier models are no longer just getting better at answering questions. They are becoming capable of completing complex tasks over extended periods of time with increasingly high levels of accuracy. That distinction matters. The real story is not that models can work longer. The real story is that they can work longer while maintaining performance.
In other words, they are becoming more reliable. And reliability is what turns technology from an experiment into infrastructure. The implications are difficult to overstate. Researchers tracking frontier AI capabilities have observed that the length of tasks these systems can successfully complete has been doubling every few months. More importantly, the pace itself appears to be accelerating.
An exponential curve is already difficult for humans to grasp. An exponential curve that accelerates over time is even harder. Our intuition simply isn't built for it. Yet organizations continue to plan as if progress will remain linear.
Many companies are still debating pilot projects while the underlying technology is improving at a rate that would have seemed impossible only two years ago. That disconnect creates risk. Not because AI will replace every employee tomorrow. But because organizations that learn how to integrate these capabilities into workflows today will build advantages that become increasingly difficult to catch up with later.
The sports executives I spoke with were not alone. Across industries, I see the same pattern. Most organizations are using AI as a productivity tool. Very few are redesigning processes around it. That difference sounds subtle. It is not. One approach delivers incremental gains. The other changes the economics of work. Most companies are asking how AI can help employees perform existing tasks. The more important question is which tasks should exist in the first place. That is the shift many leaders are still missing. AI is not simply changing how work is performed. It is changing what work needs to be performed at all.
In recent days, Anthropic’s Mythos 5 and Fable 5 became more than just another AI launch. It became a policy story. After headlines about a US government directive, Anthropic abruptly disabled access to Mythos 5 and Fable 5. The reason: national security concerns and fears that the models could be misused to by other countries or bad actors. For users, the result was simple. One moment, the model was there. The next, it was gone.
And that is exactly why this story matters. A highly capable model appears, generates enormous attention, and then disappears almost as quickly as it arrived. Whether one agrees with the decision or not is almost beside the point. The broader takeaway is that access to powerful AI systems can no longer be assumed.
What happened with Mythos 5 and Fable 5 highlights a reality many organizations have yet to fully appreciate: a growing share of the world's AI capability is controlled by a small number of companies operating within a small number of jurisdictions. When governments intervene, whether for national security, regulation or geopolitical reasons, organizations elsewhere have little influence over the outcome. Yet they may still bear the consequences.
For organizations, this raises an important strategic question: how dependent do you want to be on capabilities that exist entirely outside your control? Open-source models continue to improve rapidly. Local deployment is becoming more accessible. Compute is becoming cheaper. The ability to run increasingly powerful models on your own infrastructure is no longer a distant possibility. It is becoming a strategic option.
The organizations that build internal AI capabilities today are not simply investing in efficiency. They are reducing dependency. They are creating resilience. And they are ensuring that critical parts of their future do not rely entirely on decisions made in boardrooms or government offices thousands of miles away.
The conversation should therefore move beyond which model is currently best. The more important discussion is how organizations build AI capability, AI literacy and AI sovereignty. Because two gaps are widening simultaneously. The first is the gap between AI capability and organizational capability. Between what the models can do and what companies ask them to do. Between organizations that are experimenting and organizations that are transforming. Between leaders who recognize what is coming and leaders who still see AI as a side project. The second is the gap between organizations that consume AI and organizations that control AI. Between those who depend entirely on external providers and those who are building their own capabilities. Between those who rent intelligence and those who own strategic access to it. Together, these two gaps may become the defining competitive advantage of this decade.
The organizations that close them fastest will be the ones that shape the future of their industries. If you still don't feel urgency, I am not sure what signal you are waiting for.
Yesterday was the best time to start.
Today is the second-best.
🍟 Crispy bites
Fresh tech nuggets. Short, sharp, snackable.
