We were a generalist software firm. We focused.

The pivot

Why sovereign, open-source, and document-heavy, and why now.

Three things became true at once. Open-weight models (Llama, Qwen, Mistral, DeepSeek) caught up with frontier systems on the work enterprises do with AI: reading, extracting, classifying, and reasoning over private documents. Data sovereignty became a board-level concern for security and IT leaders, and regulation like the EU AI Act turned it into a compliance obligation. And the organizations under the most pressure, financial services and lending first, then healthcare, legal, and the public sector, are the ones that cannot send their documents to another company's cloud.

Frontier models lead on code generation, but that is not this problem. Most enterprise AI is document and data work. So we narrowed from a firm that could build anything to one that does a single thing: sovereign, open-source AI for document-heavy workloads, on infrastructure the customer controls. We start with financial services, lending, credit, and compliance-heavy firms, because that is where the rules on data leaving the customer's environment are tightest. The rest of our regulated work follows the same pattern.

We are direct about this because the people we work with expect it. A firm that will tell you "self-hosting is the wrong answer here" is more useful than a generalist that agrees to everything, and more useful than a vendor whose answer is its own product.

What we believe

That you should not have to choose between modern AI and keeping custody of your data. That open weights point to a more competitive market, one where established companies build their own advantage with AI instead of renting it from the few firms that train the models. That an architecture is only worth claiming if it's open. And that the honest answer, including the one that costs us the engagement, is the one that builds a practice worth having.