New Claude Code Channels DESTROYS OpenClaw

The new Claude Code Channels Destroys OpenClause. So let me just tell you what happened. This morning Anthropic dropped something that nobody saw coming. And I mean nobody.

They just gave you the ability to text your AI coding agent from your phone through telegram, through discord, without touching your computer at all. And before you scroll past this cool feature thinking, oh okay that's nice, whatever. I need you to understand what this actually means because this isn't a feature. This is Anthropic drawing a line in the sand and telling OpenClaude the most viral piece of software in the history of the internet.

We can do everything you can and we can do it safer and we can do it better. That's a game that's really being played here and most people don't even know they're watching it. Let's back up for a second. So right now in the AI world there are basically two tools that developers and regular people are fighting about right.

Two tools that have basically taken over every tech conversation for the last few months. The first is Claude Code. That's Anthropic's tool. Lives in your computer terminal.

You talk to it. It writes codes, fixes bugs, builds apps and edits your entire code base whilst you sit back and watch. It's been on a tear. As of February 2026 Claude Code hit a 2.5 billion annualized run rate.

A number that more than doubled since the start of the year. Weekly active users doubled since January. It's surged to 29 million daily installs inside Visual Studio Code and about 4% of all github commits worldwide are now written by Claude Code. Read that last number again.

4% of everything on github written by one AI tool in less than a year. The second tool is OpenCore and OpenCore is something different. OpenCore is wild. OpenCore is chaos and is now basically broken the internet.

OpenCore crossed over 300,000 github stars surpassing React to become the most starred non-aggregator software project on github. React took about a decade to get there. OpenCore did it in about 60 days. That's not normal.

It's a trend line. It's a phase change right. OpenCore started as a weekend project then it went viral and you've got this wild chaotic viral tool the most popular open source software in the history of github now being backed by OpenAI and it's growing by the day. Now the thing is people were saying OpenCore could do things that Claude Code couldn't right.

So OpenCore lives on your phone, you message it like a friend, you can do it through whatsapp, through telegram, through discord, through iMessage. Runs 24 hours a day, seven days a week in the background and remembers things from previous days and weeks. It reaches out to you when something needs your attention. It's always on, always watching, always working.

Claude Code lived in your terminal. You had to go to it, you had to sit at your computer, open a command line and interact with it like a developer. If you stepped away the session ended. If you wanted to check on a task from your phone you couldn't, not easily anyway.

That was the gap, that was the argument. OpenCore is a tool for regular people, Claude Code is a tool for nerds who live in their terminal. And then this morning happened. Anthropic released Claude Code channels and research preview, a new feature that lets users send messages to Claude Code directly from telegram and discord.

That's it, that's the move. Simple, clean and devastating. One tweet from Tariq, the Claude Code product manager, with 440,000 views in hours. You can now message your Claude Code agent from your phone, from anywhere, without touching your computer.

You set it up through telegram or discord using MCP servers. That's the technical bridge that connects them and from that point forward your AI coding agent is available in the chat window, the same way you'd message your co-worker. And Tariq explained the angle clearly. He said they want to give people lots of different options to talk to Claude remotely.

This is more focused on developers who want something hackable. So let me translate that from tech language to human language. Hackable means you can customize it, you can build on top of it, you can connect it to your own workflows, your own tools, your own systems. It's not a locked product with guardrails and preset flows, it's infrastructure you can shape.

That matters more than it sounds because one of the biggest criticisms of enterprise AI tools is that you're renting a product someone else controls. With Claude Code channels for example, Anthropic is saying here are the pipes, build whatever you want with them. Now let's talk about what makes this announcement especially interesting because this isn't just about convenience, this is about a power struggle that's been building for months. OpenCore has been selling itself on three things.

First, it's free, or at least the software itself is free. You pay for the AI models underneath it. If you run local models on your computer, for example like Kimi K25, then you can run OpenCore with zero recurring task, sorry cost, to any AI company at all. Second, it runs everywhere, so messages, whatsapp, telegram, discord, slack, signal.

It connects to over 30 different platforms. You can reach it from your platform, from your phone, your laptop, your smartwatch, if you really wanted to. And third, it does live stuff, not just code. It will manage your calendar, triage your emails, monitor social media, post updates, summarize your inbox every morning, and control your smart home device, personal life essentially, operating system, not just a coding tool.

