Podcast Summary: Behind the Craft

Episode: Full Tutorial: Use AI Agents for Coding AND Product Management | Eno Reyes (Factory)

Host: Peter Yang
Guest: Eno Reyes, Co-founder of Factory
Date: February 15, 2026

Episode Overview

In this episode, Peter Yang welcomes Eno Reyes, co-founder of Factory, creators of a leading AI coding agent called Droid (also referred to as Joy in the conversation). Eno offers a deep dive and live demo on how advanced engineers leverage AI agents for software development and product management. The discussion centers on Factory’s unique enterprise-first approach, best practices for working with AI coding tools, the evolving nature of product management roles, competing in a rapidly growing AI agent space, and how tools like Droid democratize codebase access and collaboration for all roles—not just developers.

Key Discussion Points & Insights

1. Why Factory Focused on Enterprises from Day One

Timestamps: 01:24–02:52

• Factory differentiated itself by addressing the needs of large companies (10k+ employees) and their engineering VPs, focusing on security, enterprise controls, ROI analytics, and integrations.
• Reyes notes the distinction: “We set out on a long journey to build these AI systems… There’s sort of a lot of layers that make us feel maybe like the full enterprise solution to software development agents.” (01:36)
• Factory’s Droid agent offers full-stack integrations beyond just the terminal or IDE, doing cross-company codebase analysis and surfacing blockers for agent-driven development.

2. Live Demo: Building a Speed Reading Web App with Droid

Timestamps: 02:55–07:11, 08:01–11:59

• Reyes shares an example project: a fast reading (speed read) web app.
• The workflow:
  • Use the terminal extension for Droid, set “high autonomy” (agent can act with little user intervention).
  • Paste in a conversational transcript as the build spec—Droid interprets and generates the prototype, autonomously planning, building, testing, and QA’ing changes, including opening browsers, taking screenshots, and running tests.
  • Factory’s autonomy modes let users control agent permissions flexibly (allow all/some commands).
  • Reyes shows how Droid supports multiple models (Opus, GPT-5.2, etc.) for planning vs. execution.
• Quote: “Where Droid shines the most is on things like basically long running tasks … we’ve done a lot around things like compaction or compression and prompt caching to make the experience feel really nice.” (04:54)

3. Spec vs. Plan: How Droid Structures Work

Timestamps: 08:01–10:45

• Droid has a unique “spec mode” (different from ordinary planning mode).
  • Spec: What should be built.
  • Plan: How to build it.
  • Droid queries the user for clarifications (e.g., input formats, features), then proposes and saves a spec that can be edited in VS Code.
• Reyes highlights the importance of separating “what” from “how,” enabling better collaboration and reusable specs.

4. Models, Costs, and Validation Best Practices

Timestamps: 10:45–12:08, 12:08–14:28

• Droid enables model selection mid-session (e.g., plan with Opus 4.5, execute with GPT-5.2), optimizing for both capability and cost.
• Extensive focus on validation:
  • Automatic linting, type checking, QA (including visual validation with screenshots and console message reporting).
  • Quote: “Agents are fundamentally bottlenecked by the ability to validate their own work … Our view is that the agent is the one that needs to validate its work to move to the next step. The quality of the output is way higher.” (13:23)
• Workflow smoothness: Higher autonomy can be set or changed on the fly for convenience or control.

5. Design System Integration and Grounding

Timestamps: 15:02–16:29

• Droid is adept at reading and grounding its UI in a codebase’s established design system—automatically applying branding, styles, and components without explicit “design system” skills.
• Quote: “What a lot of people underestimate is that building stuff that’s in your design system… doing it well is actually fairly difficult. Droid can do that quite well.” (15:42)

6. How (and When) to Use Skills, Hooks, Subagents, and MCP

Timestamps: 16:29–18:44

• Reyes demystifies Factory’s extensibility: skills, hooks, sub-agents, and MCP (multimodal command processing) are all supported and power shared workflows.
• Usage insights: Most enterprise users rely on a handful of customizable skills and integrations, and Factory enables org-wide management of these at scale.
• Quote: “From enterprises… a couple of people focus in on making skills, MCPs, tools for their whole organization or big teams… It’s just easy to get everyone in your 10,000 person company outfitted with a skill that meaningfully changes dev productivity.” (18:11)

7. Product Management Skill: A Deep Dive

Timestamps: 19:00–22:34

• Reyes demos his Product Management skill (used for reviewing PRDs, specs, design docs, feature prioritization).
  • Gathers live content from Notion (native integration), including product principles, core value prop, “11-star experience” framework (inspired by Airbnb), templates, etc.
  • Ensures generated product docs reflect Factory’s philosophy (“Our five-star experience two years ago is now our baseline.”).
• Quote: “This is probably one of my favorite skills that I have. … The structure of it ends up looking a lot more like the types of things that if you’ve been in the room at Factory for a year, you would say, instead of just what Opus 4.5 is randomly opining on.” (21:12, see also start of transcript)

8. Evolving Roles: Product Engineer vs. “Regular PM”

Timestamps: 22:34–25:05

• Factory’s org model only hires “product engineers”—no traditional PMs. All roles are expected to drive most workflows with AI agents.
  • Even sales/AE roles are power users of Droid (using skills for analysis, customer work, CRM, etc.).
  • Quote: “What it means to be really any role has changed a lot… it’s quite clear for us.” (23:44)
• AI agents as general productivity overlays—less about code, more about being able to articulate tasks and review or plan in natural language.

