Google Just Killed OpenClaw & Hermes

0:00Google just released a ton of products at their IO event yesterday and in this video we're going to be covering them all and why builders should be building with these new models that just came out because this is just more than just a model drop. We need to be paying attention because if you're building with any type of AI agents, whether open claw, Hermes, these kind of drops are tools that could really help advance your workflows. So, you know, they're really basically trying to own the AI agent layer. Google has been silent for the last few months and they've been cooking up a bunch of stuff with the new anti-gravity release, the new Gemini flash. And if you're building, like I said, with open claw or Hermes or Claude code, Codex, cursor, or any agent framework right now, you need to understand what Google just announced

0:45because this was not just another simple launch of whatever. It actually cost $75 million to even run this event. So, they had a ton of headline features. You know, yes, they announced the new Gemini 3.5 series. Yes, they showed the new cool Omni video nano banana feature, which is amazing. And they also showed a ton of new developer tools. There were a lot of shiny keynote words in this keynote as well. So, you don't want to be like shocked and surprised by just a simple demo, but buried inside in the developer keynote was a real shift. Google said specifically, we have moved from AI that simply assists you with AI agents that can independently navigate complex tasks across your entire

1:31workflow. So, that is cool and that sentence is basically the whole story. The biggest companies in the world are no longer pitching AI as just a smarter auto complete. They're pitching AI as a worker that can actually move through tools, use the tools, run the tasks, use the sandboxes, touch code, test apps, migrate projects, and inspect browsers, and then deploy software. Real quick, before we keep going, if you're watching this and you want to actually build with some of these tools, not just watch videos about them, then you're going to want to check out our community down below, Shipping School. We have a full Cloud Code course, a full Open Cloud course, and four live bootcamps every single week, where we actually help you

2:16get set up from scratch. Like, actually set this thing up, not just watch a tutorial and figure it out by yourself. And we also provide one-on-one coaching, so you could book a call with me, we could share screens, and I can help you get Cloud Code or Open Cloud running on your machine. That's it, no fluff. I built this community because watching YouTube's only gets you so far. We launched it just 3 days ago and we have over 55 members. >> [music] >> You need people around you who are actually building, people who hold you accountable, and coaches who can help you when you get stuck. I'll put the link in the description down below. Get in now before the price goes up. So, that is the same wave Open Cloud has been operating on for the last few months, and that is the same reason Hermes Agent is blowing up, and that is

3:01why Cloud Code matters, as well, and that's why Codex matters, and that is why the market is converging on the same idea. Agents are becoming the new software interface. So, what did Google actually announce? Like, there was a lot. The first big thing is Anti-gravity IDE, and this is 2.0. And the new Anti-gravity CLI, which is pretty cool. Google calls Anti-gravity its first agent platform. With the new version, they say builders can spin up specialized sub-agents to tackle complex workflows. And they wrap that with a cross-platform terminal sandboxing, credential masking, and hardened Git policies. That last part is kind of

3:46boring, but I mean, but that last part is the product. Because anyone can make a demo where an agent writes code. Anybody. You could download a local LLM and start writing code today. The hard part is letting an AI agent touch a real machine without destroying your repository or even leaking environment variables or making changes that you can't review. Google understands this. That is why sandboxing, credential masking, and Git policy are right next to the agent orchestration pitch. And this is where I think open claw users really need to be paying attention. The future of sub-agents is not just better reasoning, it's better containment, better permissions, better hand-offs, and just better overall audit trails, especially better recovery when

4:31something breaks. If you are building a business on top of AI agents, reliability and control matter more than one magical answer. The big piece is managed agents in the Gemini API. Google says managed agents remove the friction of infrastructure, the setup, and give you the power of anti-gravity agent harness through managed agents. So, a single API call gives you a fully provisioned agent with a remote sandbox. And that's a huge one. Not because every builder should use it for everything, but because it shows where the platform is going. Open AI has hosted agents, Anthropic has managed agents as well, and GitHub has co-pilot cloud agent. Now, Google is saying, "Here's a managed

