At a lobster-themed event for AI lovers, happiness with a side of cocktail sauce
After Anthropic strongly suggested that Steinberger change its name to avoid any legal issues, the project kept its lobster-themed legacy and eventually settled on the OpenClaw moniker. The software has skyrocketed in popularity over the past few months, and several ClawCon attendees who started using it in January called themselves “veterans.”
The software serves as a bridge between today’s powerful AI systems, such as Claude or the OpenAI GPT family of models, and the real-world tasks that people actually want AI systems to accomplish.
After setting up their OpenClaw agent, either on a laptop or through a virtual server, users can send text or WhatsApp messages to it, directing it to perform various tasks within the wheelhouse of today’s AI systems. For example, users say they tell their OpenClaw agents to listen to their favorite podcast episodes and send keynote summaries to users’ inboxes, negotiate with car dealers on the price of a new car, and even order and pay for groceries, all without direct human input.
Many of the ClawCon attendees had signed up for the event after catching a whiff of the seafood of this advanced and hands-on OpenClaw. The meeting, which served as a powerhouse meet and greet, featured a number of main stage presentations, a rap performance, an open dance floor and – at the top – a less crowded VIP area with live streaming of the event taking place one floor below.
“There’s a kind of electricity and energy that you can just feel in the room,” said Thomas Taylor, ClawCon showrunner and organizer. “OpenClaw has always been the go-to source for personal AI programs, and I think personal AI will be incredibly important to the evolution of all AI.” Taylor used his OpenClaw system to help organize ClawCon and communicate with vendors.
Designed to be accessible to anyone, OpenClaw can be used with paid AI systems from OpenAI and Anthropic or freely downloadable AI models, many of which come from Chinese companies such as DeepSeek or Alibaba. Agents can also teach themselves new tasks and keep detailed notes about user preferences, allowing agents to practice user preferences over time. OpenClaw itself relies on a small army of volunteers to maintain its code, respond to user issues and patch any security bugs.
One of these volunteers, Vincent Koc, emphasized that the technology is still in its infancy, although it is already having a real impact on many experienced coders and novice developers alike.
“We’ve had the personal computer era as well, but now it’s with real personal AI systems,” Koc shouted over the roar of the party. “I hear stories from mothers, artists and everyday people who can really do things with AI. And I think that’s magic.”
As the deep techno beats of a DJ shook cups of cocktail sauce at a nearby table, Koc, a software engineer by day, ushered hundreds of OpenClaw students onto the dance floor and argued that excitement was more than just a passing fad.
“I believe in this very much. I will die on the sword for this,” said Koc. To help meet his tax liability earlier this year, Koc instructed his OpenClaw agent to find an accountant and request quotes. “The program sent emails to many different tax attorneys, and they came back to me with actual quotes for their work.”
Yet many in the male-dominated crowd were less trusting of systems, seeking fame – the ability to perform rational actions without human supervision – may also be its Achilles’ heel, or its closest crustacean equivalent.
The free-wheeling nature of OpenClaw programs recently made headlines after Summer Yue, lead AI security researcher at Meta, nearly lost her inbox to her OpenClaw agent. Because OpenClaw can be linked to personal email or financial accounts, weaknesses in the system could easily expose sensitive user data to hackers around the world.
“These systems are not mainstream,” Koc said, referring to the masses of everyday people who are less familiar with advanced AI techniques and AI in general. “You actually have an AI that literally takes over the machine. That might sound difficult, because you’re giving it access to information. But people have to use their brains. Take baby steps with this.”
As ClawCon prepares for future stops in Austin, Tokyo, and London, even the most enthusiastic crowd acknowledged that this technology comes with significant risks.
“In Claw we trust!” said Mark Mollé, a lawyer who works in psychology, watching the scene from the second floor while proudly holding up a lobster-shaped necklace. “At least, until the AI Hindenburg.”
“We see people blindly trusting untested and unsafe agent tools, which will continue until there is some kind of disaster,” said Mollé. Several participants revealed that they had set up cryptocurrency accounts for their agents and asked them to try to make money on prediction market websites like Polymarket.
Downstairs, after the big talks were over and the chrome-glad guitarist started the stage, the catering staff engaged in creative discussions about work flow and guardrails, trying to find the owners of the leftover lobster tails.
