The Lobster That Started It All
In November 2025, an Austrian developer named Peter Steinberger published an open-source AI assistant called ClawdBot. It had a lobster mascot named "Clawd," a playful nod to Anthropic's Claude. The pitch was simple: an AI that actually does things. Not just chat. Not just answer questions. An AI that manages your calendar, controls your browser, handles your files, and responds to messages on your behalf.
It went viral. Developers loved it. The community grew fast. And then Anthropic noticed.
Act 1: The Name That Couldn't Stay
On January 27, 2026, Anthropic sent a polite but firm trademark request. "ClawdBot" was too close to "Claude." The name had to go.
Steinberger renamed it to Moltbot the same day. A lobster molting its shell to grow. Clever branding. But the rename created chaos. Scammers launched impersonation campaigns within hours. Malwarebytes documented the attacks. The community was confused about which download was real.
Two days later, on January 30, Steinberger renamed it again. OpenClaw. "Moltbot never quite rolled off the tongue," he said. This was meant to be the final name. A stable reset.
Three names in three months. Same project. Same developer. Same AI that actually does things.
Act 2: The Poach
On February 13, 2026, we deployed OpenClaw on a cheap VPS and connected it to Telegram. It worked. We could text our AI from anywhere and it would respond, search the web, and handle tasks on demand.
Two days later, on February 15, Sam Altman posted on X: "Peter Steinberger is joining OpenAI to drive the next generation of personal agents. He is a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people."
The creator of the tool that Anthropic forced to rename was now working for Anthropic's biggest competitor.
Steinberger said he could have turned OpenClaw into a huge company, but "it's not really exciting for me." Teaming up with OpenAI was "the fastest way to bring this to everyone." OpenClaw would continue as an open-source project inside a foundation.
The irony was hard to miss. Anthropic killed the original name. OpenAI hired the person who built it. And the open-source community was left holding a project whose creator had moved on.
Act 3: The Acquisition
While OpenAI was hiring indie developers, Anthropic was buying startups.
On February 25, 2026, ten days after Steinberger joined OpenAI, Anthropic acquired Vercept. Vercept was a Seattle-based startup founded by alumni of the Allen Institute for AI. The team of nine people, led by co-founders Kiana Ehsani, Luca Weihs, and Ross Girshick, had built a macOS application called Vy that used AI to understand on-screen content and complete tasks through natural language commands. The technology worked locally, without plugins or cloud processing.
Vercept had raised a $16 million seed round in January 2025 at a $67 million post-money valuation. Fifty Years led the financing. Angel investors included former Google CEO Eric Schmidt, Google DeepMind chief scientist Jeff Dean, Cruise founder Kyle Vogt, and Dropbox co-founder Arash Ferdowsi.
Anthropic's announcement said the acquisition was to advance "Claude's computer use capabilities." Vercept's product Vy was shut down within 30 days. The entire team joined Anthropic.
Two different strategies for acquiring AI agent talent. OpenAI hired the indie developer. Anthropic bought the venture-backed startup. Both happened within ten days of each other.
Act 4: The Avalanche
Then Anthropic started shipping. Fast.
February 24, 2026: Claude Code Remote Control. Start a Claude Code session in your terminal, continue it from your phone. No port forwarding, no VPN. Native streaming connection. Exactly what developers had been using OpenClaw for.
March 19, 2026: Claude Code Channels. Official Telegram and Discord integration. Install a plugin, connect a bot, and Claude Code responds to messages from your phone. Research preview. The exact use case we had been running on our VPS for a month with OpenClaw, now available as a first-party feature.
March 23, 2026: Claude Cowork. Full computer use. Claude opens your apps, navigates your browser, fills in spreadsheets. Anything you would do sitting at your desk. 18.4 million views on the announcement post within hours. macOS only for now. Claude Sonnet 4.6 scored 72.5% on the OSWorld benchmark for computer use, up from under 15% in late 2024. Anthropic has not confirmed a direct connection between the Vercept acquisition and Cowork, but the timing speaks for itself: 26 days between the acquisition and the launch.
March 23, 2026: Claude Dispatch. The remote control layer for Cowork. Dispatch creates a persistent conversation between the Claude mobile app on your phone and the Claude Desktop app on your computer. You assign a task from your phone, Claude executes it on your Mac. Available to Pro and Max subscribers. The computer must be awake for it to work.
