Research.
Notes, experiments, and findings from our work on environments, evaluation, and reinforcement learning.
4 posts
Searching over learning systems
A learner improves within a design. Optimus investigates the design itself: the interfaces, algorithms, data, and feedback that determine what the learner can discover.
research · SEP 12, 2026
Discovering simpler systems through optimization
Optimization can uncover behavior that needs less machinery than the search that found it. The next question is which weights, decisions, and dependencies can disappear without losing the capability.
research · AUG 24, 2026
The limits of agent scaffolding
A scaffold controls what a model sees, which actions it can take, and when it tries again. Its value depends on how those choices change complete-task performance after their costs are counted.
research · JUL 29, 2026
Observability as a constraint on intelligence
When different states produce indistinguishable evidence but require different actions, reasoning alone cannot resolve the decision. We examine how observation design changes what an agent can achieve.
research · JUN 23, 2026
Let’s explore what comes next.
Working on a related question? We’d like to hear what you’re investigating and where Arkenos could help. Write to dev@iaconautonomics.com or book a conversation.
Meet the team