MVM
A strict, cross-platform microVM runtime for safely executing AI agents and untrusted workloads.
I’m Ari Lerner. I’ve spent a long time working across systems software, AI infrastructure, security, and product—writing code and mathematics, building hardware, and teaching what I learn.
Selected systems
Runtime boundaries, content-addressed systems, and infrastructure that makes powerful software safer and simpler to operate.
A strict, cross-platform microVM runtime for safely executing AI agents and untrusted workloads.
An object-first runtime exploring deterministic execution, portable artifacts, zero-copy data paths, and compute that travels with its guarantees.
Research into semantic, content-addressed object identity and a topology where software can address meaning rather than storage location.
Working thesis
The systems I care about make capability explicit, complexity containable, and trust inspectable.
01
Powerful workloads need hard machine boundaries, attenuated authority, and a blast radius the operator can understand.
02
The best infrastructure does not expose every mechanism. It exposes the guarantees people need to build with confidence.
03
Policy, identity, provenance, and auditability should travel with the system—not live in an operator’s memory.
Writing
Essays about secure AI systems, Rust, infrastructure, learning, and the human reasons we build.
A growing body of checked mathematics can still leave its central theorem unresolved, so the repository has to show exactly where the gap remains.
PUT and GET were enough for blobs, until a release needed names, dependencies, signatures, and one visible moment of publication.
Copying memory can restore a process on the same host; moving a program somewhere else requires its dependencies, runtime contract, and fresh authority.
The long arc
I’ve spent a long time moving between code, math, hardware, systems, and teaching. The details change, but I keep returning to the same question: how do we make powerful technology useful without making it careless?
The longer storyOpen channel
I’m always interested in hard systems problems, serious open-source work, and people building new leverage without ignoring trust.