Lenny's Podcast: Product | Career | Growth
--% time saved
PodcastLenny's Podcast: Product | Career | Growth
Publisher/creatorLenny Rachitsky
Shortcast updated
About this episode
Zevi Arnovitz is a product manager at Meta with no technical background who has figured out how to build and ship real products using AI. His engineering team at Meta asks him to teach them how he does what he does. In this episode, Zevi breaks down his complete AI workflow that allows non-technical people to build sophisticated products with Cursor. We discuss: 1. The complete AI workflow that lets non-technical people build real products in Cursor 2. How to use multiple AI models for different tasks (Claude for planning, Gemini for UI) 3. Using slash commands to automate prompts 4. Zevi’s “peer review” technique, which uses different AI models to review each other’s code 5. Why this might be the best time to be a junior in tech, despite the challenging job market 6. How Zevi used AI to prepare for his Meta PM interviews — Brought to you by: 10Web —Vibe coding platform as an API DX —The developer intelligence platform designed by leading researchers Framer —Build better websites faster — Episode transcript: https://www.lennysnewsletter.com/p/the-non-technical-pms-guide-to-building-with-cursor — Archive of all Lenny's Podcast transcripts: https://www.dropbox.com/scl/fo/yxi4s2w998p1gvtpu4193/AMdNPR8AOw0lMklwtnC0TrQ?rlkey=j06x0nipoti519e0xgm23zsn9&st=ahz0fj11&dl=0 — Where to find Zevi Arnovitz • X: https://x.com/ArnovitzZevi • LinkedIn: https://www.linkedin.com/in/zev-arnovitz • Website: https://zeviarnovitz.com — Where to find Lenny: • Newsletter: https://www.lennysnewsletter.com • X: https://twitter.com/lennysan • LinkedIn: https://www.linkedin.com/in/lennyrachitsky/ — In this episode, we cover: (00:00) Introduction to Zevi Arnovitz (04:48) Zevi’s background and journey into AI (07:41) Overview of Zevi’s AI workflow (14:41) Screenshare: Exploring Zevi’s workflow in detail (17:18) Building a feature live: StudyMate app (30:52) Executing the plan with Cursor (38:32) Using multiple AI models for code review (40:40) Personifying AI models (43:37) Peer review process (45:40) The importance of postmortems (51:05) Integrating AI in large companies (53:42) How AI has impacted the PM role (57:02) How to improve AI outputs (58:15) AI-assisted job interviews (01:02:57) Failure corner (01:06:20) Lightning round and final thoughts — Referenced: • Becoming a super IC: Lessons from 12 years as a PM individual contributor | Tal Raviv (Product Lead at Riverside): https://www.lennysnewsletter.com/p/the-super-ic-pm-tal-raviv • Wix: https://www.wix.com • Building AI Apps: From Idea to Viral in 30 Days: https://www.youtube.com/watch?v=j2w4y7pDi8w • Riley Brown on YouTube: https://www.youtube.com/channel/UCMcoud_ZW7cfxeIugBflSBw • Greg Isenberg on YouTube: https://www.youtube.com/@GregIsenberg • Bolt: https://bolt.new • Inside Bolt: From near-death to ~$40m ARR in 5 months—one of the fastest-growing products in history | Eric Simons (founder and CEO of StackBlitz): https://www.lennysnewsletter.com/p/inside-bolt-eric-simons • Lovable: https://lovable.dev • Building Lovable: $10M ARR in 60 days with 15 people | Anton Osika (co-founder and CEO): https://www.lennysnewsletter.com/p/building-lovable-anton-osika • StudyMate: https://studymate.live • Dibur2text: https://dibur2text.app • Claude: https://claude.ai • Everyone should be using Claude Code more: https://www.lennysnewsletter.com/p/everyone-should-be-using-claude-code • Bun: https://bun.com • Zustand: https://zustand.docs.pmnd.rs/getting-started/introduction • Cursor: https://cursor.com • The rise of Cursor: The $300M ARR AI tool that engineers can’t stop using | Michael Truell (co-founder and CEO): https://www.lennysnewsletter.com/p/the-rise-of-cursor-michael-truell • Wispr Flow: https://wisprflow.ai • Linear: https://linear.app • Linear’s secret