Claude Code has historically lived in category number three, which is code. No live stuff, no scheduling, no messages. You point at a code base and it works, you step away and it stops. But look at what's been happening in March 2026 alone, especially from Anthropic.

So Claude Code received massive updates, pushed to talk voice mode, a 1 million token context window, the forward slash loop command that Claude Code becomes a recurring monitoring system with. You set an interval on a task and Claude can run it automatically on a schedule like a lightweight cron job. The forward slash loop command. Now you can tell Claude, check the deploy status every five minutes or check the latest AI news every hour and it will actually just go off and do it.

And that's what OpenClaude has been selling, background information, persistent scheduled work that happens whilst you're living your life. Claude Code now does it too. Then came remote control. Anthropic introduced remote control, allowing developers to continue a local coding session from a phone, tablet or any web browser.

On February the 25th, they shipped the features developers have been waiting for. Kick off a coding task in your terminal, put your laptop in your bag and keep full control from your phone. So in the span of about three weeks, Claude Code gained voice mode, a million token context window, loop commands, remote control and now channels so you can text it from Telegram or Discord. One by one, they're basically closing every single gap that OpenClaude used to claim as its advantage.

But let's steel man OpenClaude for a second because they deserve a fair shot here too. OpenClaude isn't just a messaging interface, it's fundamentally a different philosophy about what AI should do. Model agnostic, meaning you can run it with Claude, with ChatGPT, with Gemini or completely free with open source models. You're not locked to one AI company, you're not paying a subscription forever, you own your setup and it's persistent memory, real persistent memory.

Not just context in a session, memory that lives across days and weeks. OpenClaude has persistent memory unlike Claude Code which resets memory between sessions. It can remember that you prefer certain restaurants on Mondays, it can remember you have a big presentation in three days and start prepping you, it can remember things about your life. So one story that went viral about Steinberg as an AI agent was receiving a voice message.

The system wasn't programmed to handle audio but instead of failing, the agent checked the file header, identified it as an Opus audio file, called a tool to convert it and then used the OpenAI API to transcribe it all without being asked. It improvised its way for a task it'd never been trained for. That's extraordinary, that's the kind of thing that makes engineers stop and stare. Real improvised problem solving for an AI agent and nobody told it to do anything.

So OpenClaude has real strengths, real genuine strengths that Claude Code hasn't fully matched yet. But here's where the Claude Code channels announcement gets really interesting. The reason OpenClaude exists, the reason it blew up the way it did, is that there is a gap between what powerful AI could do and where regular people could actually access it. People don't live in terminals, they live in chat apps, they live on the phones and they wanted the AI agent to live there too.

And that's basically what OpenClaude has solved, access. Nothing more complicated than that. Claude Code channels is basically Anthropic saying, we hear you, access coming. And here's the thing about Anthropic versus a community-maintained open source project like OpenClaude, especially when it comes to this race.

So security researchers discovered tons of security problems. You know, 135,000 exposed machines of OpenClaude, 50,000 vulnerable instances of OpenClaude to full remote takeover just by visiting a single web page. A security audit of Claude Hub, OpenClaude's community skills registry, actually found that roughly 12% of skills, about 341 out of 2,857, were malicious. And one in eight community plugins.

So that's malicious programs in one of eight of OpenClaude's community plugins and skills on a platform that people are using to access their email, their calendar, their home network, their entire digital life. A Meta executive even reported that her OpenClaude deleted her entire email, right? And these are real reports from real users. So this is the problem.

This is the trade-off, right? You've got freedom and flexibility on one side of OpenClaude and you've got safety and reliability on the other. And you might be there thinking, okay, I'm not a developer. I'm not running some entire enterprise code base.

I'm just a business owner or a freelancer or somebody who wants AI to help with my day. Why should I care? Well, here's why. The tools that developers use today become the tools that shape what you do tomorrow.

You know, when Claude Code gains the ability to run background tasks, schedule work, and communicate over Telegram and Discord, they don't just stay in the developer world. Those capabilities flow outward. So they become the foundation for the products regular people use. The automation tools, the business software, the personal assistants, the AI employees that founders are starting to hire instead of actual humans, right?

