Skip to content

Ryan Lopopolo

Member of technical staff at openai, specifically on openai-frontier — the team building the enterprise platform for deploying agents safely at scale. Background: Snowflake → Stripe → Citadel → OpenAI.

Spoke at AI Engineer London 2026 on harness-engineering, then followed up on Latent Space (April 2026) with deeper implementation details — see ryan-lopopolo-latent-space-symphony-2026. For the nine months preceding the talk he has "had the privilege of building software exclusively with agents" — and has banned his team from touching their editors, forcing all work through models. The proof case: a 1M-LOC Electron app with 0% human-written and 0% human-reviewed code pre-merge.

Positions

  • Self-described "token billionaire" — frames the AGI transition as democratizing this status so everyone has 5/50/5,000 engineers' worth of capacity 24/7.
  • Code is free. Produce, refactor, delete without hesitation. Maintenance burden reframes as "synchronous attention drain" that agents no longer cause.
  • Humans steer, agents execute. The talk's subtitle. Human time is the scarce resource; the job of the senior engineer is to move synchronous human time into higher-leverage activities (design, spec, guardrails, skills).
  • Implementation is no longer the scarce resource. Skill sets shift toward systems thinking, system design, and delegation.