GIVE AN AI £100 AND TELL IT TO MAKE MONEY: LONDON HACKATHON PUTS AUTONOMOUS AGENTS TO THE TEST
£100, THREE WEEKS, NO EXCUSES: LONDON PUTS AI AGENTS INTO THE REAL ECONOMY
A London experiment will give competing teams three weeks to discover whether AI agents can turn £100 into real revenue, with every human intervention tracked and a £10,000 prize pool at stake. The result could provide a revealing test of the rapidly growing claims surrounding the agentic economy.
Artificial intelligence can write code, research markets, generate marketing campaigns and increasingly perform sequences of tasks with limited human intervention.
But can an AI agent actually go out into the messy real economy and make money?
A London hackathon intends to find out.
“Make Me Money”, the latest event in the Shipyard series, will challenge participants to build AI agents, fund them with £100 and attempt to generate a profit between November 2 and 21.
A live leaderboard will track both profits and human inputs, turning a familiar technology hackathon into something closer to an economic experiment.
The competition carries a £10,000 prize pool and will culminate in a public Demo Day and prizegiving ceremony at Encode Hub on November 21.
FROM BENCHMARKS TO BANK BALANCES
The experiment arrives as agentic AI becomes one of the technology industry’s biggest investment themes.
Unlike conventional generative AI systems that principally respond to prompts, agentic systems are being developed to plan and execute multiple tasks towards a goal, potentially with substantially less human involvement.
The commercial implications could be enormous.
Juniper Research forecast in April that global agentic commerce spending could reach $1.5 trillion by 2030, although it also identified trust as the biggest barrier to deployment. More recent research from the firm forecasts 120 billion agentic-commerce transactions annually by 2031.
Those are forecasts rather than evidence that today’s autonomous agents can independently build profitable commercial operations.
That distinction is precisely what Shipyard wants to explore.
“Every major forecaster, from the IMF, to McKinsey and even Goldman Sachs, has a number for what AI agents will be worth to the economy,” said Efe Kucuk, Head of the Shipyard series.
“Nobody has actually watched one earn a pound.”
‘THE ACTUAL, MESSY ECONOMY’
The design of Make Me Money is intended to distinguish it from trading competitions or laboratory benchmarks.
Agents will have to interact with real people and economic activity, while human interventions are logged.
“Make Me Money is the first public, live experiment to find out whether autonomous agents can go from zero to real revenue, with real strangers, real money, and every human intervention logged,” Kucuk said.
“Trading competitions test agents against markets but they don’t touch the real world. This tests them against the actual, messy economy: the one economists are still trying to model.”
The distinction could be significant.
Making money through an autonomous agent potentially requires far more than producing technically impressive outputs. An agent may need to identify demand, create something people want, find customers, communicate effectively and complete transactions while navigating practical and commercial obstacles.
And if humans repeatedly have to intervene to make those processes work, the experiment may reveal as much about the current limitations of autonomous agents as their capabilities.
HOW AUTONOMOUS IS ‘AUTONOMOUS’?
That makes the decision to track human inputs particularly important.
An AI system producing £1,000 of revenue with virtually no intervention represents a very different technological achievement from one producing the same amount after extensive human direction.
The leaderboard could therefore create an unusually tangible comparison between revenue and autonomy.
“Whatever the leaderboard says on November 21st, it’s the first hard data point in a debate that’s currently running on projections,” Kucuk said.
Shipyard says findings from the event will subsequently feed into a research paper examining how funded autonomous agents could contribute to the economy.
THE BIGGER ECONOMIC QUESTION
There is already serious research suggesting AI could raise economic productivity, although such projections should not be confused with evidence that autonomous agents can currently generate profits independently.
Penn Wharton Budget Model estimated in September 2025 that generative AI could raise US productivity and GDP by around 1.5% by 2035, with the strongest boost to annual productivity growth occurring in the early 2030s. The researchers explicitly cautioned that their projections were based on limited evidence about AI’s initial effects and could change substantially.
That differs from the $430 billion-a-year figure cited in promotional material for Make Me Money, which LBJ could not substantiate in Penn Wharton’s published analysis. The institution’s published forecast is therefore the figure used here.
SOHO BECOMES THE TESTING GROUND
Participants will receive credits from ElevenLabs, Tavily and Daytona, alongside workshops from teams behind the tools.
They can build their agents from a dedicated coworking space in Soho or participate remotely before the cohort gathers for Demo Day.
The £100 starting capital makes the experiment deliberately small.
The question it is asking, however, is anything but.
If AI agents eventually become capable of autonomously identifying commercial opportunities, finding customers, negotiating transactions and generating revenue, the implications extend far beyond hackathons. They touch entrepreneurship, employment, e-commerce, payments and even the meaning of running a business.
But Make Me Money could produce another valuable result: demonstrating exactly how much human involvement today’s supposedly autonomous agents still require.
For all the trillion-dollar forecasts surrounding agentic AI, London’s experiment reduces the debate to one refreshingly simple test:
