# How Zach Lloyd actually uses coding agents

Co-founder and CEO of Warp

**If you take one thing:** `Scorer + self-improvement agents`, agents that fix agents. scorer agents grade every agent run, and other agents turn low scores into PRs that fix the factory's skills.

- **Agent:** Warp Factories (web) ([source, 2026-08-21](https://x.com/zachlloydtweets/status/2090900505300177181))
- **Agent:** Warp agent (app) ([source, 2026-09-18](https://www.chatprd.ai/how-i-ai/inside-warps-software-factory))
- **Model:** GPT 5.6 Sol ([source, 2026-09-09](https://x.com/zachlloydtweets/status/2097746902418948468))
- **Review:** Engineer reviews the agent PR and ships it; QA video attached ([source, 2026-09-28](https://x.com/zachlloydtweets/status/2104565422439932412))
- **Harness:** Warp Factories defined in one config file: Slack request, triage, spec, implement, review, QA ([source, 2026-09-30](https://x.com/zachlloydtweets/status/2105364556482511042))
- **MCP servers:** figma, granola ([source, 2026-09-18](https://www.chatprd.ai/how-i-ai/inside-warps-software-factory))

## In their words

> Here, we use the Warp Terminal to kick off a run in a Factory and see it update on our web app alongside hundreds of other agent runs from our team.

> So, instead of fumbling with the UI, he uses a coding agent in Warp that has access to the Figma MCP (Multi-modal Command Palette).

> - Benchmarking models on our own data to find the most efficient cost/performance pick. GPT 5.6 Sol (high) today

> One big change is how we approach code review - for low risk changes, we now let engineers review agent PRs and ship themselves, rather than requiring a second dev’s review.

> We run all of our company's agents out of this configuration file. It's Terraform for agents


Archetype (our label): Orchestrator. Newest source: 2026-10-01.
Card: https://www.prehistoriq.com/stacks/zachlloyd/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/zachlloyd.json