The first group thesis

What a group thesis actually is

Most organisations have a product thesis. A product thesis says: here is the specific problem, here is who has it, here is what we are building to solve it. It is a narrow statement, deliberately. The narrowness is what makes execution possible. You hold one thing in mind, you move toward it, you find out whether you were right.

A group thesis is something else. It does not describe one product. It describes the territory that an institution is trying to occupy: the underlying belief about where value can be created, repeated, across multiple attempts, across multiple contexts. The group thesis is not a strategy document. It is closer to a founding conviction.

Mustard Seed Group's group thesis, in its earliest and most honest form, is this: human capability is the operating variable. Not attention, not engagement, not demand, not traffic. Capability. Our working belief is that the most durable and significant work we can do is build systems that increase what people and organisations are genuinely able to do: think through harder problems, execute at higher standards, maintain performance under pressure, make real decisions with real information.

Everything else flows from that.

Why a group and not a product company

The question that came up repeatedly in the early months was the simplest one: why not just build one product? If you have a conviction, find the one expression of it that scales fastest and commit everything to it.

It is a reasonable question. The conventional answer is focus. The unconventional answer is range.

Some theses are genuinely best expressed through a single focused product. If your conviction is that payments infrastructure is broken, you build the payments infrastructure. One product, one problem, relentless depth. The discipline of the single focus is itself the strategic advantage.

But the capability thesis does not behave that way. Capability is not one broken system. It is a condition that applies across domains. A professional executing commercial work has a capability problem. A person trying to manage their physical performance and accountability has a capability problem. A research team trying to synthesise knowledge and act on it has a capability problem. These are different surfaces of the same underlying challenge, and they require structurally different products to address them. A single product cannot carry all of that weight without becoming incoherent.

What you risk, if you try to force the capability thesis through a single product, is building something broad enough to claim the thesis but not deep enough to actually fulfil it. A platform that says it helps people perform better without ever doing the hard work of understanding what performing better actually requires in a specific domain. A general-purpose tool that flatters everyone and serves no one especially well.

So the group structure is not an organisational compromise. It is the right container for this kind of thesis. The products are distinct. The research is shared. The commercial logic accumulates across them rather than requiring each one to justify itself in isolation.

Three structures taking shape

By October 2018, three internal structures were beginning to crystallise, not as formal divisions yet, but as distinct modes of operation that the work was settling into.

The first is research. Before you can build reliably, you need to understand what is actually true about the problem. Not what is assumed, not what has been written in a think piece, but what is operationally true when you are close to the work. Benediction Lab is where this lives: agents, memory systems, how intelligence can be embedded into workflows without reducing the person operating them to a passive observer. Research is the slowest structure and the most important one. It is what separates a portfolio of unrelated bets from a coordinated body of work.

The discipline Benediction Lab is trying to develop is not simply technical. It is a discipline of asking harder questions about what capable systems actually look like in use. The gap between a system that works in a research environment and one that functions reliably inside a real workflow is significant. Most of the interesting problems live in that gap.

The second is product. Product is where the research findings get turned into something a person can use. This is Orbit's territory. The challenge with commercial execution, the lead-to-launched-product workflow, is not that there are no tools. There are many tools. The challenge is that no existing tool holds the full surface together. Sales activity lives in one place, project logic in another, product decisions somewhere else, client relationships somewhere else again. Orbit's thesis is that these do not need to be separate. Commercial work has an internal logic, and a product built around that logic will always outperform a collection of point solutions that have been integrated by hand.

The third is commercial validation. It is one thing to have research. It is one thing to have a product. It is another thing entirely to test the underlying patterns in live environments with real clients. TUXX exists to do that. Custom systems, real briefs, real delivery. The learnings from TUXX flow back into the product work and back into the research. Pattern Up is a sub-product under TUXX, an early attempt to formalise the patterns that emerge from that live work into something repeatable and teachable.

These three structures, research, product, and commercial validation, are not separate departments. They are a loop. And the loop only works if the centre is clear.

