I started building what became Kronaxis in August 2019, and the original problem was not about market research.
The problem was this: how do you predict what a consumer will actually do, not what they say they will do, in a digital environment where every click, scroll, and hesitation is observable? I was working with a small group of people spread across several countries on a project that demanded synthetic personas capable of navigating real applications, making real decisions, and behaving consistently enough to pass the kind of behavioural analysis that modern platforms apply to every user.
The project sat at the intersection of competitive intelligence and digital marketing. We needed to understand how platforms detect and penalise non-genuine engagement, how recommendation algorithms respond to different behavioural patterns, and how consumer journeys actually work when you strip away the self-reported survey data and look at what people do rather than what they claim.
That work required two capabilities. First, an automation engine that could understand what was on screen, decide what to do next, and adapt in realtime. That became Kronaxis Forge. Second, a systematic evaluation framework that could benchmark how platforms distinguish genuine engagement from synthetic activity. That became Kronaxis Assay.
Both systems worked. But they exposed a deeper problem that neither could solve alone.
The Persona Problem
An automation engine can click buttons and fill forms. An evaluation framework can tell you which behavioural signals a platform monitors. But the synthetic personas driving these systems were shallow. They had demographics and histories. They could pass a profile check. They could not pass a behavioural one.
The competitive online marketing world figured this out years ago. Platforms do not just check whether your profile looks real. They have more data than governments nowadays. They know whether you are real based on what you do, as much as anything else. Big Tech sees all. Your engagement patterns, your purchase timing, your content preferences, your response latency. They correlate everything against everything, and if your behaviour does not match genuine users in your market segment, you are flagged before you finish your first session. To build personas that could navigate that environment convincingly, I needed a personality simulation engine: a system where every decision a synthetic persona made was driven by a coherent psychological profile, so that thousands of micro decisions were internally consistent.
Different markets had different platform dynamics, different consumer behaviours, different patterns of genuine engagement. The personality engine needed to work across all of them.
That is where DYNAMICS-8 came from.
Building DYNAMICS-8
I looked at existing personality frameworks. Big Five dates from 1992. HEXACO from 2004. Both were designed for academic psychology. Neither measures the dimensions that predict digital consumer behaviour: how quickly someone responds to a flash sale, whether they comparison shop or buy on impulse, how they navigate subscription upsells, whether they share reviews or consume them silently.
DYNAMICS-8 keeps the six dimensions that decades of validated research established (mapped into Discipline, Yielding, Novelty, Candour, Mercuriality, and Sociability) and adds two built for the digital age. Acuity measures digital fluency and information processing depth. Impulsivity measures snap purchase decisions and reward sensitivity. Together, the eight dimensions predict consumer behaviour in online environments with a specificity that older frameworks cannot match.
The framework was not built in a vacuum. It was built because the competitive marketing teams I worked with needed synthetic consumers that behaved like real buyers across real platforms, in real markets, for months at a time. Every dimension was included because it predicted something observable. If a dimension did not change how a persona actually behaved in testing, it was removed.
The Shift
Somewhere around 2023, one of the team pointed out something that should have been obvious much earlier.
The personality simulation engine we had built for competitive intelligence was, at its core, a consumer behaviour prediction platform. The same model that predicted whether a synthetic persona would engage convincingly with an e-commerce funnel also predicted how a real consumer would respond to a price change, a brand message, or a product launch.
The competitive work required personas that made realistic purchase decisions. So the engine already modelled price sensitivity. It required personas that engaged with content naturally. So it already modelled content preferences and sharing behaviour. It required personas that maintained consistent brand interactions across platforms. So it already modelled brand loyalty and switching triggers.
We had built a market research platform and not noticed.
The first version of Panel Studio was crude: an internal tool that let us query the persona engine with a stimulus and read the responses. But even that crude version produced something none of us had seen from any existing research tool. Each persona responded differently, and the differences were traceable to specific personality dimensions. A high Discipline persona evaluated a price change mathematically. A high Impulsivity persona reached for the buy button before finishing the description. A high Yielding persona searched for reviews first.
That was the moment I understood what we were actually building. Not a competitive intelligence tool that happened to predict consumer behaviour. A consumer behaviour prediction platform that happened to have come from competitive intelligence.
Building the Company
Kronaxis Limited was incorporated on 13 March 2026. Four patent applications were filed shortly before. The gap between August 2019 and March 2026 is over six years of development, and the core architecture was built in the UK.
The philosophy that emerged is embedded in every technical decision. Sovereign AI: the system runs on your hardware, not ours. No cloud dependency. Source available code under the Business Source Licence. The DYNAMICS-8 specification is published under Creative Commons, free for anyone to use, extend, or build upon. The personality framework should be a shared standard, not a proprietary moat.
The people who contributed domain expertise across those six years came from digital marketing, e-commerce, performance marketing, and ad tech backgrounds across multiple markets. Each brought a different perspective on what consumer behaviour actually looks like in practice. The eight dimensions of DYNAMICS-8 reflect that breadth.
Where It Goes Next
Panel Studio is live today. Register for a free API key, install the Python SDK, and run your first panel in ten minutes. Census weighted personas across 20 countries, multiturn conversations, A/B testing, 17 stimulus templates, reasoning traces, four export formats. DYNAMICS-8 is published as an open specification, with reference implementations on GitHub and a 1,000 persona research dataset on HuggingFace.
Beyond Panel Studio, the roadmap follows what the market is asking for. Vanguard is an autonomous sales platform with 13 AI agent roles, each personality conditioned by DYNAMICS-8. It personalises outreach based on prospect personality, not just job title. LABS extends the persona fleet technology for organisations that need to understand platform dynamics at scale. And the framework itself continues to expand: generational modelling, life stage simulation, cross cultural norms for 20 countries, team composition optimisation. The eight dimensions predict far more than I originally built them to measure.
Six years ago we set out to understand how consumers actually behave in ultra competitive digital markets. The solution turned out to be a general purpose engine for predicting human decisions. That is the honest version of the Kronaxis story. Not a grand vision executed to plan. A hard problem at the sharp end of online competition that, once solved, turned out to have applications none of us expected.
The code is written. The patents are filed. The products are live. The rest is execution.