Met the founder of OpenClaw at the hackathon: what else can lobsters do?

Original | Odaily Planet Daily (@OdailyChina)

Author | jk

In March 2026, the UK AI Agent Hackathon 2026, initiated by the Blockchain Association of Imperial College London, was held in London. This hackathon centered around OpenClaw as the core technology framework, attracting over 1,200 registered participants. On Demo Day, it set a record with 5,000 live online viewers, briefly topping the global trending charts on X platform.

Many participants regarded it as the “World’s First University OpenClaw Hackathon”. OpenClaw’s creator, Peter Steinberger, personally flew to London for this event.

Which projects are the most interesting?

On March 7, teams from various universities showcased prototypes they built over a week, covering a broad spectrum from agriculture to biosafety, from urban governance to DeFi protection. Here are six projects worth paying special attention to:

AgroMind: Satellite Data + AI Agent Making Agricultural Risk Hedging a Reality

AgroMind integrates satellite crop monitoring, weather data, and market signals to create a predictive and automated hedging system for agricultural supply chain risks. Its core scenario is an automated hedging workflow.

Information asymmetry in agriculture supply chains has always been a money issue. Commodity prices fluctuate wildly, often due to climate risks in certain regions that have been brewing for months, but the market only reacts after news reports. AgroMind aims to fill this gap. It combines satellite crop monitoring, weather data, and market signals so that when satellite images show early drought stress signs in a soybean region in Brazil—before any official reports—the system is already running. It checks the user’s inventory and current market volatility, drafts hedging plans, and if conditions are right, places orders directly on commodity exchanges. Rather than just an AI tool, it’s like an analyst sitting in front of satellite images, watching the market 24/7.

ClawBio: Bioinformatics with a Hugging Face

Bioinformatics has a long-standing problem: top analysis tools and knowledge are locked within a few universities and pharmaceutical companies, making it inaccessible to ordinary researchers. ClawBio aims to do for bioinformatics what Hugging Face did for AI models. It is an open repository of verified, reproducible biological analysis skills, which any Agent can call upon, including toxin screening and dangerous organism identification. A very interesting scenario: a user takes a photo of a drug package, and the Agent queries ClawBio’s skills against local genomic archives, returning a personalized medication dosage card in seconds. All data is processed locally, without uploading to any server. This “Local-First” approach is especially sensitive in healthcare and vital for privacy protection.

BioSentinel: From Pathogen Identification to Drug Candidate, End-to-End Automation

BioSentinel’s ambition is even greater. It starts with global public health data, continuously scraping information from WHO, CDC, CIDRAP, and other sources. When a new threat is identified, it automatically locates the pathogen’s target proteins, then uses computational biology tools like RFdiffusion and ProteinMPNN to design potential therapeutic molecules. Each candidate undergoes screening against toxin databases to ensure safety before proceeding. The entire process can be driven via a chat interface. Researchers don’t need to run commands manually; they just specify their needs, and the Agent orchestrates the tools. This significantly lowers the barriers in computational biology.

“London Nervous System”: From Smart City to “Thinking City”

This project starts with a simple idea: London generates massive sensor data daily—traffic, air quality, infrastructure status—but these data are fragmented, and no one truly knows the city’s real-time state.

The team used OpenClaw to connect real-time traffic flow, air quality sensors, and financial market data monitoring. If air quality drops suddenly in a district, the system doesn’t just log it; it proactively pushes low-pollution route suggestions to nearby schools and commuters. If a streetlight or sensor fails, the response is much faster than manual reporting. The team’s long-term goal is to open this framework to local governments, integrating with existing city systems rather than building a new one from scratch.

Highstreet AI: Building “Digital Employees” for Small Shops on London Streets

Most AI products are designed for tech companies, not small businesses like the seafood shop on Kingston Street. Highstreet AI aims to address this gap.

It targets small and medium enterprises that receive emails, WhatsApp messages, and phone orders daily but lack any IT system. Highstreet’s solution deploys a set of collaborative Agents: one understands the incoming request, another checks real-time inventory, a third drafts invoices and payment links, and finally, a dashboard button for the owner to approve.

The entire process requires only the owner’s final confirmation. Highstreet claims that this system can save a shop owner over 10 hours a week and requires no technical knowledge.

AlphaMind AI: Bringing Institutional-Grade Investment Logic to Retail Investors

There’s a deep divide between retail and institutional investors, not just due to capital but also analysis capabilities and response speed.

AlphaMind aims to bridge this gap. Users can compare their portfolios with public holdings of investors like Buffett, but the system does more than just show a comparison. It uses OpenClaw Agents to analyze concentration risks across multiple brokerages and exchanges, then automatically rebalances the portfolio. Its mission: not just tell you what happened, but explain why, and handle the adjustments for you.

Peter Steinberger, the “Lobster Godfather,” personally attended

In November, Austrian developer Peter Steinberger launched a project called “Clawdbot,” which can be messaged via Telegram or WhatsApp to help manage calendars, handle emails, run scripts, and even browse web pages. No one expected this project to sweep the global AI scene in just two months. OpenClaw exploded in popularity at the end of January 2026, and on February 14, Steinberger announced he joined OpenAI to develop the next generation of personal AI Agents. The OpenClaw project was handed over to an independent open-source foundation for continued development. Such a developer, who just became a key figure in the AI world, came to London because of this hackathon.

His trip to London nearly didn’t happen. The organizers revealed that just before departure, Peter suddenly discovered a visa issue, and “the whole team was panicking,” until two days before the event, when it was finally resolved. After securing the visa, he even changed his flight to ensure he could attend all sessions as planned. When he first entered the Imperial College classroom, he was just looking down at his phone, taking notes and preparing his speech, with no hint of an “AI celebrity” attitude.

During the hackathon,

At the subsequent Sequoia venture capital party, a developer who couldn’t get a ticket was standing outside in the rain. Peter noticed and didn’t hesitate—he approached and started chatting. When asked how the explosion of Agents might change the future of foundational large models, his honest answer was: “I don’t know. I’m better at building interesting things with the tools I have.” The speech was scheduled for only 30 minutes, but the atmosphere was so lively that the audience kept asking questions, and he stayed for over two hours. The organizers later said, “This meant a lot to us. Honestly, we owe him an apology.”

When leaving London, Peter left a final message: “You are not here to find meaning; you are here to create meaning.” Perhaps, this is the most important message for everyone seeking to make a difference in the AI era.

OpenClaw × Web3: Huge Potential, But Security Is the Biggest Hurdle

Steinberger himself isn’t very fond of the crypto scene, but the projects submitted for this hackathon starkly contrast his personal stance. On DoraHacks’ project page, several Web3-specific directions emerged:

• Agent identity and sovereignty are the most common themes. clawOS is built on the Nostr protocol, with each Agent holding an independent identity and wallet, not relying on any platform; Cortex.OS attempts to solve the black-box problem of AI in Web3, making each decision traceable on-chain.
• Direct money management is another direction. Trading Narwhal and Vibe4Trading are betting on Agents upgrading from auxiliary monitoring to executing trades directly, although OpenClaw’s architecture isn’t very friendly to private keys.
• Governance and public oversight also feature interesting projects: WatchDog uses six autonomous Agents to continuously scan UK government contracts for anomalies; CivicLift enables citizens to interact with local governments via Agents; GreenClaw is a multi-Agent city safety operation center.

However, security remains the biggest obstacle for OpenClaw entering Web3. Agents can access your files, APIs, and systems, but nothing monitors what they are actually doing. In scenarios involving real assets, users must be very cautious when adopting OpenClaw.