to building beloved B2B products | Nan Yu (Head of Product): https://www.lennysnewsletter.com/p/linears-secret-to-building-beloved-b2b-products-nan-yu • Cursor Composer: https://cursor.com/blog/composer • Replit: https://replit.com • Behind the product: Replit | Amjad Masad (co-founder and CEO): https://www.lennysnewsletter.com/p/behind-the-product-replit-amjad-masad • Base44: https://base44.com • Solo founder, $80M exit, 6 months: The Base44 bootstrapped startup success story | Maor Shlomo: https://www.lennysnewsletter.com/p/the-base44-bootstrapped-startup-success-story-maor-shlomo • v0: https://v0.app • Everyone’s an engineer now: Inside v0’s mission to create a hundred million builders | Guillermo Rauch (founder & CEO of Vercel, creators of v0 and Next.js): https://www.lennysnewsletter.com/p/everyones-an-engineer-now-guillermo-rauch • Cursor Browser mode: https://cursor.com/docs/agent/browser • Google Antigravity: https://antigravity.google • Grok: https://grok.com • Zapier: https://zapier.com • Airtable: https://www.airtable.com • Build Your Personal PM Productivity System & AI Copilot: https://maven.com/tal-raviv/product-manager-productivity-system • The definitive guide to mastering analytical thinking interviews: https://www.lennysnewsletter.com/p/the-definitive-guide-to-mastering-f81 • AI tools are overdelivering: results from our large-scale AI productivity survey: https://www.lennysnewsletter.com/p/ai-tools-are-overdelivering-results-c08 • Yaara Asaf on LinkedIn: https://www.linkedin.com/in/yaarasaf • The Pitt on Prime Video: https://www.amazon.com/The-Pitt-Season-1/dp/B0DNRR8QWD • Severance on AppleTV+: https://tv.apple.com/us/show/severance/umc.cmc.1srk2goyh2q2zdxcx605w8vtx • Loom: https://www.loom.com • Cap: https://cap.so • Supercut: https://supercut.ai ...References continued at: https://www.lennysnewsletter.com/p/the-non-technical-pms-guide-to-building-with-cursor — Recommended books: • The Fountainhead : https://www.amazon.com/Fountainhead-Ayn-Rand/dp/0451191153 • Shoe Dog: A Memoir by the Creator of Nike : https://www.amazon.com/Shoe-Dog-Memoir-Creator-Nike/dp/1501135910 • Mindset: The New Psychology of Success : https://www.amazon.com/Mindset-Psychology-Carol-S-Dweck/dp/0345472322 — Production and marketing by https://penname.co/ . For inquiries about sponsoring the podcast, email [email protected] . — Lenny may be an investor in the companies discussed. To hear more, visit www.lennysnewsletter.com
Loading episode data...
Episode summary
Shipping product without writing code? That’s where we’re headed today. My guest is Zevi Arnawitz, a PM at Meta who built real apps solo by orchestrating AI tools, and he’s sharing the exact prompts and slash commands so non‑technical PMs can do it too. Welcome, Zevi.
I have no engineering background. A year ago I saw people spinning up apps with AI and it felt like someone handed me superpowers, so I ran home and started building—and my goal today is that listeners leave and start building themselves.
Give us the high‑level playbook. What tools, and how do you run your workflow day to day?
I began with Projects to keep contexts separate and created a strict “CTO” agent to challenge my thinking and force planning before code. Treating models like teammates helped me avoid people‑pleasing answers and premature implementation.
Start gently: a project in a chat UI, then graduate to builder tools, and only later step into an IDE. That staged exposure made code less scary.
You now build in Cursor with Claude Code. Why move beyond Bolt or Lovable once you got comfortable?
I outgrew the guardrails when things like payments and data changes showed up. Code is just text, so I carry the same repo across tools, and I mix models—Claude for reasoning, Composer for speed, Gemini for UI—depending on the task.
Let’s walk through your slash‑command system, and we’ll link the prompts so people can use them right away.
