I was in Hong Kong this week for FCM Wallet Launch Event. Huge thanks to @fcmoffice , @NonsyncA and @CoinRank_io for giving me the opportunity to speak at this amazing event. @fcmoffice wallet is currently in the internal testing phase and will soon be available on app stores.

What makes the $ZIG model interesting is how the different parts connect. Capital flows into vaults and earn products, ZIG Markets generates revenue, ZIG Chain adds another revenue layer, and a portion goes back into open-market buybacks. If adoption grows, the whole flywheel has room to scale. That’s the part worth watching.

What really got me is @nectarsocialai being able to pick up brand mentions directly from videos by analyzing the audio and text on screen, even when nobody tags you. If you worked as outreacher or growther, you know how many opportunities get missed simply because you never see them in the first place. This is the kind of tool that makes a lot of sense once you understand the problem it's solving. So this is amazing product especially for the people like me

As Elon has said, until we give AI products a physical form, they’ll remain products trapped behind a screen. That’s why we’re seeing more progress toward making AI tangible. I shared one example of this on my profile yesterday named @roboticscorp.

One thing that’s always bothered me with AI coding tools is having to choose between model capability, speed, cost and privacy. @Recursive_SI’s Flint feels like a step toward solving that problem. A 40B coding model built to run locally, with impressive inference speeds and open weights. If this keeps improving, the way we build software could look very different. Worth keeping an eye on what @RichardSocher and the team build next.

AI made creating visuals easier, but somehow made everything look the same. @anurahk and @cocoailabs are tackling the actual problem: designs built around your brand, with real assets and editable layers. AI should make design more original, not more generic that's what i want and will definitely use it for our designs:

People who use AI agents for coding know how frustrating it can be when you're waiting on model responses instead of getting things done. I think one of the more overlooked parts of inference is that generating tokens isn't just about compute. moving model data fast enough matters just as much. that's why i like what @general_compute and @positron_ai are doing here. Atlas first, then Titan, with inference capacity available through General Compute from Q1 2027. Using different chips for different stages of inference makes a lot of sense, especially as coding agents handle longer, more complex tasks with more model calls along the way.

I didn't think a robot's personality would matter this much to me, but here we are. @make_robot_go and @roboticscorp are building Voxel, a droid butler designed to actually feel at home in your home. not just what it can do, but how it feels to live with it. The little gentleman energy is a pretty good touch. finally, a robot i'd actually want around not a cassandra type :)..

When i ask the devs on our team about the biggest challenges they face, one thing that comes up often is how a design can look right in 3d but still not be buildable in the real world. complex products aren't just 3d shapes. every component needs to fit together, follow mechanical constraints, and be built in the right sequence. that's why @skyfallai caught my attention. SKY-001 breaks complex objects down into individual components and builds them back up into a complete CAD model. the reported results are pretty wild too: 100x faster and 150x cheaper than GPT Astra on CAD and BIM tasks. if world models can eventually help automate the process from design to production, that's a much bigger deal than just generating 3d models.

The strongest tokenomics are backed by real business, not just hype. @ZIGFinance has now acquired 5.37M $ZIG through revenue-driven buybacks, while expanding real-world finance partnerships and distribution:

If you've been sitting on an idea because you didn't want to spend time or money figuring out how to build it, this is a good chance to give it a shot. @boltdotnew is free this weekend, and there's $10K in prizes for builders too. Might as well stop overthinking it and see what you can actually achieve in 2 days;

Anyone who has labeled video, reviewed site photos or checked footage by hand knows how much visual work is still manual. @ElorianAI’s first model is built for this kind of work: reasoning over images and video, not just describing what’s in the frame. Robotics, factory inspection, job sites and aerial imagery are among the use cases it lists. The waitlist is open, with free chat access and a paid API: elorian.ai/ @AndrewDai @dustinvtran @ElorianAI

There is a huge innovation for filmmakers. VN Studio @buildwithVN has just been launched by @nelsonchu_ and @varien. VN Studio is the first AI production platform built around the responsible use of IP and human authorship. You can develop characters, wardrobe, and environments; compose scenes and shots; generate and compare video takes; work through post-production; and prepare a scanned final. It has an AI Producer, which works with the context of your project, helping develop ideas, organize production, and carry out supported actions within defined spending controls. If you are interested in this, definitely check out their website here: vn.ai

Annoying thing about stock photos is finding one that's almost perfect, except for that one detail you can't change. and generating everything with AI isn't always the answer either. you end up tweaking prompts forever trying to get the exact result you had in mind. @envato's approach makes a lot of sense. start with one of 29M+ human-made assets, then use AI to change the background, remove objects or adapt it for a completely different campaign. way better than spending an hour trying to generate the same thing from scratch.

Most businesses have already moved online. The way we interact with them hasn't caught up. You can browse a website, compare options, and find exactly what you need, only to hit a wall when it's time to actually get something done. Call a number, navigate a menu, wait for someone to pick up. @usebland’s Blandie points toward a different model: a simple text conversation that can access live website information, take action, process payments, and bring a human into the conversation when needed. The important shift is from AI that merely answers questions to AI that can help complete the task itself. If that experience becomes seamless, messaging could become much more than another customer support channel. It could become the interface between people and businesses.

AI agents are getting better at doing things for us. but autonomy is only useful if we can trust it with access to the real world. payments and logins are a major missing piece. giving an agent your credentials creates obvious risks, while locking those capabilities to one platform defeats the point of having multiple agents. @selfxyz is tackling this with Agent Vault: one portable layer for payments and logins across agents, with spending limits, approval rules, and credentials kept out of the agent's view. the future isn't just agents that can do more. it's giving them the ability to act without giving up control.

The interesting part isn't just a $4B+ stablecoin. It's what happens when stablecoin distribution starts generating real revenue. $USD1 is moving into payment finance through ZIG Markets, with the first vault already live. Distribution → TVL → Revenue → $ZIG buybacks. That's the model to watch. If adoption translates into sustainable revenue, the connection between real usage and token value becomes much more tangible. @ZIGFinance 🤝 @worldlibertyfi

The real value of AI isn't in giving people better answers. it's in removing the need to manage everything themselves. travel is a perfect example. finding options is easy. using loyalty points, navigating 100+ integrations, securing better rates, making bookings, and monitoring price drops or cancellations requires much more than a chatbot. @OdessiaTravel brings it together in one conversation. it can use your Hilton, Marriott or Hyatt perks, access benefits across 2,000 top luxury hotels, and find smarter routes that can cut trip costs by up to 40%. the bigger idea is what interests me: AI moving from a tool we operate to a system that acts on our behalf. the next generation of AI won't just help us think faster. it'll change how much of life we actually have to manage ourselves.