AI EngineerVideoSource Linked

AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix

Netflix’s performance-agent playbook feeds production profiles and exact deployed code to agents, then relies on canaries and engineers to validate proposed optimizations.

AI Engineer · Jul 28, 2026

Open Source Open Markdown Open JSON

Source Summary

Netflix found a hidden **quadratic-time pattern** by giving an agent call-stack profiling data, then locating the matching code at the production commit. The proposed workflow turns profiles into candidate fixes instead of relying only on manual flame-graph inspection.

Practical Implication

Build a shared Git-backed catalog of observed performance patterns, and let coding or review agents consult it before inefficient code lands. A **stateful catalog plus a stateless LLM** can spread production lessons across services without requiring a vector database.

Agent-Ready Context

Netflix found a hidden **quadratic-time pattern** by giving an agent call-stack profiling data, then locating the matching code at the production commit. The proposed workflow turns profiles into candidate fixes instead of relying only on manual flame-graph inspection.

Build a shared Git-backed catalog of observed performance patterns, and let coding or review agents consult it before inefficient code lands. A **stateful catalog plus a stateless LLM** can spread production lessons across services without requiring a vector database.

Treat the profiler as an estimate and the **canary as ground truth**; error rates and an engineer’s decision remain release gates. The talk favors fixed, level-two orchestration because greater autonomy demands more evaluation, sandboxing, and prompt-injection defenses.

Connected Context · Feed7 Judgment

This turns the prior signal-to-fix pattern into a performance-specific harness: join profiler evidence to the exact production commit, consult a durable Git catalog, propose a bounded change, and treat canary results plus engineer approval as release gates. It also narrows the appropriate autonomy level, arguing that richer orchestration carries additional evaluation, isolation, and injection-defense costs.

From Signal to PR: Anatomy of a Self-Improving Agent — Jason Lopatecki, ArizeNetflix instantiates Arize’s general signal-to-PR loop for performance work, using profiles as the production signal, commit-matched code as context, and canaries to verify proposed fixes.Don't Build Agents You Can't Answer For — Addy OsmaniProfiler evidence, canary measurements, error rates, and an explicit engineer decision are concrete forms of the explainability, verification, and ownership Osmani requires before agent-authored changes ship.How Forward Deployed Engineering is done at Factory — Eno ReyesThe workflow reinforces Factory’s case for instrumenting the path from operational signal to deployment and placing validators at the autonomy boundary; here the canary is the decisive validator.Effective harnesses for long-running agentsBoth use Git-backed state to carry knowledge across otherwise stateless agent runs, but Netflix stores reusable performance patterns rather than session progress and feature status.

Context Map

agentcoding#harness-engineering#agent-memory#agent-reliability

Uncertainty

Treat the profiler as an estimate and the **canary as ground truth**; error rates and an engineer’s decision remain release gates. The talk favors fixed, level-two orchestration because greater autonomy demands more evaluation, sandboxing, and prompt-injection defenses.