Get the Weekly IT + AI Roundup
What changed this week in NinjaOne, ServiceNow, CrowdStrike, and AI. One email, every Monday.
No spam, unsubscribe anytime. Privacy Policy
Five major product launches in 28 days. An indie developer's vision, a startup's technology, and a $2.5 billion product line, all converging on the same idea: AI agents that do real work on real computers.
What We Were Actually Doing Through All of This
On March 22, the day before Cowork launched, we set up Claude Code Channels on the same VPS that was already running OpenClaw. Same server, same $7/month, new tool.
The setup was not smooth. The official docs made it look like six steps. It took considerably longer. Plugin installs that fail from the command line. OAuth tokens that have to be generated on the server, not locally. Permission prompts that freeze the bot when nobody is watching. It took a full day of troubleshooting to get it running reliably.
But it works. We text our Telegram bot from our phone and Claude responds. We forward a resume PDF and get back a structured analysis in 60 seconds. We ask it to check if services are running and it runs the commands on the server and tells us the results.
The same thing we were doing with OpenClaw in February. Now with Anthropic's official tooling.
Here is the comparison that matters: Dispatch lets you text Claude from your phone and it does work on your Mac. A self-hosted setup lets you text Claude from your phone and it does work on your server. Dispatch requires your computer to be awake. A VPS runs 24/7 for under $10/month. Dispatch is zero-setup. Self-hosting takes a day of configuration. Different tradeoffs, same result: AI that works while you are away from your desk.
The Pattern Nobody Is Talking About
Here is what happened in four months:
- ●An indie developer builds something people want
- ●The platform owner kills the brand
- ●A competitor hires the creator
- ●The platform owner acquires a $67M startup doing similar work
- ●The platform owner ships their own version of everything
Two talent acquisition strategies running in parallel. OpenAI hired the indie developer who built the agent people loved. Anthropic bought the funded startup that solved the hard perception and interaction problems. Both happened in the same ten-day window in February 2026. Both fed into products that launched in March.
This is not new. It happens in every platform ecosystem. But the speed is new. Four months from indie project to corporate product. Sixteen million dollars of venture funding absorbed in an acqui-hire. In the same timeframe, Claude Code went from a developer tool to a $2.5 billion annualized revenue product with 29 million daily VS Code installs.
The AI agent space moved faster between November 2025 and March 2026 than most industries move in four years.
What We Actually Learned
The tools change. The skills don't. We built a resume authenticity scoring framework that works on OpenClaw, Claude Code Channels, or a plain Claude chat window. The framework scores resumes on five categories: specificity, judgment, failure-recovery, language fit, and uniqueness. It does not care which tool runs it. We also built a trading analysis framework using the same approach. Skills are portable. Tools are temporary.
Self-hosting gives you control. Managed services give you convenience. A self-hosted setup requires more configuration but lets you run custom skills, control the model, and keep data on your own server. Claude Dispatch requires clicking a button. Both get you an AI on your phone. There is room for both.
The real moat is not the tool. Anyone can set up Claude Code Channels. Anyone can use Dispatch. The value is in what you build on top of it. The scoring frameworks. The automation workflows. The skills that encode your domain expertise into something an AI can execute consistently.
If you are waiting for the "right" tool, you have already missed three generations. ClawdBot, Moltbot, OpenClaw, Remote Control, Dispatch, Channels, Cowork. Seven names for the same idea in four months. Pick one. Build something. The tool will change. The skill will not.
Where It Goes From Here
Cowork is macOS only right now. Dispatch is available to Pro and Max subscribers but requires the computer to stay awake. Channels is still a research preview. OpenClaw is open-source but its creator works at OpenAI now. Vercept's product is gone, absorbed into Anthropic. The dust has not settled.
But the direction is clear: AI agents that do real work on real computers, controlled from your phone, running 24/7. Whether that runs on Anthropic's infrastructure, your own VPS, or something we have not seen yet does not matter. The era of "chat with an AI" is ending. The era of "work with an AI" has started.
We have been running it since February. If you want to start, the tools are ready. The question is what you will build.
*Want to set up your own always-on AI assistant? Download our VPS setup guide or read the technical breakdown.*
*For pre-built automation skills and workflows, check out our Pro Tracks.*
*Not sure where to start with AI? Take the free quiz and get matched in 2 minutes.*