UC Berkeley talk says agent systems need ontologies to catch errors prompts miss
AI Engineer · youtube · 2026-07-23
Frank Coyle argues agent systems need ontologies as guardrails
In this UC Berkeley talk, Frank Coyle says many agent failures come from a missing layer outside the model: a formal ontology that acts as logical guardrails.
Problem examples
He points to failure modes such as:
- a second refund on the same order,
- a payout going to the support desk instead of the buyer,
- an order status like “probably shipped.”
Proposed fix
His answer is a neurosymbolic stack:
- probabilistic reasoning stays inside the model,
- logic and constraints live outside it,
- and tool calls are validated before the agent can act.
He suggests using standards such as RDFS and OWL, plus practical checks like Pydantic and ontology validation around a Claude tool-use loop.
Main takeaway
English instructions are too fragile for many real workflows. Typed entities, relationships, and constraints can catch errors that prompt engineering cannot.
More from coding & agent
- Alex Townsend posts 200 open problems in numerical linear algebra for humans and AI agents — IgorCarron · 2026-09-11
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11
- Looking for a classifier of software engineering task shapes to pick models per task — StewartalsopIII · 2026-09-11
- Steal this idea: prompt-to-hardware where agents assemble custom devices — paraschopra · 2026-09-11
- Model Is the Least Interesting Part: A Guide to Six Core AI Architectures from RAG to Multi-Agent — goyalshaliniuk · 2026-09-11
- Non-coder builds layered memory architecture: 20k tokens tracks a year of agent conversations — matteoianni · 2026-09-11