5:16agent in your Gemini API. One call, remote sandbox, go build." This is the convenient side of the market. You don't want to set up a Mac mini. You don't want to have to manage the gateway all the time. You probably don't want to configure Docker, and if you've done it before, you probably know, "Wow, that was a pain." You just want to hand a task to an agent and get the work right. For a lot of companies, that is going to be very attractive. But, Google also announced the anti-gravity SDK, and that is the other side of this. They said you could host your own anti-gravity agent and get programmatic control over the harness, customize the agent, and deploy it on your own infrastructure. That tension is super important. Managed

6:01agent when you want speed, self-hosted harness when you want more control. That is the same conversation we've been having with Open Claw. Cloud convenience versus operator ownership. There is no perfect answer here. If I'm testing a throwaway coding task, a managed sandbox is great. But if I'm connecting my business email, my customer data, revenue numbers, calendar files, private workflows, I want way more control than that. That is why Open Claw still matters. That is why local first agents still matter. But if you should really not ignore the managed side because managed agents will keep getting easier, cheaper, and more capable. The builders who win with both will understand both and use hosted agents for disposable work. Use self-hosted agents for the

6:46workflows that become part of your business operating system. The next part is probably the most important for web builders. Google announced Web MCP. Web MCP is a proposed open web standard that lets developers expose structured tools like JavaScript functions and HTML forms so browser-based AI agents can execute complex tasks with more speed, reliability, and precision. And that is a huge deal because right now browser agents are often guessing. You know, they'll look at the page. They'll click around at buttons and they'll read labels. They'll try to infer what is happening, but the DOM or the screenshots, it's just not accurate. It works sometimes. It fails most of the time. And when it does fail, it usually fails in the most annoying way possible.

7:31Web MCP is basically Google saying, "What if websites exposed to already agent actions direct, not just human UI?" Machine usable functions. So if this takes off, it changes how builders should think about website. Your app is not only for humans anymore. Your app is also for AI agent. That means the best websites might need clean actions, clean forms, and a clean state. Great accessibility as well. Predictable tables and structured capabilities an agent can actually call. This is not only an AI story, this is a product design story. If you're struggling to keep up with content, well, I'm about to save you about 40 days worth of work. [music] I built something called Content Machine. It's 10 AI agents that run on the open claw orchestration, and they

8:16handle everything. Scripts, thumbnails, X [music] posts, blogs, outreach, clips, newsletters, all of it. So, I went from 1,000 subscribers to 4,000 subscribers on YouTube in 7 days using this exact system. Every single morning I wake up, and the content's already done. I spend [music] maybe 15, 20 minutes reviewing and approving them, and I move on with my day. [music] It works for any niche, fitness, finance, real estate, marketing, [music] whatever you are building, and it is 100% completely customizable to your use case. So, you get the mission control dashboard, all of the cron jobs, everything I've built over the last 40 days helping me gain more and more people to subscribe and join the community. So, you plug in your

9:02own thing, and it molds it to you. It learns how you talk, and it writes so it doesn't sound like AI slop. $97 one-time. It's not a subscription. [music] I'll put the link down below, and you'll thank me later. If agents become a major way people interact with software, then agent readable interfaces become a competitive advantage. Think about that for a second. If two SaaS tools do the same thing, but one is easy for an agent to operate, and the other one is a mess, which one gets used inside of automated workflows? The agent-friendly one. This is why I think Web MCP is worth watching, not because you need to rebuild everything today, but because it points to where the web is going. Then also, Google announced Chrome DevTools for agents. This one is super practical. They are bringing

9:48Chrome DevTools capabilities to AI agents, so agents can then verify, debug, and optimize the code in real time. They mentioned quality audits, uh basically emulating real-world user experiences, session handoffs with autoconnect, and more without manual oversight. That is exactly the kind of tool agents need. Not because like it's for fun, but because actually the problem with agent-built apps is not usually that the first code never appears. The problem is verification. Does the button actually work? Does it work on mobile? Does the page load fast? Does the form submit? Does the modal trap focus? Does the checkout flow break after login? A human developer uses DevTools to figure that out. Now, Google wants agents to use DevTools directly.