The common thread

The mistake would be to describe these structures as a diversified portfolio in the venture capital sense: a set of independent bets distributed across different sectors to manage risk. That is not what this is.

Every part of the portfolio is oriented toward the same underlying question: what would it take for this person, in this context, to operate at a genuinely higher level? For a commercial team, that means better intelligence and a more coherent operating surface. For someone managing their physical performance, that means clearer accountability structures, smarter coaching, and feedback that is honest rather than flattering, which is what CheekyGains and Naira are built around. For an organisation trying to deploy AI without losing sight of what the human is actually doing, that means the kind of thinking Benediction Lab is attempting to formalise.

The common thread is not AI. AI is an enabling technology, and a significant one, but it is not the thesis. The thesis is capability. AI happens to be the most powerful tool available right now for extending capability, so it is present throughout the work. But the work is not in service of AI. AI is in service of the work.

This distinction matters more than it might seem. Companies and institutions that organise themselves around AI as the thesis tend to drift with the technology, chasing what is new, what is impressive, what generates attention in the moment. Companies organised around a substantive thesis about human performance use the technology more deliberately. They are slower to adopt things that do not serve the thesis and faster to adopt things that do.

What a coordinated portfolio requires

The appeal of a group structure can also become its failure mode. If the thesis is not genuinely shared across the portfolio, the group becomes a holding company for unrelated projects, each with its own logic and no meaningful connection between them. The centre holds the narrative together but adds nothing structural to the individual components. That kind of group creates overhead without creating advantage.

A coordinated portfolio requires a different relationship between its parts. Research findings need to travel. Patterns discovered in one product need to inform another. Commercial work needs to surface real-world friction that the product team cannot see from the inside. The intelligence layer built to serve one product should, over time, be capable of serving others. The group creates compounding returns only if there is genuine exchange across its components.

This is why Orion, the AI intelligence layer that powers Orbit commercially, matters beyond Orbit itself. The work of building a reasoning and memory system that can handle the complexity of commercial workflows is also the work of building a system capable of other complex tasks. The investment made in Orion is not made for Orbit alone. It is made for the group.

The same logic applies to the consumer side. The work on CheekyGains is not separate from the broader capability thesis: it is a test of what accountability and coaching systems can actually do when they are built with genuine seriousness rather than as feature additions on top of a social feed. What is learned there about how people respond to intelligent performance feedback has implications well beyond the consumer fitness context.

Coordination, in practice, is not a meeting structure or a set of shared OKRs. It is a shared vocabulary for what you are trying to do, and a culture of genuine curiosity about what the other parts of the organisation are discovering.

The discipline of a founding conviction

There is something that needs to be said honestly about the early months of articulating a group thesis: it is tempting to make it too broad. A conviction about human capability could, if you let it, expand to encompass almost anything. Add enough qualifiers and you have justified every possible product decision in advance. That is not a thesis. It is a permission structure.

The discipline is in the specificity of the claim. Not "we care about humans": every organisation says that. Not "we want to improve performance": that is not actionable. The thesis that is actually organising this work is narrower than it first appears: we are building systems, and those systems have to increase genuine operational capability, and that increase has to be measurable in terms of what a person or team is able to do that they could not do before. Not how they feel, not how engaged they are, not their satisfaction scores. What they are genuinely able to accomplish.

That specificity is what allows you to say no. To a product direction that sounds interesting but does not serve the thesis. To an opportunity that would generate revenue but pull the portfolio in a direction that weakens its internal coherence. To a partnership that would add a logo but not add anything structural to the work.

Saying no coherently requires a clear thesis. Building a group without a genuine centre is just building a collection.

The beginning of something longer

A group thesis is not a plan. It is a working conviction that you test over time, refine under pressure, and hold firmly enough to prevent the portfolio from drifting wherever the current trend points.

In October 2018, the thesis is not fully formed. The structures are visible but not fully operational. The research is real but early. The products are further along in conception than in execution. What exists at this point is the conviction: clear enough to organise around, not yet tested at any meaningful scale.

The conviction is this: systems that increase human capability are worth building, and a group structure is the right vehicle for building them well.

Everything else is the work of proving it.