PART II OF III - THE HUMAN STORY
Aymeric Zhuo and Osman Ramadan were doing serious AI research, the kind that gets you cited alongside OpenAI. Then they walked away from it. Here's why that decision matters more than the product and why their move away from a traditional OpenAI Career path matters even more.
In Part I, we introduced you to Codewords - the London AI Startup processing 500,000 automated tasks a month, backed by €7.6 million and some of the most credible names in European tech.
We told you what it does.
Now we tell you where it came from.
Because this product didn't emerge from a product roadmap, it emerged from a genuinely strange decision.
Before Codewords, there was Agemo.
And before you'd heard of either, Aymeric Zhuo and Osman Ramadan were doing the kind of AI work that gets you cited alongside OpenAI.
So, Agemo was an AI research laboratory focused on neurosymbolic reasoning - a field that attempts to combine the pattern-recognition of modern neural networks with the logical, rule-based thinking of classical AI systems. In December 2024, Agemo's work appeared on the ARC-AGI benchmark, one of the most respected measures of AI reasoning capability, in the same company as OpenAI.
That's not nothing.
That's the sort of recognition that funds careers, builds reputations, and keeps most researchers comfortably inside academia, the kind of trajectory that often leads to an OpenAI Career.
Zhuo and Ramadan chose to leave it behind anyway.
The reason, as best we can understand it, was a question that kept nagging at them: what's the point of building something brilliant if the people who need it most can't actually get near it?
Businesses were drowning in repetitive tasks - chasing leads, monitoring competitors, drafting content, managing inboxes, and the tools that could help them were either too simple to handle real complexity, or too technical to use without an IT team standing by.
The growing AI Automation space had solutions, but not ones most businesses could easily adopt.
"We wanted to enable non-technical people to create software simply. You don't need technical expertise anymore, because today coding is solved for us." - Aymeric Zhuo, Co-founder
So they pivoted. Not a gentle pivot - a full rebuild of identity, product, and purpose.
Agemo became Codewords.
The research laboratory became a no-code infrastructure platform. And the target audience shifted from peer-reviewed journals to marketing managers in Manchester.
The gap they saw
The insight at the heart of Codewords is deceptively simple.
On one side of the market, you have vibe-coding tools - platforms like Lovable that let non-technical users create prototypes and basic applications without writing a line of code. Fast, accessible, but limited. Brilliant for getting something off the ground. Not built for the messy, branching complexity of real business operations.
On the other side, you have serious workflow automation platforms - the kind that can handle genuine complexity but require a level of technical knowledge that rules out most of the people who actually need them.
The gap between those two things - accessible but shallow versus powerful but complicated - is where Codewords has planted itself as an AI Startup.
The ambition, as Ramadan describes it, is to make automation as easy as vibe-coding a website, but with the same complexity as a serious engineering tool.
The ambition also places Codewords inside the wider AI Automation conversation.
The question is whether the product actually delivers on that ambition.
The answer is more complicated than either side would like.
The next part is where it gets honest.
Why Investors backed founders who left an OpenAI career
The €7.6 million seed round is notable not just for its size but for who put money in. Alongside lead investor Visionaries and institutional participants Firstminute Capital and Sequel, the round included angel investors who are, by any measure, an unusually high-calibre group.
Andrey Khusid, CEO of Miro. Mati Staniszewski, CEO of ElevenLabs. Ilkka Paananen, CEO of Supercell. Robert Gentz, CEO of Zalando. François Chollet, co-founder of the ARC Prize - the very benchmark Agemo appeared on. And leaders from OpenAI, Mistral, and Zapier, a direct competitor.
That last point is worth sitting with. Zapier - the company Codewords is most directly competing with - has people inside this funding round.
That either means they see Codewords as a complement rather than a threat, or they've decided to back the competition before it gets away from them. Either interpretation is interesting. We explore what that means for the competitive landscape properly in the final part.
"The best operators don't wait to be asked. We built Cody on the same principle - an agent that learns about your business, sees what needs doing, and delivers outcomes." - Aymeric Zhuo, Co-founder
We believe the calibre of the angel roster, as we noted in Part I - tells you something important. These are people who have built category-defining companies. They don't write cheques for fun. They write them when they've seen something that reminds them of an earlier version of themselves.
For a fast-growing AI Startup, that kind of backing carries weight.
Now, we've told you who built it and why they walked away from something most researchers would kill to keep.
What we haven't told you yet is whether what they built actually holds up and where it quietly doesn't.
The best is saved for last.
