- AI you can check
- We want AI you can hold to account: running on your own machine, giving the same answer twice, and keeping its reasoning close enough to the surface that you can disagree with it. Where a question has an exact answer, we would rather compute it than ask for one.
- Owning the mechanism
- We would rather understand a thing than depend on it. The engines under our tools are written from the format up, with almost nothing pulled in, because the parts we wrote ourselves are the parts we can fix, explain and keep running for years. It is the slower way to build, and it ages far better.
- Privacy first
- The strongest privacy guarantee is a product with no reason to collect anything. Ours are bought outright, which keeps the incentives simple: your files stay on your disk, your work stays yours, and privacy is the default state rather than a setting you have to go and find.
- Fair pricing
- Every tool we sell follows one fair-pricing policy: a price set so that the tool is accessible to everyone who could use it, and ends up in daily use rather than on a shelf. One fair payment, and it is yours. Fair is not the same as cheap. It is the price at which a tool reaches the most people, and we would rather be used by many than paid the most by a few.
- It works on what you already have
- Our tools meet your work where it already is — the windows already open, the files already on disk, in formats other software wrote and other software can still read. A tool should fit around the work that exists, rather than ask the work to rearrange itself around the tool.
- Research that reaches somebody
- A good idea about software deserves both halves: an implementation people can actually use, and the evidence standing next to it. We build our experiments as products you can hold, and publish the numbers behind them where anyone is free to check the claim.