The mathematics behind Datasent
Datasent is built on a small set of precise ideas: deterministic representation, structure-aware encoding, and residual-only transmission. This is where we publish the full framework — the proofs, the system design, and the open questions we’re still working through.
Our mission
Keep raw data where it belongs
Most systems move data first, then try to protect it. Datasent takes a different approach. By encoding data against a shared model basis, only the unpredictable residual ever needs to move. The original stays in place — exact, recoverable, and under control.
Privacy-First Data
How Datasent approaches data
Datasent replaces raw data movement with a structured encoding layer. Sender and receiver agree on a shared model basis upfront, allowing data to be represented through its structure rather than transferred in full.
Articles
Research, ideas, and practical guides
Feb 2026
John Rhye
Why Raw Data Shouldn't Travel
The assumption that raw data must move between systems to be useful is baked into almost every layer of modern infrastructure. Datasent was built to break it.
Feb 2026
Ben Wilson
Lossless by Design: How Datasent Guarantees Exact Reconstruction
Most compression systems trade fidelity for size. Datasent refuses that bargain. Here is how exact reconstruction is guaranteed, mathematically, for every data type.
May 2026
John Rhye
The Trusted Setup: How Datasent Keeps Raw Data Local
A trusted setup is often associated with cryptographic ceremonies and secret key generation. Datasent's version is simpler — and more powerful. No secrets required.
Apr 2026
John Rhye
AI's Data Problem — and Why It Starts at the Infrastructure Layer
Federated learning keeps gradients local. Clean rooms control access. But neither keeps raw data where it belongs and supports lossless, governed, multimodal sharing. Datasent does.
Founder’s Note
Why we stopped moving raw data
Most data systems assume the same thing: data has to move before anything useful can happen. That assumption drives cost, risk, and complexity across every layer of infrastructure.
Datasent was built to remove it. By separating structure from unpredictability, we make it possible to work with data without transferring it in full — keeping raw data local while still allowing exact reconstruction when needed.
Join our newsletter
Get updates on research, product news, and privacy-first data insights.