My flow lives in slash commands saved in the repo: create issue, exploration, create plan, execute, review, peer review, then update docs. I demoed this on my side app, StudyMate, and specced a new fill‑in‑the‑blank quiz type by dictating the request and capturing it to Linear in seconds.
That voice spec felt like talking to an engineer. Then Claude pulls the ticket, asks smart clarifying questions, and we move to a build plan, right?
Exactly. Exploration lets the agent read the codebase, map affected files, and pressure‑test scope, data model, UX, validation, and grading. When it clicks, I run create plan to generate a markdown plan with status checkboxes that any model can follow; I often have Composer handle back end work and Gemini sketch the UI.
Then execute runs the plan. It writes changes, I boot the app locally, and we iterate until the feature behaves as intended.
You’ve said the real challenge isn’t writing code anymore—it’s reviewing it. How do you handle quality?
I do quick manual QA, then run slash review so Claude audits its own work. I also ask GPT‑based Codex and Cursor’s Composer to review the branch, because each model catches different issues.
Peer review is my favorite move: I paste each model’s notes into a prompt that treats Claude like the dev lead. It must either refute others with context or fix the findings. I lean into their personalities—Claude is a collaborative lead, GPT is the silent bug crusher, Gemini is a bold designer—and let them “argue” until we converge.
Once you’ve closed the gaps, how do you prevent repeat mistakes and reduce slop next time?
Post‑mortems are non‑negotiable. I ask the agent why it stumbled, then we update prompts, docs, and tooling so that class of bug doesn’t recur; a de‑slop pass helps clean lingering messiness.
For folks feeling intimidated, what would you tell them about getting started and building confidence?
This is a golden moment. My brother replaced a stack of no‑code tools by building a custom CRM with AI, and you can do the same—begin with a chat project, ask it to map decisions, learn as you go, and step up tools when you’re ready.
How would you adapt this inside a larger company without stepping on landmines?
Make the codebase AI‑native with clear markdown docs and patterns so agents can operate safely. PMs shouldn’t ship risky data migrations, but they can prototype contained UI, open a PR, and partner with devs for the final mile; teams that invest early in workflows will feel years ahead.
Does relying on AI dull PM skills or flood teams with low‑quality output?
Only if you abdicate ownership. The job isn’t to be the lone genius—it’s to marshal everything that gets customers the right outcomes. Used well, AI gives juniors more reps across strategy, UX, and execution; used sloppily, it shows.
Quick quality tip for better outputs?
Set the stage like you would for a new teammate: goals, style, constraints, and examples; then run a clean‑up pass to strip generic cruft.
You also used AI to land your Meta role. What worked in your prep?
I spun up a Claude project as my coach, fed it frameworks I liked, and did relentless mocks; I even built a segmentation quiz game to drill weak spots, mined question banks with Comet, and had the model critique my answers and show exemplar responses.
The biggest unlock was still human mocks, but AI got me sharper, faster, and more focused on the right questions.
Let’s do Failure Corner. Share a time you stumbled and what changed.
Early at Wix, I tried to dazzle in a product review by working alone and bombed. I learned the expectation wasn’t being a ten‑times genius—it was being a ten‑times learner—so I sought mentors for their superpowers and our results improved together.
That learning mindset runs through everything you’ve shared today.
Final thought: it’s an amazing time to be junior and curious. If you’re kind, communicative, and willing to grind, you can create outsized value—go build something and let it teach you.
Lightning round highlights: favorite reads include The Fountainhead, Shoe Dog, and Mindset. Must‑watch TV is Severance, and a recent indie pick. A product I’m eyeing from Zevi’s list is Kap as a lightweight Loom alternative, and two mottos stick—“you can just do things” and “nobody knows what they’re doing.”
I once ran a thermal‑clothing hustle in school with a chant that included my phone number; years later people still remember it. If you want to connect, reach out on LinkedIn or X, and please send me what you build.
Zevi, this was fantastic. We’ll link the prompts in the show notes—can’t wait to see what listeners create. Thanks for tuning in; subscribe on your favorite app, and we’ll see you next time.
Ready to save time?
Download Shortcast and get started today