10:33That is the correct direction. If your agent can inspect the browser, watch the network calls, check the performance, and verify the behavior, it is guessing. And guesswork is what makes AI-built apps feel fragile. This also connects to Openclaw. Openclaw builders should not only ask, "Can my agent write code?" Ask, "Can my agent verify the result? Can it open the browser? Can it test the form? Can it see the error? Can it compare screenshots?" And can it tell when something is actually done? That is where the money is being made. And the next announcement is modern web guidance. Google says it is set on expert-vetted skills for coding agents. Launching in early preview with support for over 100 use cases. It integrates

11:19with baseline and can be installed with one click in antigravity or via the CLI. That sounds small, but I think it's very important because the agent problem is not only the model intelligence anymore. It is instruction quality. A raw model might know a lot, but a skill that is packaged best with the best practices in mind for a specific workflow, that's what wins. Open claw people already understand this. Skills are how you turn repeated knowledge into reusable behavior. Google is now bringing that same idea into web development. Expert vetted skills for agents, that is the playbook here. So, if you are a builder, you should be thinking about your own skills, your own rules, your own procedures, and your own Q&A checklist. Basically, think about how your own prompts and, you know, and how they work

12:05for repeatable workflows. That is how you compound this thing. Then, there is the Android side of things. Google announced a stable Android CLI that lets AI agents tap into the Android Studio power. It could download the Android SDK, run apps on Android devices, and help create high-quality Android apps using any agent LLM and tool of your choice. They also open-source Android skills for complex workflows like migrating to Jetpack Compose or Jetpack Navigate 3. And, they previewed a Android Studio migration agent that can migrate an app to a native Kotlin Android app, even if the source is React Native, web, or iOS. Google says migrations that took weeks could become hours now. That is the kind of claim

12:50people will test immediately. But, even if it's imperfect, the direction is obvious here. Agents are moving from writing snippets to operating full development environments. Download SDK, run the device, migrate the architecture. It could test the app, use the skills, and that is basically a worker, not a chatbot window anymore. So, what does this actually mean for Open Claw and Hermit? Hmm. I think Google is not replacing them directly, not yet anyways. Google is building the infrastructure layer around agents, models, and managed sandboxes. Basically, a harness, a CLI, and an SDK. It has browser standards, dev tool access, web guidance skills, Android skills, and that is a full stack move here. Open claw is getting stronger as personal operating system. Hermes is

13:35still exploding as a self-improving personal agent with real community workflows. And Claude code and Codex are still strong for coding. But Google is trying to make the environment itself agent native. And that is the biggest threat. If the browser becomes agent native, if Android Studio becomes agent native, and if Google AI Studio becomes agent native, and if Gemini API gives managed agents with one call, then the agent layer becomes normal, not niche and not weird, but actually normalized. And that is good for builders who are already learning this stuff because most people are still debating whether AI agents are useful. Meanwhile, Google is designing the web, browser, Android, and cloud stack around them. My advice is super simple here. Watch webMCP. If you

14:21build websites or SaaS tools, start thinking about whether an agent could operate your product cleanly. Watch Chrome devtools for agents. If you are using any coding agent, verification is the next bottleneck. Watch managed agents in the Gemini API. Use hosted sandboxes for work that is temporary, isolated, or easy to review. And watch anti-gravity SDK because self-hosted harnesses are where serious operators will still want control. The mistake would be turning this into Google versus Open Claw. They're all working together. This is too simple, right? The real story is that the entire software stack is being rebuilt for agents. Open Claw gives you an operator layer. Hermes gives you a fast-rising rival with persistent memory and a growing use case

15:06library. Google is pushing agent native infrastructure into browser, Android, cloud, and developer tools. Those things can compete, but they can also feed each other. A smart builder does not worship one tool. A smart builder asks what part of my workflow gets better because of these tools exist? For me, like the takeaway is super simple. Like if you're using AI like a smarter search box, you're already behind. The big players are building agents that can run real workflows. They open-source the projects. They are building agents that can live with you and the browser is becoming a tool surface for it all. The IDE is becoming the agent cockpit and the web itself might start exposing actions for agents to call. And that is not a small update. That is the next interface shift here. So, don't ignore

15:52it. Like pick one workflow this week, make it agent-friendly, and then you're going to add a scale, you're going to add verification, add approvals, and make it repeatable. Because the builders who learn this now are going to look insanely far ahead this by this time next year. If you need help getting your first AI agent up and running, join us inside the community. We have over 150 builders in there. We do unlimited one-on-one coaching calls. And we also do seven live boot camp calls a week. We'd love to see you there, and we'll see you in the next one. Have a blessed day.