Extreme Harness Engineering for Token Billionaires: 1M LOC, 1B toks/day, 0% human code, 0% human review — Ryan Lopopolo, OpenAI Frontier & Symphony
April 7, 2026
AI Summary
5 min read"You Can Just Do Things": Inside OpenAI's Zero-Human-Code Engineering Lab
Ryan Lopopolo's team at OpenAI Frontier spent five months building a production Electron app—over a million lines of code, 1,500 pull requests—without writing a single line themselves. The constraint was self-imposed: "If we're trying to make agents that can be deployed into end enterprises, they should be able to do all the things that I do." The result is a working system where humans review code post-merge, agents write their own specs, and the team's daily stand-up is 45 minutes long because they need to fan out understanding of what the agents have done.
The Harness Engineering Mindset
The core insight is that as models improve, the bottleneck shifts from model capability to human attention. Lopopolo's team started with Codex CLI and the Codex Mini model—deliberately weaker than today's models—and found the first month and a half was "ten times slower than I would be." But paying that cost upfront forced them to build the right abstractions. The key move: whenever the agent couldn't do something, they didn't write the code themselves. Instead, they "double-clicked into the harness" and built smaller building blocks the agent could reassemble.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Latent Space: The AI Engineer Podcast
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **🎙️ Introduction: Ryan Lopapolo from OpenAI**
- 2 (01:32) **The Origin of the "Zero Lines of Code" Constraint**
- 3 (03:27) **The Painful First Month: Building the "Assembly Station"**
- 4 (04:42) **Adapting to Model Upgrades (5.2, 5.3)**
- 5 (07:46) **Humans Become the Bottleneck: The Shift to Post-Merge Review**
- 6 (09:35) **Inverting the Environment: The Agent as the Entry Point**
- 7 (11:35) **The Power of Markdown Skills and Agent-Guided Documentation**
+ Full timestamped outline available in the app
Show Notes
We’re proud to release this ahead of Ryan’s keynote at AIE Europe. Hit the bell, get notified when it is live! Attendees: come prepped for Ryan’s AMA with Vibhu after.
Move over, context engineering. Now it’s time for Harness engineering and the age of the token billionaires.
Ryan Lopopolo of OpenAI is leading that charge, recently publishing a lengthy essay on Harness Eng that has become the talk of the town:
In it, Ryan peeled back the curtains on how the recently announced OpenAI Frontier team have become OpenAI’s top Codex users, running a >1m LOC codebase with 0 human written code and, crucially for the Dark Factory fans, no human REVIEWED code before merge. Ryan is admirably evangelical about this, calling it borderline “negligent” if you aren’t using >1B tokens a day (roughly $2-3k/day in token spend based on market rates and caching assumptions):
Over the past five months, they ran an extreme experiment: building and shipping an internal beta product with zero manually written code. Through the experiment, they adopted a different model of engineering work: when the agent failed, instead of prompting it better or to “try harder,” the team would look at “what capability, context, or structure is missing?”
The result was Symphony, “a ghost library” and reference Elixir implementation (by Alex Kotliarskyi) that sets up a massive system of Codex agents all extensively prompted with the specificity of a proper PRD spec, but without full implementation:
The future starts taking shape as one where coding agents stop being copilots and start becoming real teammates anyone can use and Codex is doubling down on that mission with their Superbowl messaging of “you can just build thin
More from this podcast
Latent Space: The AI Engineer Podcast →