Experience Graphs: The Data Foundation for Self-Improving Agents

2026-06-29

AI Summary

Existing agent frameworks struggle to persist and reuse exploration experience accumulated over long-horizon tasks. This work proposes Trellis, a system that, for the first time, models agent experience as first-class graph database objects, enabling structured, queryable, and governable storage. By integrating vector-guided graph retrieval, materialized views, and time-travel queries, Trellis unifies cross-session experience reuse, crash recovery, horizontal scalability, and closed-loop training within a single architecture. Deployed in Meta’s KernelEvolve production system, Trellis accelerates task completion by approximately 10× while reducing token consumption by 52%.

Abstract

The database community has repeatedly advanced the state of the art by recognizing that new workloads demand new system architectures. We argue that long-horizon agentic tasks -- code generation, scientific discovery, hardware design -- are such a workload. These agents explore: they generate artifacts, execute tools, observe failures, branch, and repair over hundreds of steps. This search produces a structured object we call an experience graph: executable artifacts, tool outputs, rewards, sibling comparisons, and causal lineage. Yet existing agent frameworks treat this experience as disposable state -- JSON checkpoints and session logs that cannot be recovered after a crash, queried across users, or materialized into training data. We propose Trellis: a data foundation that treats the experience graph as first-class, governed, queryable database state. The core insight is that search over experience graphs is a database access pattern. Frontier selection is a query, cross-session reuse is vector-seeded graph retrieval, training-data extraction is a materialized view, and reconstructing what an agent knew at any past step is a time-travel query. When the database owns the experience graph, agents become stateless compute, and crash recovery, horizontal scaling, and a closed-loop training flywheel emerge as architectural byproducts. We ground the design in KernelEvolve, a production accelerator-kernel optimizer at Meta, where cross-session reuse reaches a target speedup roughly 10x faster at 52% lower token cost. More broadly, Trellis turns inference-time search from disposable computation into a durable institutional asset: logs made databases reliable; experience graphs may make agents cumulative.

Problem

• experience graphs
• self-improving agents
• long-horizon agentic tasks
• agent frameworks
• durable experience

Innovation

• experience graphs
• self-improving agents
• database-driven AI
• vector-seeded graph retrieval
• materialized views for training