Partners needed to evaluate long-horizon LLM agent failures in real workflows.

We are looking for partners to evaluate and reduce long-horizon failure modes in LLM agents performing multi-step tool use.

Challenge: “horizon blindness” and drift in extended sequences, especially when errors are irreversible.

What we provide: a governance layer (FIONA) and an evaluation approach focused on deterministic, auditable traces and clear failure definitions.

We seek workflow owners and research teams to collaborate on reproducible evaluations and applied pilots. NDA available for implementation details.

We are also open to conversations with strategic investors aligned with this space.