Fable 5: A frontier model, then an access freeze
TL;DR
Anthropic released Claude Fable 5 as a “Mythos-class” model that’s been made safe enough for broad release, with Opus 4.8, released few weeks earlier, as a fallback on sensitive topics like cybersecurity. The hype was immediate: people called it a clear major-version leap and “the best coding model.” Then came the real plot twist: the US government stepped in with an export control directive suspending all access to Fable 5 and Mythos 5 for foreign nationals, even foreign Anthropic employees. This isn’t just a model launch; it’s platform power, access politics, and national security all in one release.
Why this matters
Fable 5 is explicitly positioned as “Mythos-class, made safe for general use”: the top capability exists, but only reaches users through guardrails. That’s a new kind of product strategy.
Anthropic openly says certain queries (including cybersecurity) are routed to Opus 4.8 instead. That’s not a footnote; it’s a built-in safety gate at the capability level.
The market reaction wasn’t “small iteration,” but “major-version bump” and SOTA “by a margin,” including a qualitative jump. That raises the stakes of every restriction.
The government pushback is unusually concrete: an export control directive suspending access for all foreign nationals, inside and outside the US. That makes “who gets to use it” as important as “what it can do.”
The conversation shifts toward inevitability: if cloud access can be switched off, local/open compute immediately becomes the alternative story.
My thoughts
This is the moment when “model releases” stop feeling like product news and start feeling like geopolitics. Anthropic is trying to sell a sharp balance: Mythos-level capability, packaged as Fable 5 with safeguards, and with Opus 4.8 as an automatic downgrade when things get risky. That’s a mature framing: not everything you can build should ship without brakes. And they’re saying it out loud.
But precisely because Fable 5 seems to be such a step forward, serious testers are calling it SOTA and a real leap, the question of “who gets access?” becomes the main story immediately. The export control directive makes that painfully visible: not just rate limits or policy text, but a hard boundary based on nationality, even for the company’s own employees. That’s “platform control” in its most literal form: the capability exists, but access is a policy layer.
And then you get the second-order effects: once users experience that a frontier model can disappear or be restricted overnight, attention shifts to what can’t be switched off, local models, owned infrastructure, and alternatives outside a single centralized gatekeeper. The irony is that safety gates are meant to reduce risk, but they also reveal exactly where the power sits. And once you see that, you start planning around it.
Google’s new Omni,“anything-to-anything”, starts with video
TL;DR
Google’s Gemini Omni pushes multimodal generation toward “anything-to-anything,” starting with video. The headline isn’t that video got better. It’s that Google is turning creation into a native capability of the Gemini stack, where inputs (text, image, audio, video) collapse into one pipeline, and the product becomes the platform.
Read it yourself
Why this matters
Omni is positioned as “create anything from any input,” which is a bigger claim than “better video model.” It’s an interface shift.
The first wave is video, but the strategic move is unifying modalities so workflows don’t hop between tools.
Early reactions show the gap between “looks real at a glance” and “is physically consistent,” which is exactly where trust and pro use-cases live.
The hands-on press frames it as wild and powerful, but explicitly not the singularity, meaning the ceiling is visible, and that matters for adoption.
The discussion is already splitting into two camps: “this is sci‑fi” vs “it still breaks on realism,” which is the same pattern every time a model crosses a new threshold.
My thoughts
The easiest way to understand Google Omni is: it’s not “a video generator.” It’s Google trying to make media creation work like a single, unified input/output system. Google’s own framing is “create anything from any input”, meaning you can start with text, an image, audio, or an existing clip, and push it through one model pipeline to get a new video out. That’s a different product bet than “type a prompt, get a clip.” It’s closer to: give me your messy creative intent, plus references, and I’ll assemble something coherent.
Why that matters is workflow. If Omni becomes good enough, the center of gravity shifts from “which editing app do you use?” to “which model layer do you build your creation process on?” The Verge’s hands-on basically shows the promise: it can remix reality fast and it feels like a new kind of interface for making things. But it also shows the limit: it’s impressive without being flawless, and you still hit moments where it doesn’t fully hold up.
The public reaction is already split in a predictable way. Some testers describe it as straight-up sci‑fi because the multimodal flow feels fluid and integrated when it works. Others are more skeptical and focus on the “too much” vibe, the sense that we’re crossing into a world where generating believable media is becoming normal, which raises the stakes on trust, provenance, and misuse. My take: Omni is Google planting a flag that the next platform war isn’t just about chat or coding. It’s about who owns the creation layer, and who can ship it with enough quality and guardrails that real teams will rely on it.
