Lenny's Podcast: Product | Career | Growth

The non-technical PM’s guide to building with Cursor | Zevi Arnovitz (Meta)

January 18, 2026

Key Takeaways Copied to clipboard!

  • Non-technical individuals can build sophisticated, real products by adopting a structured, AI-assisted workflow using tools like Cursor and specialized LLMs. 
  • The main challenge in AI-assisted development shifts from writing code to effectively reviewing it, which can be mitigated by using multiple AI models (like Claude and Codex) in a 'peer review' system. 
  • The current era presents an unprecedented opportunity for junior professionals to become builders, as AI tools lower the barrier to entry for creating startups and complex projects. 
  • The default professional approach for those growing up with AI is to immediately leverage AI first to solve any new challenge or problem, such as interview preparation. 
  • AI-assisted interview preparation is powerful for creating personalized coaches and mock interviews, but human mocks remain the biggest game-changer for competitive roles like Meta PM interviews. 
  • The most valuable mindset for junior professionals is prioritizing being a '10x learner' over being a '10x doer,' as demonstrated by Zevi Arnovitz's early career success at Wix. 

Segments

AI Superpowers for Non-Technical Builders
Copied to clipboard!
(00:00:00)
  • Key Takeaway: AI tools like Cursor with Claude Code grant non-technical individuals superpowers, enabling them to build sophisticated products independently.
  • Summary: Non-technical PMs can now build real products, challenging the notion that technical skills are prerequisites. The main hurdle for AI-assisted builders is reviewing the generated code. Techniques like using slash commands for AI review (e.g., slash review) are essential for quality control.
Zevi’s Background and AI Catalyst
Copied to clipboard!
(00:04:48)
  • Key Takeaway: Exposure to AI app-building tools like Bolt or Lovable served as the catalyst for Zevi to immediately start building, feeling like he gained superpowers.
  • Summary: Zevi has zero technical background, having studied music in high school and not being in a tech unit in the army. Watching videos of others building apps with AI inspired him to start building immediately upon returning from a trip. His goal for sharing his workflow is to inspire others to open their computers and start building.
CTO Project for Technical Guidance
Copied to clipboard!
(00:07:41)
  • Key Takeaway: Creating a dedicated GPT project with a ‘CTO’ custom prompt mitigates the risk of AI agents being overly agreeable and writing poor code without proper planning.
  • Summary: Early AI builders often face issues because coding agents are too eager to write code without planning complex features like payments. Zevi created a ‘CTO’ persona with a custom prompt to act as the technical owner, challenging his ideas and ensuring architectural soundness before execution.
Gradual Tool Adoption Path
Copied to clipboard!
(00:12:19)
  • Key Takeaway: New builders should adopt AI coding tools gradually, starting with less intimidating environments like GPT projects before moving to full IDE integration like Cursor.
  • Summary: The progression path involves starting slow with a GPT project to focus on conversation and learning, then graduating to tools like Bolt or Lovable, and finally moving to Cursor in light mode before going full developer mode. This gradual exposure acts as essential ’exposure therapy’ for those intimidated by code.
Structured AI Workflow via Slash Commands
Copied to clipboard!
(00:14:41)
  • Key Takeaway: A reusable workflow, automated via slash commands within Cursor, structures the development process from ideation to documentation.
  • Summary: Zevi uses slash commands (e.g., /create issue, /explore, /plan, /execute, /review) to automate his entire development lifecycle within Cursor, powered by Claude Code. This system ensures structured ideation, planning, execution, and review, even for non-technical users.
StudyMate Feature Build Demonstration
Copied to clipboard!
(00:17:18)
  • Key Takeaway: Zevi demonstrated building a new feature (fill-in-the-blank questions) for his revenue-generating side project, StudyMate, using his AI workflow.
  • Summary: StudyMate allows students to upload materials and generate interactive tests, currently only offering multiple-choice questions. The live build focused on adding fill-in-the-blank functionality with a drag-and-drop interface, initiated by the /create issue slash command linked to Linear.
Model Specialization for Task Execution
Copied to clipboard!
(00:30:22)
  • Key Takeaway: Different LLMs should be strategically used for specific parts of the plan, leveraging strengths like Gemini 3 for UI and Composer for speed in execution.
  • Summary: The markdown plan generated by Claude is split to assign front-end work to Gemini 3 due to its superior UI capabilities, while Composer is used for fast execution of less complex tasks. This strategy maximizes efficiency by playing to each model’s strengths.
AI Peer Review and Model Personification
Copied to clipboard!
(00:38:33)
  • Key Takeaway: Code review is formalized by having the primary agent (Claude) review its own work, supplemented by reviews from competing models like Codex to ensure quality.
  • Summary: Zevi personifies models: Claude as a communicative CTO, Codex as the reclusive expert for tough bugs, and Gemini as a terrifyingly artistic scientist. The ‘peer review’ slash command pits these models against each other, forcing the primary agent to justify or fix issues flagged by others.
Workflow Post-Mortems for Continuous Improvement
Copied to clipboard!
(00:45:40)
  • Key Takeaway: Continuously updating AI tooling and prompt documentation based on post-mortems of model failures is a critical hack for long-term productivity gains.
