Alex Krentsel presents Exo, a self-improving AI agent architecture
Overview
UC Berkeley PhD candidate and Google Research senior researcher Alex Krentsel presented Exo at Arena HQ, according to a post from Arena.
Exo borrows the exokernel idea from operating systems, separating an agent into a protected mechanism layer (state, history, secrets, and a "guardian") and a fully evolvable policy layer.
Arena says Exo was 31-62% cheaper than Codex and Pi on open-world RuneScape tasks. The post does not give further details of the tests, and the figures come from Arena's own account.
Written by AI from the articles below · updated Oct 29, 1:58 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
Arena.ai@arenaOfficialExo lets AI agents self-improve their harness using an exokernel designAIUC Berkeley PhD candidate and Google Research senior researcher Alex Krentsel presented Exo, an agent architecture that separates a protected mechanism layer from an evolvable policy layer. Arena says Exo was 31-62% cheaper than Codex and Pi on open-world RuneScape tasks.
Heat Trend
Not enough continuous observations to show a trend yet.