The non-technical PM’s guide to building with Cursor | Zevi Arnovitz (Meta)
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
This episode features Zevi Arnovitz, a product manager at Meta who has developed a revolutionary workflow for building sophisticated products using AI without any technical background. (04:48) Zevi shares how he transitioned from being afraid of code to shipping real features using Cursor, Claude Code, and a series of custom slash commands. The conversation covers his complete AI-powered development workflow, from ideation to shipping, demonstrating how non-technical PMs can leverage multiple AI models to build production-ready applications. (07:41)
- Core theme: Democratizing software development through AI, enabling non-technical product managers to build and ship real products independently
Speakers
Zevi Arnovitz
Zevi is a product manager at Meta with a completely non-technical background who discovered his passion for building with AI after watching YouTube videos about AI app development while traveling in Japan. (06:06) Previously a PM at Wix and a psychology student, he has built multiple side products including StudyMate (a platform for creating interactive tests from study materials) and has developed a sophisticated AI workflow that even engineers at Meta ask him to teach them.
Lenny Rachitsky
Host of Lenny's Podcast and author of Lenny's Newsletter, one of the leading publications for product management insights. Lenny regularly interviews top product leaders, entrepreneurs, and innovators to share actionable insights with ambitious professionals.
Key Takeaways
Start with AI Projects, Not Code Editors
Zevi recommends beginning with ChatGPT projects rather than jumping straight into Cursor or other coding environments. (12:18) He created a "CTO" project with custom instructions to act as a technical co-founder who challenges ideas rather than just agreeing. This approach provides exposure therapy to code while maintaining a conversational interface. The key is gradual progression: ChatGPT projects → Bolt/Lovable → Cursor, allowing non-technical people to build confidence before diving into more advanced tools.
Use Multiple AI Models for Code Review
Rather than relying on a single AI model, Zevi employs a "peer review" system using different AI models to review each other's code. (38:40) He personifies each model with distinct characteristics - Claude as the communicative CTO, GPT as the brilliant but uncommunicative coder, and Gemini as the creative designer. This multi-model approach catches more bugs and provides diverse perspectives on technical solutions, compensating for his non-technical background.
Create Systematic Workflows with Slash Commands
Zevi developed a complete development workflow using reusable slash commands in Cursor: create issue → exploration → create plan → execute → review → peer review → update docs. (14:55) These commands automate prompts and ensure consistent processes. Each phase has a specific purpose, from quickly capturing ideas in Linear to thorough technical planning before execution. This systematic approach prevents common AI pitfalls like jumping straight to coding without proper planning.
Treat AI Learning as Continuous Improvement
When AI makes mistakes, Zevi uses a "learning opportunity" slash command and conducts postmortems to understand root causes. (46:04) He asks the AI what in its system prompt or tooling led to the error, then updates documentation and prompts accordingly. This creates a feedback loop where the AI workflow becomes progressively smarter and more reliable, turning failures into systematic improvements rather than recurring problems.
Focus on Being a 10x Learner, Not a 10x Doer
Based on his early career failure at Wix, Zevi learned that junior professionals should prioritize learning over trying to impress with immediate output. (62:54) He identified each team member's strengths and used them as mentors for specific areas. This approach made his eventual success feel like a shared victory rather than competition. With AI tools, this mindset becomes even more powerful as non-technical people can learn complex concepts through AI tutoring while building real products.
Statistics & Facts
- Zevi built StudyMate, a functional educational platform with quiz generation, user authentication, and payment processing, entirely using AI tools despite having zero technical background. (17:37)
- He completed full localization of StudyMate from Hebrew to English in just two days, a process that would typically take a development team weeks to complete. (35:17)
- Zevi built a personal website from no domain to live deployment in just one and a half hours using AI tools. (35:17)
