zmazz.dev

Project

FHE Trace

Experiment · Active · 2026

Capability-level case study

Exploration of asymmetric proofs of compute for an AI-agents economy, using fully homomorphic encryption as the confidentiality boundary.

Role
Author
Context
Personal
Capabilities
Evaluation, Observability, Security and access control

Overview

If agents are going to sell compute, someone has to prove that the work happened without showing the work. This exploration treats fully homomorphic encryption as a confidentiality boundary for that proof, not as a slogan.

I am building it alone. The private repository description is the only source used here.

Problem

Agent economies need receipts. Plain logs leak; trusted hardware is a different bet. FHE is a way to keep inputs closed while still emitting something checkable. Whether that something is useful is the open question.

Current state

Active exploration (updated 2026). The source repository is private and is not linked from this page. No claim that the cryptography is complete or audited.

Limitations

A description is not a proof. This page must not be read as “I shipped production FHE.” It is a research prototype with an honest gap between intent and evidence.