We are entering an era where AI tools are writing roughly 41% of all code. That number is going to keep climbing and the AI agents who write the code are going to start looking a lot more like co-workers who you message on Slack and text on your phone than tools you sit in front of at your desk. Claude Code is becoming that co-worker right now in front of your eyes. And let me tell you what the full picture of this March looks like for Claude Code specifically, because it's important to understand the pace, right?

It went from version 2.1.63 to 2.1.76. They added voice mode in 20 languages, a forward slash loop recurring task system, 1 million token context window, Opus 4.6 as a default model, and it's got the ultra think keyword for maximum reasoning and the forward slash effort command for adjusting how hard the AI thinks about a problem. And now you've got channels, Telegram, Discord, and from your phone. So they're not just casually shipping features.

This is a team in a war, a team that looked at OpenCLAW, looked at the headlines, and decided, right, let's take over again here, right? They can feel themselves falling behind, I would say. Now here's something worth pausing on, because there's a real difference in philosophy between these two tools. And it's not just about features.

Claude Code is built with the assumption that humans stay in the loop, right? Every significant action it asks, every command it runs, it shows you. It asks for permission before implementing commands. It sandboxes operations when possible.

It maintains strict separation between conversation context and system access. OpenCLAW is built with the assumption that the AI should run autonomously. It acts without asking. It makes decisions.

It improvises. That's what makes it feel like magic when it works. And that's also what makes it terrifying when it doesn't. So this is interesting.

One Reddit user said it perfectly. OpenCLAW is basically Claude Code without the corporate safety guardrails, you know, the ones that protect you from bad actors. So both philosophies have value. But for most people, most regular people, who aren't deep in the tech world, which one do you actually want touching your email, your files, your schedule, and your business?

The channel announcement tells you Anthropic is going, right? They want to give you the access and the convenience that OpenCLAW offered, but they want to do it in a system that has security reviews, enterprise support, dedicated engineers, and a company with a legal address you can hold accountable if something goes wrong, right? That's a different product than a weekend project with 1,000 volunteer contributors shipping code every week that sometimes breaks things and most people don't even check. So now let me talk about what that means for you if you're not a developer, right?

I want you to think about your job for a second. Whatever it does, you know, sales, marketing, customer support, whether you're running a small business, maybe managing a team, writing, designing, how much of your time is spent on tasks that are repetitive, you know, sending follow-ups, updating spreadsheets, scheduling meetings. If you could text an agent the same way you text a co-worker and it would handle those things, how much time would it save you? Because you can do that right now today with cloud code sessions, with channels, right?

It does it. It's not a demo, it's actually doing it today. And so I'll keep saying this, and I'm going to keep saying it until it's obvious, learning AI automation is not optional anymore. It's a skill that separates the people who thrive in the next few years from the people who struggle.

That's why I built the AI Profit Boardroom. Not to teach you how to be a developer, not to teach you how to write code, but to teach you how to actually use these tools, right? The way business owners and freelancers and creators and operators need to use them to get real work done faster than your competition. And every week we're going through the actual workflows that matter, the setups that work, the ways to build systems that run whilst you sleep.

If you're watching this and thinking you should probably learn this stuff, you're right. And the AIProfitBoardroom.com is where I'd start. Link in the comments description or go to the AIProfitBoardroom.com and come see what we're building. So back to the war between OpenClaw and cloud code.

So let's talk about OpenClaw's actual position after today. This announcement doesn't, you know, end it overnight. I'm being dramatic with the title and we all know it, right? OpenClaw is still growing.

As of the time I'm recording this, it has about 300,000 stars on GitHub. The contributor's active and the skill library keeps growing. What today's announcement does is close the access argument, right? The one thing OpenClaw had going for it with non-developers and regular users was you can use this whilst you're on your phone from normal apps without being a tech person.

And that was the amazing pitch. That's what Andrew Karpathy called the most incredible sci-fi takeoff adjacent thing. That advantage just got cut in half and OpenClaw still wins on day-to-day automation, right? Emails, reminders, calendars, things that cloud code was never built for.

It's also still completely free if you use local open source models and those real advantages aren't going away forever. But the narrative shifted. Anthropic just told the world we're in the phone and chat app game now. We're coming for the access layer.

