Building personal intelligence
for everyone.
We are a team of systems engineers, ML researchers, and product builders who believe AI should work for individuals — not the other way around. Founded in late 2024.
Why we exist
The current generation of AI assistants is reactive. They wait for prompts, require you to context-switch, and add yet another tool to an already overloaded digital stack.
We believe the next paradigm is not conversational — it is anticipatory. An AI that truly understands you observes your digital context and synthesizes actionable proposals before you even recognize the need.
What we believe
We believe in artificial personal intelligence — AI that is narrow in scope but deep in understanding. For the first time, individuals can aggregate their own data and leverage intelligence for themselves, rather than relying on platforms to provide value.
Privacy is non-negotiable
User data is sacred. If it can be processed locally, it will be. No exceptions, no compromises, no backdoors.
Deterministic over probabilistic
We value reproducible, testable outcomes over stochastic guesses. Every system we build must be auditable.
Builders, not talkers
We ship weekly. The best argument is working code. Meetings are expensive; pull requests are cheap.
Radical transparency
Our architecture is documented. Our privacy model is verifiable. We believe sunlight is the best debugger.
How we build
Boxy is built on three pillars:
Everything that can happen inside the Box, happens inside the Box. NER, obfuscation, and embedding all run locally. Raw data never leaves your machine.
Our agents follow pre-trained manifests. Every action is recorded, reproducible, and auditable. No hallucination and no guessing.
When cloud computation is necessary, we operate on obfuscated data only. The math works without the metadata.
Want to build with us?
We are hiring across engineering, ML, and design.