Practical AI, available to everyone — not just people who can afford another subscription.

But there is a problem. Much of today's AI depends on expensive cloud infrastructure, recurring subscriptions and external AI providers. That creates a barrier.

We believe AI should be available when people need it, affordable enough for everyone, and capable of doing useful things without requiring a permanent subscription to a powerful cloud AI service.

We are building a different approach:

Lightweight AI agents that can run on affordable, portable hardware and perform practical tasks locally — with external AI used only when genuinely necessary.

Because much of the intelligence can run locally, the device doesn't need to send every request to a large cloud AI model.

This isn't about replacing ChatGPT. It's about making practical intelligence available beyond the people who already have access to it.

Consider someone trying to make an appointment at their GP surgery. Or a parent helping their child learn. Or an older person who needs reminders throughout the day. Or somebody living with dementia who benefits from simple prompts, reminders and familiar interactions. Or a small business that needs a digital worker to carry out repetitive administrative tasks.

These are not futuristic problems. They are everyday problems — and they don't necessarily require a trillion-parameter AI model to solve them.

Our approach is deliberately lightweight. Instead of sending every task to an expensive external AI service, our system first asks a simple question. If the answer is yes, it handles the task itself. If the task needs more intelligence, it can use a small local model. And if something genuinely requires a powerful external AI model, it can call one selectively.

This means the majority of routine tasks can potentially be performed without continuously paying for external AI inference.

A major part of our proposition is that we don't believe every AI task requires a massive model. A business doesn't need a supercomputer to recognise a button, read a simple document, identify a task, check a calendar, trigger a reminder, follow a workflow, classify information, monitor a system, calculate a route, or perform a predefined action.

Many of these tasks can be handled through conventional software, rules, OCR, speech recognition, computer vision and small local models. The result is what we call Micro-Agent Intelligence — small, focused digital workers designed to do specific jobs extremely efficiently.

We are exploring a new generation of portable AI hardware. The exact hardware specification is still being developed — that is deliberate. We want the software and agent architecture to establish what the hardware actually needs, rather than designing an expensive device first and trying to find a use for it afterwards.

Today's AI can look like this — every interaction creating infrastructure and inference cost:

We don't want a person's ability to use useful AI to depend on whether they can afford another monthly subscription. That is particularly important for families, children, older people, vulnerable people, carers, and people who simply cannot justify another monthly bill.

We believe there should be a way to build sustainable technology while keeping essential functionality accessible. The commercial opportunity comes from the wider ecosystem: hardware + specialist services + business agents + enterprise deployments + partnerships + optional advanced functionality.

Our model is closer to: build the intelligence once, put it into affordable hardware, make useful functionality available, and build commercial applications around the platform. This gives us the opportunity to generate revenue from businesses and specialised applications while keeping core public-facing functionality free.

The same underlying agent technology can power both. A lightweight agent that understands a user's request and performs an action can be adapted to a family, a school, a care environment, a healthcare navigation service, a retailer, a logistics company, an office, or a customer service operation. The difference is the workflow — the underlying technology remains similar.

Our focus is efficiency. We ask a different question:

What is the smallest amount of intelligence required to successfully complete this task?

That philosophy can reduce infrastructure requirements, AI inference costs, dependency on external providers, latency and unnecessary cloud processing. It can also make AI more practical on portable devices.

Local processing creates another important opportunity: where appropriate, information can remain on the user's device rather than automatically being sent to an external AI provider. That matters particularly when dealing with personal information, family information, education, sensitive conversations and everyday routines.

The architecture can decide what needs to stay local and what, if anything, needs to be sent externally.

We want to build a platform for practical intelligence — a platform where thousands of small agents can perform thousands of useful tasks. Some for individuals. Some for families. Some for education. Some for care. Some for businesses. Some running entirely locally. Some using external AI when necessary.

The user shouldn't have to understand any of this. They should simply be able to say:

“Help me.” And the system should know what to do next.

• Core agent technology
• Local AI model development and optimisation
• Portable hardware research and prototyping
• Software development
• Speech and computer-vision capabilities
• Security and privacy architecture
• User testing
• Healthcare and safeguarding requirements
• Education applications
• Business-agent pilots
• Partnerships
• Intellectual property
• Initial market launch

The immediate objective isn't to build a huge AI company. It's to prove that useful AI can be delivered dramatically more efficiently.

This is not simply an investment in another AI application. The biggest AI companies are building enormous models and enormous infrastructure. We're asking a different question:

What happens if intelligence becomes small enough to live with us?

If we can make that work, the potential applications are enormous.