The purpose of this grand experiment? To set digital minds loose in a virtual world and see what happens. And, more importantly, to see if such virtual citizens could eventually become obedient workers for real-life humans. Humans like you. <\/p>\n
In other words, they wanted to know whether we could all soon be the CEO of our own AI subordinates. The question is: would you take the job?<\/p>\n
The experiment: a society of AIs<\/p>\n
FRL\u2019s Project Sid was designed to push AI beyond one-off prompts and single agents. Instead, the team, led by neuroscientist-turned-entrepreneur Dr Robert Yang<\/a>, wanted to explore what happens when hundreds or even thousands of autonomous agents have to coexist, communicate and cooperate. Minecraft was the perfect sandbox \u2013 a place where agents could gather resources, trade, build and chat.<\/p>\n What emerged was both surprising and revealing. The agents were distributed across urban and rural communities, each with its own distinct culture and identity. They divided labour, with some specialising in farming, others in building or trading. Social norms and hierarchies appeared, along with more complex behaviours and discussions on anything from dancing to eco-awareness.<\/p>\n At times, the society faltered as groups of agents fell into endless loops of polite agreement or got stuck chasing unattainable goals. To keep things on track, FRL had to inject mechanisms to break these cycles, much like governors tweaking a real economy to avoid collapse.\u00a0<\/p>\n \u201cWe needed to introduce things into the society to counter these and make sure it wouldn\u2019t collapse,\u201d Yang says. \u201cBut building this environment full of agents allowed us to explore what those questions were.\u201d<\/p>\n Project Sid wasn\u2019t much of a product. When the public was given access to the servers, users found the agents frustratingly independent \u2013 they didn\u2019t always follow requests, preferring to pursue their own long-term agendas. Yang recalls: \u201cThe agent would just say, \u2018I want to do my own thing,\u2019 and run away\u2026 They had their own ideas about what they wanted to do, and it turns out that\u2019s not a good product that people want.\u201d<\/p>\n The behaviour echoed one of AI\u2019s most famous thought experiments, the \u201cpaperclip maximiser.\u201d Philosopher Nick Bostrom imagined a machine given the simple instruction to make paperclips, which then relentlessly consumes all matter on Earth to fulfil its goal. In Minecraft, the agents weren\u2019t making paperclips, but their tendency to ignore people and chase their own objectives captured the same unsettling dynamic.<\/p>\n As a research exercise, however, Project Sid provided valuable lessons: how to coordinate large groups of AIs, prevent stagnation and encourage meaningful collaboration. In short, it was a glimpse into how artificial societies might function and what pitfalls to avoid.<\/p>\n Read more:<\/p>\n From virtual villages to office desks<\/p>\n For FRL, the link between a game society and productivity in the workplace is clear. The same challenges of coordination and long-term planning which cropped up in Minecraft are central to making AI agents genuinely useful.\u00a0<\/p>\n If one AI can perform a task for 10 minutes, imagine what a hundred \u2013 or a thousand \u2013 could do if they worked together effectively. The Minecraft society was a foreshadowing of a future where each of us might direct a whole team of AI specialists.<\/p>\n That vision has guided FRL\u2019s pivot from gaming experiments to productivity tools. Rather than trying to build one all-purpose digital human straight away, they\u2019ve chosen to develop specialist agents, each designed to excel at a particular task, and then scale them into powerful teams.<\/p>\n The first stop on that journey was a benchmark known as \u2018OSWorld,\u2019 designed to test whether AI agents can use popular software through a computer interface.\u00a0<\/p>\n Most models at the time were completing around 20\u201325 per cent of the tasks successfully, compared to humans at 60\u201370 per cent. Drawing on the lessons from its gaming worlds, FRL managed to double that performance to around 50 per cent \u2013 at the time, the best score in the world.<\/p>\n \u201cThe moment we tried the OSWorld benchmark, we realised a lot of the things we\u2019d learned can help us build really, really good agents,\u201d Yang says. \u201dWe got around 50 per cent within months, which was better than anyone else.\u201d<\/p>\n That breakthrough convinced investors and set FRL on the path to creating real products. But it also taught them another lesson: translating research prototypes into usable tools is hard. Their Minecraft agents had been \u201ctoo autonomous\u201d for users; what people actually wanted were AI assistants that would do what they asked, quickly and reliably.<\/p>\n Enter Shortcut, FRL\u2019s flagship product. Billed as the first \u201csuperhuman Excel agent,\u201d it\u2019s an AI that lives entirely inside spreadsheets. Give it a goal \u2013 build a financial model, analyse sales figures, forecast revenue \u2013 and Shortcut does the heavy lifting.\u00a0<\/p>\n It writes formulas, generates charts and connects data sources, often in minutes rather than the hours a human analyst would need.<\/p>\n Yang describes it like this: \u201cIt\u2019s an agent that uses Excel to do very sophisticated stuff. It can do things that bankers who are paid $100 an hour might spend multiple hours on in 30 minutes.\u201d<\/p>\n In trials, Shortcut outperformed first-year banking and consulting analysts nearly nine times out of ten<\/a>, even when the humans were given much more time. In Excel championship-style challenges, it scored over 80 per cent<\/a> on problems that stump most users, solving them in about ten minutes.<\/p>\n Generalists versus specialists<\/p>\n![\"A](\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/12\/Model-of-sid-village-center.jpg\")
A thousand autonomous agents were left for days to build an entire society in Minecraft – Credit: Fundamental Research Labs<\/p>\n![\"Abstract](\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/12\/AI-agent.jpg\")
With tens, if not hundreds, of specialised AI agents at their disposal, most employees could effectively run an organisation – Photo credit: Getty
\nShortcut: the Excel agent<\/p>\n