Starchild-1 sets new benchmarks for real-time simulated intelligence
TL;DR
Starchild-1 arrives as the first multimodal world model, capable of generating audio and video in real-time based on user input. It surpasses traditional models by capturing both visual and auditory inputs, paving the way for advances in robotics, education, and gaming. Unlike previous models, it adapts continually to interactive environments. This step toward general world intelligence could reshape how machines learn and respond to the world.
Why this matters
Starchild-1 is the first multimodal model generating real-time audio and video together.
It uses a new multimodal causal training stack for real-time simulations.
Traditional models had fixed outputs; Starchild-1 adapts to user input dynamically.
Enables more natural AI interaction with the world, improving education and gaming.
Facilitates long-term multimodal stability and interaction in simulations.
My Taste
Starchild-1 highlights how integrating sound into world models is a game-changer. While it's easy to focus on visuals, sound is a critical part of our interaction with the world. This model's real-time adaptability shows how AI can become more aligned with our sensory experiences. As this technology matures, I'm excited to see if it addresses the increasingly complex demands of personalized and interactive AI, potentially unlocking new avenues in devices and applications yet to be conceived.
Google harmonizes shopping with Universal Cart
TL;DR
Google's Universal Cart will function across Search, YouTube, Gmail, and other Google services. It automatically finds deals, tracks price changes, and uses AI to suggest product alternatives when needed. The cart is built on Google's Gemini models for smarter functionality as these models advance. By understanding user payment perks via Google Wallet, the system uncovers hidden savings. This rollout marks a significant step in agentic commerce, broadening transaction efficiency and potentially altering retail landscapes.
Why this matters
Universal Cart functions across Search, YouTube, and Gmail, enhancing shopping versatility.
Google's Gemini AI models power the cart, driving smarter shopping innovations.
The Shopping Graph leverages over 60 billion product listings to optimize the user experience.
UCP ensures seamless agentic checkouts with brands like Nike, Target, and Shopify merchants.
The system deciphers payment perks for savings, integrating with existing Google Wallet features.
My Taste
Google's shift towards creating a seamless, smarter shopping ecosystem with Universal Cart is an impressive move. Integrating various services means shopping becomes more intuitive and proactive, not just responsive. This is about more than convenience, it's a step towards fundamentally redefining how digital commerce operates. If executed well, businesses tied to Google's ecosystem could see serious transformations in customer engagement and transaction efficiency. What's next might not just be about buying smarter but living smarter as shopping becomes an integral part of our digital lifescape.
Google steps up in AI Search
TL;DR
Google is revamping its Search capabilities by integrating advanced AI features like the Gemini 3.5 Flash model. This marks the Search box's largest transformation in over 25 years. With AI-enhanced tools, users can pose complex questions and engage dynamically with information agents. This rollout is pivotal as AI-augmented Search queries have doubled every quarter, now surpassing one billion monthly users. The implications are vast, poised to redefine how information is accessed and utilized worldwide.
Why this matters
Over 1 billion users are already on AI Mode, with query numbers doubling quarterly.
Gemini 3.5 Flash model now powers AI Mode, ensuring peak performance in search tasks.
Information agents, launching for Google AI Pro & Ultra subscribers, offer personalized updates.
Personal Intelligence in AI Mode expands to nearly 200 countries, across 98 languages.
New AI-powered Search box supports text, images, files, and Chrome tabs as inputs.
My Taste
This move to integrate AI deeply into Google Search is timely and significant. As AI capabilities expand, how we access and process information changes fundamentally. What's especially intriguing is the seamless shift towards personal agents capable of proactive information gathering. It's a glimpse into a future where digital tools operate increasingly like personal assistants. We’ve reached a point where the convenience and intelligence of AI are not optional luxuries but essential tools. As these features roll out, the key question is how quickly users will adapt and what new opportunities this behavior shift will create for businesses like Google did more than 20 years ago.
🧀 Cheesy pick
A cheesy selection of three tools and one tasty rabbit hole.
🍱 Leftovers
A roundup of updates that are too cheesy to ignore.
ByteDance ships Seed2 3D generation from a single image or text prompt, faster path from idea to assets.
Higgsfield CLI ships a lean way to run marketing “Skills” without bloated schemas, lower agent spend, higher-quality creative.
Higgsfield Supercomputer is a self-learning agent for end-to-end task execution, 40+ tools, three memory layers, runs in browser or Telegram.
Higgsfields ships 25+ UGC video hook templates (cold opens, POV, interrupts, story hooks), usable in-platform or via Claude MCP.
Higgsfield launched Ad Reference: upload your best videos and it recreates the format via MCP across agents.
Higgsfield launched Supercomputer, a cloud-native AI agent unifying models, tools, and creative workflows to ship campaigns end-to-end.
Higgsfield plugins for Premiere Pro and After Effects are live, generate assets, reframe, remove backgrounds, edit by drawing, and upscale to 4K.
Runway Characters turns one image into a conversational video agent streaming HD at 24fps, with just 1.75s end-to-end latency.
Runway Aleph 2.0 adds single‑frame video edits that propagate across the whole clip, plus a new web Edit Studio for faster iteration.
Runway’s Agent is a chat-based AI creative partner that ideates and produces fully finished, sound-designed, edited videos.
Runway MCP plugs Runway into Claude/ChatGPT/Cursor/Replit, so you can generate Gen-4.5 images/videos without leaving your workflow.
Anthropic’s Claude ships keyless auth, browser CLI login or cloud identity (AWS/GCP/Azure/OIDC), so teams can stop juggling API keys.
Anthropic’s Claude shipped ready-to-run finance agent templates, pitches, valuation reviews, month-end close in Cowork/Claude Code or run as Managed Agents.
Anthropic partnered with SpaceX to boost compute, enabling higher usage limits for Claude Code and the Claude API.
Anthropic’s Claude Managed Agents adds “dreaming” plus outcomes, multi-agent orchestration, and webhooks in public beta.
Anthropic’s Claude for Excel/PowerPoint/Word is now GA, with Outlook in public beta, shared context across Microsoft apps keeps work flowing.
Anthropic’s Claude Code shipped Agent View: a single list of all sessions, making it easier to track work across threads.
Anthropic Mythos reportedly helped three researchers build a macOS kernel exploit bypassing Apple’s M5 MIE in six days.
Anthropic’s launched self-hosted sandboxes and MCP tunnels so agents run inside your perimeter.
Anthropic’s Claude Code added dynamic workflows, it plans, spins up hundreds of parallel subagents, and self-verifies for big migrations.
Anthropic’s Claude Code shipped a security-guidance plugin that flags and fixes vulnerabilities while you code.
Inworld’s new Realtime TTS-2 voice model that listens, takes voice direction, keeps one identity across 100+ languages, more natural agents.
SpaceXAI’s Grok 4.3 is live on the xAI API, fastest yet, topping tool-calling and instruction-following leaderboards; #1 in enterprise domains.
SpaceXAI’s Grok Voice Think Fast 1.0 customer-support voice agent handles complex workflows fast, stays accurate in noisy settings, and scales tool-heavy calls.
SpaceXAI’s Grok Build early beta agentic CLI for coding, app-building, and workflow automation.
Microsoft’s Copilot Cowork adds iOS/Android support plus new skills, plugins, and connectors.
Microsoft’s Copilot got a redesign, simpler, faster, more intuitive, aimed at keeping you in the flow instead of fighting the UI.
OpenAI GPT-5.5 Instant is rolling out in ChatGPT, smarter, clearer, more personalized replies with a warmer tone, plus tighter concision.
OpenAI’s ChatGPT add-on for Excel and Google Sheets: analyze messy data, write formulas, update sheets, and explain steps.
OpenAI’s GPT-Realtime-2 smartest voice model in the API, GPT‑5‑class reasoning for real-time voice agents that listen and solve live.