  • Summary: When an AI makes a mistake, Zevi asks the model to introspect on the root cause within its system prompt or tooling. This knowledge is then used to iterate on the slash commands and documentation, ensuring the AI system becomes progressively smarter and avoids repeating errors.
AI Impact on PM Craft and Junior Roles
Copied to clipboard!
(00:53:42)
  • Key Takeaway: AI tools enhance, rather than atrophy, PM skills by providing reps at a higher level of strategic thinking and execution, especially benefiting junior PMs.
  • Summary: The PM role is about harnessing tools to deliver user solutions quickly, not about always having the right answers. Junior PMs gain invaluable experience by using AI to explore complex strategy and marketing decisions on side projects, effectively getting high-level practice.
AI-Assisted Interview Preparation
Copied to clipboard!
(00:58:15)
  • Key Takeaway: Zevi treated AI as a personal coach, using a dedicated Claude project and custom-built games (like one in Base44) to practice specific interview challenges like product segmentation.
  • Summary: Zevi adopted an ‘AI-first’ approach to problem-solving, consulting an AI coach project for interview strategy and frameworks from experts like Ben Ares. He even built a specific game to drill segmentation questions, treating AI interaction as a natural part of professional development.
AI-Assisted Interview Preparation
Copied to clipboard!
(00:58:36)
  • Key Takeaway: Zevi used Claude to create a dedicated ‘coach’ project fed with top interview frameworks, like those from Ben Ares, for consultation and mock interviews.
  • Summary: Zevi prepared for his Meta PM interviews by creating a dedicated project in Claude, feeding it top frameworks from experts like Ben Ares, and using it as a coach for each preparation phase. He also built a quiz game in Base44 to practice product segmentation challenges. The AI-driven preparation was supplemented by human mock interviews, which Zevi found to be the biggest game-changer.
AI Feedback Loop Value
Copied to clipboard!
(01:01:02)
  • Key Takeaway: AI excels at providing objective, critical feedback on recorded interviews, filling a crucial gap where humans rarely offer negative critique.
  • Summary: Research indicates people use AI to record interviews and receive specific feedback on areas for improvement, which is often missing in real-world scenarios. Zevi also used Claude to play the role of the candidate to generate perfect answers for questions he lacked time to mock with humans. This approach allows learning directly from an ideal response.
Generational Shift in AI Use
Copied to clipboard!
(01:02:21)
  • Key Takeaway: The younger generation’s default is to immediately seek AI assistance for preparation and problem-solving, embodying the principle that one will be replaced by someone better at using AI.
  • Summary: The default behavior for those growing up with AI is to immediately turn to it for preparation or problem-solving, similar to how a child defaults to an iPhone gesture. This reinforces the idea that professional replacement will come from those more adept at leveraging AI tools, not necessarily from AI itself.
Failure Corner: Wix Onboarding
Copied to clipboard!
(01:03:19)
  • Key Takeaway: The expectation for junior PMs should be to be a ‘10x learner,’ not a ‘10x PM,’ leading to success by actively seeking mentorship based on individual strengths.
  • Summary: Zevi initially failed his first product review at Wix by trying to impress senior PMs alone instead of collaborating. He realized the expectation was for him to be a 10x learner, not a 10x performer. He then strategically used each senior PM as a mentor for their specific strength (product sense, methodology, systems thinking), which accelerated his learning and built team alignment.
Best Time to Be Junior
Copied to clipboard!
(01:06:25)
  • Key Takeaway: Despite market challenges, it is the best time to be a junior or learner because AI enables individuals to build startups or deliver high value quickly.
  • Summary: Contrary to reports of fewer junior roles, Zevi argues it is the best time to be a junior because AI lowers the barrier to building startups independently. Kind, hardworking, and communicative individuals can now provide more value to companies than many with decades of experience.
Recommended Books and Media
Copied to clipboard!
(01:07:54)
  • Key Takeaway: Key recommendations include The Fountainhead for fiction, Shoe Dog for business inspiration, and Mindset for psychological development.
  • Summary: Zevi recommends The Fountainhead (fiction), Shoe Dog (business), and Mindset by Carol Dweck (psychology), noting that the latter fundamentally shifted his perspective from a fixed to a growth mindset. He also highly recommends the TV show Severance and recently enjoyed The Pitt.
Favorite Products and Mottos
Copied to clipboard!
(01:09:21)
  • Key Takeaway: Zevi favors Cap and Supercut as well-crafted, open-source alternatives to Loom, and lives by the mottos ‘You can just do things’ and ‘Nobody knows what the fuck they’re doing.’
  • Summary: Zevi favors Cap, an open-source Loom alternative, for its attention to detail, and also likes Supercut. His mottos reflect both the current capability enabled by technology (‘You can just do things’) and a necessary humility about organizational complexity (‘Nobody knows what the fuck they’re doing’).
High School Entrepreneurship Story
Copied to clipboard!
(01:11:12)
  • Key Takeaway: Zevi successfully negotiated directly with an importer for thermal clothing in high school and used creative marketing, including a custom basketball chant, to drive sales.
  • Summary: In 10th grade, Zevi bypassed multiple layers in the supply chain for thermal clothing by negotiating directly with the importer, achieving 100% profit margins. He created a memorable marketing campaign by writing a thermal clothing chant with his phone number embedded, which became locally famous in Jerusalem.