What does it actually take to build a software factory that delivers better software, faster? As AI coding agents become more capable, the challenge is no longer just getting agents to write code. It is building a measurable, repeatable system that can move work from an idea to a shipped feature while continuously improving along the way.
In this episode of AI That Works, Vaibhav, Dex, and Zach, founder and CEO of Warp, debate what a software factory really is, how it differs from simply adding AI tools to an existing development process, and why defining success is harder than it sounds. They explore the idea of a cloud-based loop where agents can triage issues, write specifications, implement code, review changes, run tests, and help teams ship software.
The conversation gets practical as the group discusses measuring engineering productivity, automating existing workflows, building factories as code, and using real work to improve agent performance. They also examine how teams can gradually introduce automation without rebuilding everything from scratch, why engineers need to stay connected to customer outcomes, and how feedback loops can help organizations identify what is working.
Finally, the discussion turns to the human side of software factories: working in public, psychological safety, risk-based code review, better specifications, and knowing when human judgment is still essential. The central challenge is not simply producing more code. It is building a system that consistently delivers meaningful outcomes for users and the business.
Subscribe to AI That Works — a new episode drops every Friday. Follow the hosts too: Vaibhav (https://x.com/vaibcode)) and Dex (https://x.com/dexhorthy))
GUEST BIO & LINKS
Zach — Founder and CEO of Warp
Zach leads Warp, a development tools company focused on helping developers ship better software more quickly. Warp is known for its terminal and its Warp Factories platform, which helps companies build and improve internal software production workflows.
Website: https://www.warp.dev/
HOSTS
Vaibhav — Co-host of AI That Works
X: https://x.com/vaibcode
Dex — Co-host of AI That Works
X: https://x.com/dexhorthy
KEY TAKEAWAYS
A software factory goes beyond AI coding tools by connecting the full development workflow.
Define success before optimizing performance.
Measure customer outcomes and business impact, not just code shipped.
Build workflows as code to simplify testing and improvement.
Improve existing workflows before redesigning the entire process.
Use agent feedback to identify failures and improve workflows.
Psychological safety and transparency help teams adopt agent workflows.
Apply risk-based code reviews, focusing on critical changes.
Clear specifications and verifiable results reduce review effort.
Focus on delivering consistent, meaningful outcomes—not just more code.
TIMESTAMPS
– What Is a Software Factory?
– Meet the Hosts and Zach from Warp
– The Software Factory Debate Begins
– Defining the Modern Software Factory
– Building Factories as Code
– Measuring Yield and Engineering Productivity
– Why Engineers Need to Own Customer Outcomes
– From Individual AI Tools to Automated Workflows
– Start Small and Improve the Existing Process
– Metrics, Feedback Loops, and Continuous Improvement
– Agents That Improve the Factory Itself
– Measuring Automation, Human Steering, and Cost
– Multiplayer Workflows and Psychological Safety
– Code Review as Risk Management
– Risk-Based Reviews and Smaller Pull Requests
– Better Specs and Verifying Agent-Built Software
– Planning, Prototyping, and Turning Specs into Code
RESOURCES
Warp — Developer tools and Warp Factories: https://www.warp.dev/
Terraform — Discussed as a comparison for defining workflows declaratively.
Claude Code and Codex — Examples of coding agents used in development workflows.
DORA metrics — Discussed in the context of engineering productivity and software delivery.
Terminal-Bench and SWE-bench — Referenced as examples of benchmarks that can inspire evaluations built around real company tasks.
Slack, Jira, and MCP — Mentioned as ways to connect work and agent interactions to a broader software workflow.
TOPICS COVERED
Software factories, AI coding agents, agent orchestration, software development lifecycle, engineering productivity, engineering metrics, customer outcomes, workflow automation, continuous improvement, factory-as-code, cloud-based development, human oversight, code review, risk management, software testing, specifications, psychological safety, collaborative workflows, developer experience, and business impact.
HASHTAGS
#AIThatWorks #ArtificialIntelligence #CodingAgents #SoftwareEngineering #SoftwareFactories #DeveloperTools #Automation #DevTools #AIEngineering
Transcript
I've watched this one closely. Ask me anything about it and I'll point you to the exact moment.