When you're Anthropic, a company with billions in revenue, dedicated security teams, enterprise infrastructure, and the backing of Amazon and Google versus a community project maintained by volunteers, the race for the access layer is not even close, right? There's also the question of where does OpenClaw go from there? Remember, the project is transitioning to an open source foundation with OpenAI backing after Steinberger joined OpenAI on February the 14th. OpenAI backing an AI agent that runs on Cloud, that runs on Anthropic's models, and using that to find Anthropic's own tool.

If that sounds messy, it is messy. The AI world in 2026 is extremely messy, right? Companies are funding each other's competitors. Tools are running on each other's infrastructure.

The lines between enemies and partners blur constantly. But here's the thing about the dynamic. OpenAI backing OpenClaw gives it resources. It gives it credibility.

It gives it staying power. The project isn't going to collapse because a few security vulnerabilities got found. It's too big, too popular, too funded. And the narrative a few months ago was OpenClaw does things Cloud Code can't do, and regular people can use OpenClaw from the phone.

Today, the story is Cloud Code can be accessed from anywhere, right? Through Telegram, through Discord, the same apps OpenClaw has been living in. And Cloud Code comes with Anthropic's security, Anthropic's infrastructure, and Anthropic's engineers making sure your inbox doesn't get deleted. That's a different story, a significantly different story.

So we're in the middle of the fastest moving platform shift in the history of technology. Faster than mobile, faster than the internet, the platform that wins the agentic AI layer, the layer where AI agents do the work. That platform is going to define the next decade. Anthropic is responding.

Cloud Code Channels is not a small update. And behind Anthropic, you have the $4 billion Amazon investment, right? You have the Google partnership. You have the enterprise sales team.

You have the security infrastructure. And bear in mind as well, Claude's mobile app just soared to the top of the US App Store charts, overtaking ChatGPT after Anthropic refused to allow the Department of Defense to use its AI for domestic surveillance or autonomous weapons, right? So I'm not here to tell you to trust, that's your call, but I'll tell you what I see. The companies that treat safety as a feature rather than a limitation are going to win the mainstream market.

The mainstream market is not developers who can audit their own security setups. The mainstream market is just people who want the thing to work and not go haywire, right? So let's talk about what you should do and how you should actually implement this. If you're a developer, I would sign in to update Cloud Code right now.

You can run the NPM update and then just set up channels with Telegram or Discord. The setup guide is live on GitHub right now. You can actually use Cloud Code just to help you set up. If you're a non-developer, a business owner, creator, freelancer, etc, here's your move.

So start learning how to use these tools, not how to code, how to use them. There's a difference. You don't need to understand the technical setup to benefit from these tools. You just need to understand what they're capable of so you can start designing your workflows around them.

If you're a manager of a team leader, I would start thinking about how your team works. Start thinking about where the bottlenecks are. Where are your people spending time on things that AI could handle? Because you want to free up your people to do the things that AI can't do.

That's the main thing and the companies who figure this out now are going to have a huge advantage. Now the other thing to note here as well is like seven months ago, Cloud Code was a terminal tool that developers used to write code faster. Today it's an AI agent. You can text from your phone whilst you're on vacation.

It will monitor your systems, run scheduled tasks, communicate through Telegram, and keep working whilst you're away from your desk. Where is it going to be in seven more months? And I genuinely don't know. That's the thing.

I've been watching the space every single day and I just don't know. The pace is too fast, right? The moves are too unpredictable. The competition is too intense.

What I do know is this. Most people who wait to figure this out are going to look back and wish it started earlier, right? Not because the technology is going away, because the competitive advantage of understanding these tools narrows as everyone catches up. Right now there's still a gap.

Right now most business owners and most professionals and most freelancers are still on the sidelines. They're watching, they're waiting, saying maybe next year. And the people who are actually using these tools, building workflows, automating the workflows, sending messages to the AI agent from the phone, those people are quietly building a lead that's going to be very hard to close. That's the real story behind Cloud Code Channels.

Not just so you can text your AI now. Not just so OpenCloud has some new competition. The real story is this. The gap between people who use AI well and the people who don't is getting wider every week.

Every feature like this widens it further. Every feature like this is another way that AI users can offload work that non-users still have to do themselves. And offloading work is offloading time. And time is the only resource you can't make more of.

Come learn with us at the AI Profit Boardroom. We're implementing this every week. We're learning, we're growing, we're supporting each other, and we're learning how to actually implement this stuff inside our businesses. The lobster has some competition now.

Thanks for watching and I'll see you on the next one. Cheers.