OpenAI’s Codex now runs directly in Chrome on macOS/Windows, works better with sites, and can operate across tabs in parallel.
OpenAI Deployment Company unites 19 firms to help businesses ship frontier AI into production faster.
OpenAI Daybreak is their new cyber-defense push combining frontier models, Codex, and security partners to help teams secure software faster.
OpenAI’s Codex shipped a Chrome extension that writes and runs code to automate repetitive browser workflows.
OpenAI adds SynthID watermarks plus a public verification tool for AI images, alongside C2PA, clearer provenance and easier detection.
OpenAI launched Guaranteed Capacity, letting customers lock in long-term access to OpenAI compute.
OpenAI’s ChatGPT now creates and edits presentations directly in PowerPoint.
OpenAI disproved the “square grid” intuition in Erdős’s planar unit distance problem, an 80-year-old open question.
OpenAI’s Codex Computer Use now supports Windows, so it can act on your PC, and you can steer tasks from ChatGPT mobile.
OpenAI’s ChatGPT added a table of contents for long chats, making sprawling “one quick thing” threads easier to navigate.
OpenAI’s Codex can now securely use apps on your locked Mac from your phone, without leaving your machine unlocked.
OpenAI launched private MCP servers. Keep MCP inside your network; ChatGPT, Codex, and Responses API connect via outbound-only HTTPS.
Manus Projects now learn from every task, suggesting updates to instructions/files/skills so teams reuse context and cut setup.
Manus shipped Scheduled Tasks 2.0: recurring work now runs in the right place with the right context, continuing in-task.
Manus Projects now work on mobile, so you can organize workflows on the go with shared files, instructions, skills, connectors.
Amazon Bedrock AgentCore adds AgentCore payments, letting agents pay for APIs/MCP/web/other agents mid-run, no custom billing.
Lovable shipped an MCP server so terminals and AI agents can create, iterate, and deploy apps end‑to‑end.
Lovable just upgraded Lovable’s design output, set typography, layout and color preferences, explore concepts, and ship landing pages fast.
Lovable shipped SEO, new apps are server-side rendered; existing ones get automatic pre-rendering for better discoverability.
Lovable shipped Google connectors, so your full-stack apps can pull from Gmail, Drive, Sheets, Calendar, BigQuery, and more.
Lovable shipped subagents: spin up behind-the-scenes helpers to research, review, and QA in parallel.
Google’s Gemini 3.1 Flash-Lite is their most cost-efficient model, tuned for high-volume agentic tasks, translation, and simple data processing.
Google spotted the first AI-developed zero-day used in the wild—early counter-discovery may have stopped a wider strike.
Google’s Gemini Intelligence automates multi-step tasks across your apps, fill out forms, turns spoken thoughts into text and much more.
Google’s first Gemini-first laptop, heavyweight performance, Gemini at the core, and tight Android phone sync, landing this fall.
Google’s Gemini 3.5 DeepMind’s new model family targets “real-world action”; first drop is 3.5 Flash, tuned for agents and coding.
Google’s Gemini 3.5 Flash Google’s new top Flash model targets better intelligence, speed, and cost, and it’s now available everywhere.
Google’s Gemini Spark 24/7 personal AI agent on Gemini 3.5, built on Antigravity for long-running background tasks under your direction.
Google’s Antigravity 2.0 ships as a standalone desktop app: multi-agent teams, scheduled tasks, native voice, and one-click Google integrations.
Google’s Gemini for Science experimental AI tool suite helps scientists explore hypotheses, validate at scale, and unpack literature faster.
Google’s Flow launched Tools: describe the editor/resizer/shader you want, and Flow helps build it for your workflow.
Google’s Gemini API Managed Agents one call spins up an agent with a Google-hosted remote Linux box.
Google’s AI Studio now lets anyone build native Android apps for free, no coding required.
Google announced Antigravity CLI, a lightweight terminal interface to run the same Antigravity agents, harness, and models.
LTX Studio shipped video-to-video: pick Pose/Depth/Edges, add a prompt + starting image, and generate structured edits via LTX-2.3.
