In March 2026, the UK AI Agent Hackathon organized by Imperial College London took place. Peter Steinberger, the father of OpenClaw, flew in personally to attend. Over 1,200 developers used OpenClaw as the core framework to build six major projects, including agricultural risk hedging, bioinformatics platforms, and smart city neural systems. Peter was scheduled for a 30-minute speech but ended up talking for over two hours.
(Background: As the AI wave surges, crypto builders may have an advantage? The three major financial giants’ actions speak volumes.)
(Additional context: The World Gold Council partnered with BCG to release a tokenized gold sharing framework, directly challenging Tether and Paxos.)
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 technology, attracting over 1,200 registered participants. On Demo Day, a record 5,000 viewers watched live online, briefly topping the global trending charts on X platform.
Many participants regarded it as the “world’s first University OpenClaw Hackathon.” Peter Steinberger, the creator of OpenClaw, personally flew to London for the event.
On March 7, teams from various universities showcased prototypes built in just one week, covering a broad spectrum from agriculture and biosafety to urban governance and DeFi protection. Here are six noteworthy projects:
AgroMind integrates satellite crop monitoring, weather data, and market signals to create a predictive and automatic hedging system for agricultural supply chain risks. Its core scenario is an automated hedging workflow.
Information asymmetry in agricultural supply chains has always been a money issue. Price volatility in commodities often stems from climate risks in certain regions months in advance, but markets only react after news breaks. AgroMind aims to fill this gap. It combines satellite imagery, weather data, and market signals so that when early drought signs appear in a soybean region in Brazil, the system detects it before official reports. It checks user inventories and current market volatility, drafts hedging plans, and if conditions are right, places orders directly on trading platforms. Rather than just an AI tool, it’s like a vigilant analyst watching satellite images 24/7—never sleeping.
Bioinformatics has long faced a problem: top analysis tools and knowledge are locked within a few universities and pharmaceutical companies, inaccessible to most researchers.
ClawBio aims to replicate the success of Hugging Face in AI models within bioinformatics. It’s an open repository of verified, reproducible biological analysis skills that any Agent can call upon, including toxin screening and dangerous organism identification.
A fascinating scenario: a user takes a photo of a drug package, the Agent calls ClawBio’s skills to query local genomic files, and within seconds, returns a personalized medication dosage card. All data is processed locally, without uploading to any server. This “Local-First” approach is especially sensitive in healthcare, where privacy protection is paramount.
BioSentinel has even bigger ambitions. 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 targets the pathogen’s proteins and uses computational biology tools like RFdiffusion and ProteinMPNN to design potential therapeutic molecules.
Each candidate molecule undergoes toxin database screening before proceeding, ensuring no dangerous compounds are accidentally created. The entire process can be driven via chat interface. Researchers don’t need to run commands manually; they just specify their needs, and the Agent schedules the tools accordingly. This significantly lowers barriers in computational biology.
This project’s humble origin lies in London’s massive daily sensor data: traffic, air quality, infrastructure status—yet these data streams are fragmented, and no one truly knows the city’s real-time state.
The team used OpenClaw to connect real-time traffic, air quality sensors, and financial market data. When air quality drops suddenly in a district, the system doesn’t just log it; it proactively suggests low-pollution routes to nearby schools and commuters. If a streetlight or sensor fails, the response is faster than manual reporting. Their long-term goal is to open this framework to local governments, integrating with existing city systems rather than building from scratch.
Most AI products are designed for tech companies, not small eateries on Kingston Street selling seafood. Highstreet AI aims to address this gap.
It targets small and medium-sized businesses that receive emails, WhatsApp messages, and phone orders daily but lack any IT system.
Highstreet’s solution deploys a set of collaborative Agents: one to understand incoming requests, another to check real-time inventory, a third to draft invoices and payment links, and finally, a dashboard with an “Approve” button for the owner.
The entire process requires only the owner’s final confirmation. Highstreet claims this system can save a shop owner over 10 hours weekly without requiring any technical knowledge.
There’s a deep divide between retail investors and institutions—not just due to capital differences, but also analysis capabilities and response speed.
AlphaMind aims to bridge this gap. Users can compare their portfolios with public holdings of Warren Buffett and others, but it’s more than just a comparison chart. It uses OpenClaw’s Agents to analyze asset concentration risks across multiple brokerages and trading platforms, then automatically rebalances.
Its mission: past tools tell you what happened; AlphaMind explains why, and handles the adjustments for you.
In November, Austrian developer Peter Steinberger launched a project called Clawdbot. You can send messages via Telegram or WhatsApp, and it will manage your calendar, handle emails, execute scripts, and even browse websites.
No one expected this project to sweep the global AI scene within just two months. OpenClaw exploded in popularity at the end of January 2026. On February 14, Steinberger announced he joined OpenAI to develop the next-generation personal AI Agents, and the OpenClaw project was handed over to an independent open-source foundation for continued development. This developer, who just became a prominent figure in AI, arrived in London because of this hackathon.
His trip almost didn’t happen. The organizers revealed that Peter discovered a visa issue just before departure, causing panic among the team. It was only two days before the event that the problem was finally resolved. After securing the visa, he even changed his flight to ensure he could attend all sessions as planned. When he first stepped into the Imperial College lecture hall, he was just focused on his phone, taking notes and preparing his speech—completely unpretentious, with no “AI celebrity” airs.
At the hackathon, Peter:
During the subsequent Sequoia venture capital party, a developer who missed the ticket stood outside in the rain. Peter noticed him, didn’t hesitate, and approached to chat. When asked how the explosion of Agents might change the future of large foundational models, his honest reply was: “I don’t know. I’m better at building interesting things with the tools I have.”
Originally scheduled for just 30 minutes, his speech lasted over two hours due to the lively atmosphere and continuous audience questions. The organizers later said, “This meant a lot to us. Honestly, we owe him an apology.”
When leaving London, Peter left a parting thought: “You’re not seeking meaning; you’re creating it.” Perhaps that’s the most important message for anyone wanting to make a mark in the AI era.
Steinberger personally isn’t very fond of the crypto scene, but the projects submitted to this hackathon starkly contrast his personal stance. On DoraHacks, several directions for concrete Web3 implementations emerged:
- The identity and sovereignty of Agents are recurring themes. clawOS is built on the Nostr protocol, with each Agent holding independent identities and wallets, 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 focus. Trading Narwhal and Vibe4Trading bet on Agents upgrading from assisting with market analysis to executing trades directly, despite OpenClaw’s architecture not being very friendly to private keys.
- Governance and public oversight projects also appeared: 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 hurdle for OpenClaw entering Web3. Agents can access your files, APIs, and systems, but nothing monitors what they’re actually doing. In scenarios involving real assets, users must be cautious when adopting OpenClaw.
