Trusted Execution · Minimal by Design

Horseman

Move to tilt · Click to scatter

A small deterministic language and virtual machine for trusted execution.

Most enclave runtimes bring a large software stack into the trusted boundary. Horseman tests a smaller model: a compact interpreter with explicit input and output.

The Research Question

What if the safest enclave runtime is the one that refuses to be general-purpose?

General runtimes favour compatibility. Horseman removes capabilities until every remaining part can be inspected, bounded, or declared.

Read the architecture
Not a container runtime A purpose-built language boundary
Not an operating system inside SGX A small VM with two explicit doors
Not a production claim A research hypothesis to measure

Horseman project overview

Core Constraints

Three rules shape everything else.

01

Minimal TCB

A core small enough to read and audit.

02

Deterministic Execution

Execution does not change with secret data.

03

A Closed World

Inputs are explicit. Results leave sealed.

Source code at first release

The interpreter and documentation will be published with the first research prototype.

2 people following
the project
HSL / Access gate Confirm the signal. Run a private verification before sending.

Release notes only. Your address is encrypted and kept by Horseman.