9. The Terminal as an Overlay, Not a Destination

Timestamps: 25:05–26:32

• Reyes explains why the agent UX moves beyond the old “all-in-one” IDE-centric mindset.
  • Agents should act as overlays, not full-screen destinations: always-on, able to operate across desktop apps, files, workflows.

10. Competing in a Crowded AI Agent Space

Timestamps: 27:07–29:48

• Factory (40-person team) competes successfully against much bigger, well-funded teams (Cloud Code, Cursor, etc.).
  • Key: Focus on hard enterprise problems (hierarchical controls, customizable security, “air-gapped” operation in highly secure contexts).
  • Success on benchmarks stems from real enterprise datasets and solving tough problems, not public leaderboard hill climbing.
  • Quote: “There’s just so much to be explored in AI for software development… just opening Twitter and reading a couple workflows, you realize the variance… is so high.” (27:46)

11. AI Agents for Legacy Code, Democratizing Access

Timestamps: 29:48–31:43

• Major value: agents shine at detail-oriented, unpleasant engineering work (e.g., legacy code refactoring) and onboarding by reading, contextualizing, and surfacing insights from gnarly codebases.
• Extends utility to QA, ops, data, and other adjacent roles—“democratizing access to what used to be a very complex and hard to understand topic.”

12. Rethinking Codebase Access and Enterprise Education

Timestamps: 31:43–33:37

• Reyes advocates for broad codebase read access: “A bunch of companies are going to pay a huge cost to design decisions like ‘we’re not a monorepo’ or ‘limiting codebase access to only these personas’… it won’t age well into the AI era.” (31:57)
• Educating enterprises about activation and advanced user journeys is as crucial as product features themselves. Power users become internal evangelists, driving broad adoption.

13. Droid for All: Getting Started

Timestamps: 34:06–34:40

• Droid is free to use (with credits for new users): “Just go to Factory AI. … One line and you’re in.” (34:16)

14. Final Thoughts & Where to Find Eno

Timestamps: 34:40–35:16

• Host praises Factory’s focus and speed: “It all comes down to focus. I’m super impressed with your progress.”
• Find Eno on Twitter: @enoreyes

Notable Quotes & Memorable Moments

• “Agents are fundamentally bottlenecked by the ability to validate their own work.” (13:23, Eno Reyes)
• “At the beginning [Droid] does this grounding step… looking at our CSS, different pages, and it’s using that to ground its UI.” (16:22, Eno Reyes)
• “The hardest software problems is like… refactoring these gnarly legacy codebases, right? … Droids just do that stuff pretty well.” (29:56, Reyes)
• “What it means to be any role has changed a lot … you definitely need a willingness to drive your workflows with AI if you want to work in any role at Factory.” (23:44, Reyes)
• “I think enterprises should just give everyone access to the codebase. … This stuff is not going to age well [without it] into the AI era.” (31:43, Yang, Reyes)
• “You can just get it digested [by the agent] for pretty cheap.” (31:40, Reyes)

Episode Timeline (Selected Timestamps)

• 01:24 – Factory’s enterprise focus
• 02:55 – Kicking off live build demo
• 03:47 – Demoing Droid autonomy modes and interface
• 08:01 – Spec vs. plan; editing and validation flow in Droid
• 11:04 – Model selection for plan/execution and cost management
• 13:23 – Validation as core agent differentiator
• 16:00 – Design system auto-integration; agent “grounding” step
• 19:00 – Deep dive: Product management skill
• 22:34 – Only hiring “product engineers”; the modern PM role
• 27:30 – Competing as a lean team
• 29:56 – Solving “the hard stuff” (legacy code)
• 31:43 – Rethinking access and onboarding in the AI era
• 34:16 – Getting started with Droid
• 34:57 – Where to find Eno

Conclusion

This episode offers a tactical and philosophical master class in deploying AI agents (like Droid) for both engineering and product management at scale. Eno Reyes provides actionable insights for enterprise teams and solopreneurs alike—emphasizing validation, adaptive workflows, custom skill-building, the blurring of traditional tech and product roles, and why the next generation of work will be driven by universally accessible agents rather than exclusive “power tools.” If you want to see what the near future of product leadership and software development looks like with AI at its core, this episode is a must-listen.

Find out more:

• Factory AI: factory.ai
• Eno Reyes on Twitter: @enoreyes