Today, we’re opening Matilda to a broader group of beta users. Matilda is available at matilda.maincode.com and on the Apple App Store.
Matilda is Maincode’s Australian AI system: built from metal to model, designed for thoughtful use, and held to a higher standard for how AI should behave.
We’re building Matilda because Australia should have more than access to powerful AI. We should have AI systems that understand our context, reflect our standards, run on infrastructure we can trust, and give people and organisations more control over how they use this technology.
That matters more now than it did even a few years ago. AI is becoming part of the foundation for how people work, learn, communicate, create, and make decisions. At the same time, the geopolitical environment around AI is shifting quickly. Compute, models, data, supply chains, regulation, and national capability are all becoming questions of strategic importance.
In that environment, relying entirely on systems designed, deployed, and governed elsewhere is a real limitation. Australian users and organisations need AI that is not only capable, but locally grounded: shaped by Australian expectations around safety, privacy, clarity, accountability, and voice.
Over the past few months, we’ve been working closely with early users through the Matilda Insiders Club. Their feedback has helped shape the product in practical ways: how Matilda responds when it is unsure, how it handles everyday writing and research tasks, how it works with files, how it preserves context, and where the experience still needs to become faster, clearer, and more useful.
The open beta is the next step. It gives more people the chance to try Matilda, tell us where it works, and help us improve where it does not.
Why Australian voice matters
Every assistant has a voice. It has a sense of what counts as helpful, polite, confident, cautious, funny, formal, direct, or appropriate. Those choices show up in small moments: how a system gives advice, how it handles uncertainty, how it pushes back, how much it flatters, how it interprets tone, and how it responds when the user is asking for judgement rather than information.
For Australian users, these details matter.
A system can be fluent in English and still feel culturally off. It can be technically correct and still sound too promotional, too deferential, too certain, too generic, or removed from the way people here actually communicate. The answer may be acceptable, but the interaction does not quite fit.
Matilda is being designed with Australian voice as part of the product, not a cosmetic layer added at the end. This Australian voice is not a performative caricature, with forced slang or a blanket casual register. It means designing a behavioural system that is practical, clear, warm, direct, and contextually appropriate.
The goal is not to make Matilda sound Australian in every sentence. It is to make the interaction feel right for the context.
What we mean by Australian-made AI
When we say Matilda is Australian-made, we mean something specific.
Our claim is not that every underlying model component must be invented from scratch in Australia. It is that Matilda is an end-to-end Australian AI system in the ways that matter for trust, control, and use: deployed on Australian infrastructure, adapted for Australian contexts, governed by Australian safety expectations, evaluated against Australian product standards, and delivered through an experience designed here.
Where open-weight models are appropriate, we treat them as inputs into the system, not as the system itself. The differentiation is in how these models are selected, adapted, deployed, safeguarded, evaluated, and made useful for Australian users and organisations.
That distinction matters.
AI systems are not just model weights. They are infrastructure, data handling, safety behaviour, product design, evaluation, governance, and operational control. They are also shaped by the decisions a team makes about what the system should do, what it should refuse, how it should handle uncertainty, and how much confidence it should project when the answer is not clear.
Those choices are not incidental. They are the product.
Why this matters now
People do not experience AI as a model card or a benchmark. They experience it as a conversation, a workflow, a draft, a decision support tool, or a moment of trust.
That is why we have been focused on the full system around Matilda, rather than the raw capability of any single model.
For Australian users and organisations, trust depends on more than whether an AI system can produce a fluent answer. It depends on where the system runs, how information is handled, what standards it is evaluated against, how it responds to ambiguity, and whether the product is designed for the context in which people actually work.
It also depends on control.
As AI becomes more embedded in everyday operations, organisations will need to understand more about the systems they rely on: where inference happens, how data is handled, what behaviours are evaluated, which safeguards are in place, and who is accountable when the system changes.
This is especially important for governments, regulated industries, enterprises, educators, researchers, and any organisation working with sensitive information or public trust. Foundational AI capability is becoming part of national infrastructure. It should not be treated as a generic software subscription with no local context and limited visibility.
We want Matilda to be capable, but also calibrated. Useful, but not over-eager. Clear about what it knows and what it does not. Designed to support people’s judgement, not replace it by default.
Built from metal to model
Maincode’s approach is grounded in control over the AI stack.
Matilda runs on Australian infrastructure, with inference on Australian shores. That gives us a different foundation for building trust, reliability, and governance into the product from the start.
Running inference in Australia is not only a symbolic choice. It affects what we can offer users and organisations in practice: clearer data boundaries, lower dependency on offshore systems, more direct operational control, and a stronger foundation for meeting Australian expectations around privacy, security, and accountability.
That infrastructure matters because AI is not static. Models change. Usage patterns shift. Safety expectations evolve. The more control we have over the system, the better we can evaluate it, improve it, and respond when we need to.
But infrastructure alone is not enough.
The real work is in how the system comes together: the models we use, the way they are adapted, the safety behaviours we design, the evaluations we run, the user experience we build, and the feedback loops we create with the people using Matilda in practice.
This is what we mean by an end-to-end system.
It is not a claim that every piece of the AI supply chain begins from a blank page. It is a claim that the product is built, operated, evaluated, and improved as an Australian-controlled system.
What beta users can expect
Matilda is still in beta, and that means two things.
First, it is already useful for a range of everyday work: drafting, rewriting, summarising, exploring ideas, working with files, preparing communications, and helping turn rough thoughts into clearer outputs.
Second, it is still learning from real use. Some parts will feel polished. Others will need improvement. There will be moments where Matilda is too cautious, not contextual enough, or not yet as smooth as we want it to be.
That is exactly why we are opening the beta.
We are especially interested in feedback on the moments that matter for trust: when Matilda is helpful, when it gets something wrong, when it should ask before answering, when it should be more direct, and when its behaviour feels different from what you expect from an AI assistant.
The beta gives us a way to learn what reliability looks like in practice, across real users, real tasks, and real expectations.
Matilda is available at matilda.maincode.com and on the Apple App Store.
An Australian AI system people can help shape
The open beta is an invitation to use Matilda and help shape what it becomes.
We are building for people who want AI that is powerful, but also locally grounded, carefully governed, and designed with taste, restraint, and transparency. We believe Australia should have AI systems that reflect those priorities, not just access to products built for someone else’s market, norms, and assumptions.
Matilda is our contribution to that future.
It is early. It will keep changing. And the feedback we receive during this beta will directly shape the product.
Welcome to Matilda.