Eidola

Eidola envisions AI as a tool for liberation rather than subjugation — a technology that individuals can wield in assertion of their own agency.

That future will only exist if we demand it now, and Eidola is laying the groundwork.

Intelligence should be intimate

The most useful AI is the one that knows you best — your habits, your work, your conversations, the shape of your days. That knowledge is powerful and deeply personal. It should stay under your control, and yours alone.

Most AI tools ask you to trade that intimacy for capability — to trust that whoever runs the systems will handle what they learn about you well. Eidola is built on the opposite premise: that the usefulness of such a tool is undermined when trust is coerced.

We designed our product to offer the confidentiality and integrity of an application running locally, even when workloads demand the resources of a datacenter:

• AI runs in client-verified enclaves sealed even against us, on publicly auditable, bit-for-bit reproducible software that we cannot change behind the scenes.
• Your content is never stored, logged, or reviewed, and the minimal metadata we keep is structured to resist linking requests to each other or to you.
• Releases require an attestation, signed by an accountable human, certifying that each change — and the motivation behind it — complies with our enumerated privacy guarantees.

In early development

Eidola is being built in the open. You can read every line, reproduce every build, and decide for yourself: start with how to think about Eidola, read the privacy guarantees, or go straight to the source.