Thinking Machines shares an AI built for real-time, multi-modal collaboration, talk, listen, watch, think—plus early results and a live demo.
Meta’s New Muse Spark voice mode: interrupt, switch topics, languages'; generates images and pulls Reels/maps recommendations.
Meta’s Incognito Chat is a “completely private” AI mode on WhatsApp and the Meta AI app, privacy-first AI chats go mainstream.
AI Radio on X launches as the first fully AI-run station, streaming 24/7 AI news with ambient breaks, hands-free signal for builders.
Kimi’s Web Bridge new browser extension lets agents browse like humans: search, scroll, click and type.
Recraft shipped V4.1: more human photorealism, dreamier gradients, and new illustration styles.
Tavus launched Image-to-Replica: turn a single image (mascot, historical figure, AI character) into a conversational replica.
Malta gives every citizen free ChatGPT Plus for a year, after an AI literacy course built by University of Malta, not OpenAI.
Telegram Bots can now talk to other bots, giving autonomous agents a native communication layer that humans can still follow.
Cursor Composer 2.5 is smarter, steadier on long tasks, and better with complex instructions.
TesanaAI’s new game-making AI model Muranyi 3 promises faster idea-to-game with better graphics/VFX, smarter NPCs, smoother animations, and reliable builds.
Blink ships an AI Chrome Extension Builder that turns your repeat web steps into an extension—describe it, it builds.
ElevenLabs Speech Engine turns any chat agent into a full voice agent with one prompt, bundling speech, transcription and orchestration.
ElevenLabs launched Music v2, better vocals, instrumentation, and arrangements across genres, plus stronger multilingual support.
ElevenLabs shipped Dubbing v2, a dubbing model that carries original emotion and performance across languages.
YVO3D WORLD turns a single image into modular, editable 3D environments you can drop into real Unreal/UE5 game workflows.
Qwen3.7-Max Alibaba’s new flagship model for the “Agent Era,” tuned for end-to-end coding and reliable office productivity assistance.
Telegram added assignable inbox bots that can read and reply for you, with granular permissions and chat access controls.
Willow Scribe Voice AI writing assistant that drafts emails, docs, and messages fast using your style and context—“say it, Scribe writes it.”
Robinhood now lets AI agents trade stocks, pushing automation into retail investing—and raising the bar on guardrails and accountability.
Robinhood Agentic Accounts let AI agents explore trade ideas, rebalance portfolios, build tools, and place trades, so strategies run 24/7.
Krea 2 shipped an API, now integrated with fal and ComfyUI, plus agent support (Hermes) and Claude/Codex/OpenClaw.
Krea 2 is their first foundation model, built from scratch for aesthetic diversity and tighter stylistic control.
Sesame shipped an iOS app preview with a collection of personal agents, new characters and features since last year’s research preview.
YouTube releases Ask YouTube as their new conversational search, which lets you ask complex queries and get guided answers fast.
YVO3D V3 generates fully modular 3D environments from a single image, rendered in Unreal Engine, handy for fast prototyping in gamedev.
Tencent’s Miora AI creative agent studio hits international beta, generate images, video, UI/UX, and 3D on one canvas, no tool-switching.
Tencent’s Workboddy desktop AI agent goes global, autonomously shipping ready-to-use project files; built-in Skills and reusable workflows.
Mistral Vibe shipped as a long-horizon productivity + coding agent, bundling Work/Code modes, a CLI, and a new VS Code extension.
How’d this digital dip taste?
This was it. Our fifty-six digital dip together. Forward this to someone who thinks using AI and building AI capability are the same thing.
Is your organization learning fast enough? Are you still testing AI around the edges, or are you starting to rethink the work itself? And maybe even more important: are you building the capability to use AI on your own terms? Because the gap is no longer only about technology. It is about readiness. Ownership. Control. This is exactly where I help organizations. Turning AI from loose experiments into real capability. Helping teams understand what is possible, where to start, and how to build the confidence to move. Need help closing that gap? Just reply to this email.
Looking forward to what tomorrow brings! ▽
-Wesley
