Our own products
We look for categories where everyone has quietly agreed that nobody checks. Those are the good ones. If the industry standard is to make it up, then actually doing the work is both the right call and very hard to copy.
Dublin, Ireland · European Union
Most software asks you to take its word for things. Ours doesn't. If our product tells you something, you can follow it back and see where it came from. That's slower to build. We think it's the only kind of company worth spending years on.
The standard
Everyone says they care about quality. Almost nobody tells you what they won't do. Here's ours. We've already left money on the table to keep them.
No made-up precision. No "world's only" anything. No number picked because it sounds good in a headline. When our software can't work something out, it says so. A blank is uncomfortable for about a second. A confident guess costs you the user forever, usually without them telling you why they left.
The quickest money in consumer software is to worry someone and then sell them the cure. We won't build that. Not in any category, not at any margin. It works brilliantly until the day it doesn't, and that day takes the whole business with it.
When someone writes down what they're actually going through, that text is encrypted, kept out of analytics, and never handed to a model in full. You can export it or delete it yourself, from inside the app, without emailing anyone. GDPR is where we start, not what we aim for.
Language models are good at explaining things warmly. They're terrible at being held to a number. So every figure and date in our output is calculated properly and then written straight over whatever the model said. After that we check the prose against the data. The model gets to phrase it. It doesn't get to decide it.
Our maths is checked on every build against published reference values. Not against what we expected. Against answers someone else worked out and published years ago. Break one and the build stops. Nothing ships.
What we do
We're not an agency that'll build whatever you ask for. We're also not a one-app company. We build our own things where we think a category has got lazy, and we take on outside work when the problem is properly hard.
We look for categories where everyone has quietly agreed that nobody checks. Those are the good ones. If the industry standard is to make it up, then actually doing the work is both the right call and very hard to copy.
Work for companies who need something built properly. AI that has to be accountable. Maths that has to be right. Apps that have to survive contact with real users. We take on very few of these and we say no a lot.
Systems where the model's answer is held in place by maths it can't argue with.
Scientific and mathematical software, checked against outside reference data on every build.
Mobile and web software that keeps people because it's useful, not because it's sticky.
Products
Astrology is a category built almost entirely on making things up. Antara Nexus works it out instead. Planetary positions come from NASA JPL data, classical Vedic maths runs on top, and every sentence you read comes from one of those results.
We picked it on purpose. If you can be rigorous in a field where absolutely nobody is asking you to be, you can be rigorous anywhere.
More is being built. We'd rather announce things once they work than once we've decided to do them. It makes for a quieter website and much better software.
Places that look like the first one. People trust the product, nobody checks the work, and there's a hard technical core everyone has decided to skip. If that sounds like something you're sitting on, tell us.
Mission & vision
To build software that earns trust instead of asking for it. You can check what it claims. You can see what it charges. Whatever it knows about you, you can take back whenever you want.
A handful of products and partnerships, in different fields, on different continents, all recognisably ours because of how they behave when nobody's watching. Features get forgotten. The standard is what people remember.
“Anyone can tell you what they're able to build. Ask what they've refused to ship.”
Contact
Partnerships, engineering work, press, or just a category you think someone should finally do properly. We read everything. We reply to anything we can actually help with.