How a Meta PM ships products without ever writing code | Zevi Arnovitz
These notes were generated by AI and may contain inaccuracies.
Product manager Zevi Arnovitz ships products without knowing how to write code or review code. Zero technical background, did music in high school. When Sonnet 3.5 came out, watched YouTube video building apps using Bolt or Lovable—felt like gaining superpowers. Now uses Cursor with Claude Code. For non-technical people, code is terrifying, but AI makes it possible. In coming years, everyone will become a builder; titles and responsibilities will collapse.
Main challenge: reviewing AI-written code. Uses /review to have Claude review its own code. Also opens Codex and Cursor to have each review the code. Quote: "It's not that you will be replaced by AI. You'll be replaced by someone who's better at using AI than you." Best time to be a junior—can build a startup right out of school.
Zevi is PM at Meta, previously at Wix. Non-technical, young, ambitious; uses Cursor and Claude Code to build significant real products alone. Created clever workflow everyone can copy. Download prompts and /commands from show notes top.
Workflow: work with Cursor to add ideas to Linear, explore with AI, develop plan, build, have LLMs review code, update documentation, use as learning opportunity.
Annual newsletter subscribers get 19 premium products free for a year: Lovable, Replit, Bolt, Gamma, n8n, Linear, Devin, PostHog, Superhuman, Descript, Wispr Flow, Perplexity, Warp, Granola, Magic Patterns, Raycast, ChatPRD, Mobbin, Stripe Atlas. lennysnewsletter.com/product-pass.
Sponsor: 10Web—AI website building before ChatGPT. Over 2 million websites generated. Vibe coding platform like Lovable for WordPress, front-end and back-end. Build any site: e-commerce, portfolios, blogs. Includes WordPress admin and thousands of plugins. API for SaaS, agencies, etc. 10web.io/lenny, code Lenny for credits and 30% off.
Sponsor: DX—developer intelligence platform. Helps answer: which tools work, usage, value. Used by Dropbox, Booking.com, Adyen, Intercom. getdx.com/lenny.
Zevi thanks Lenny, huge fan of show. Tal Raviv intro: "Zevi is the most hands-on vibe coding PM I know... His engineers at Meta ask him to teach them... everyone needs to be hearing this."
Very non-technical: music in high school, no tech unit in Israeli Army. Year ago, traveling in Asia with wife, in Japan when Sonnet 3.5 out. Watched YouTube (Greg Isenberg or Riley Brown) building apps with Bolt or Lovable—felt like superpowers. Got home, opened Bolt, been building for past year.
Claude prep: Goal is if people think "how amazing Zevi is," failed; if they open computer and start building, succeeded.
Started as GPT projects and Claude projects—shared folder of chats with custom instructions and knowledge base. GPT memory annoyed by mixing contexts (running, PM, student). Projects compartmentalize.
Bolt/Lovable eager to code (system prompt: coding agent). Fun initially, but later caused bugs (e.g., payments, database changes) due to poor planning. Created CTO project in ChatGPT: complete technical owner, owns how built, challenge user, not people-pleaser.
ChatGPT sycophantic: example, asked if Bun JavaScript like Zustand—said yes, then "thought you were making it up and was riffing."
CTO for architecture/technical decisions. Now uses full workflow in Cursor with Claude Code, no GPT. Recommend starting with GPT project for learning/exposure therapy, then Bolt/Lovable, Cursor light mode, terminal/dark mode.
Graduated tools when outgrew: Bolt great until payments, then Cursor + Claude. Code is just words/files—portable across apps/tools.
Cursor screen share: left—code files; right—Cursor AI; middle—Claude Code.
/commands: reusable prompts saved in codebase, invoke with /name.
- /create issue: capture bug/feature mid-development for Linear. Format: TLDR, current state, expected outcomes, context.
Workflow:
- Create issue (Linear).
- Exploration phase: explore issue, clarify, understand codebase.
- Create plan: markdown template.
- Execute plan.
- Review.
- Peer review.
- Update docs.
App: StudyMate—upload study materials (PDF), create interactive tests. Select pages, questions, difficulty. Sends to Gemini for quiz (challenging comprehension questions, hints). Currently multiple choice only. Wants true/false and fill-in-blank (30% fill-in-blank, competitor research).
Fill-in-blank: single sentence with 2 blanks, 6 potential answers (2 correct, 4 incorrect; 1 correct + 2 incorrect per spot?), drag-and-drop interface.
/commands from CTO project backbone, automated repeats.
Voice dictate with Wispr Flow: /create issue—mid-building, capture quick.
Idea: Add fill-in-blank (30% of tests), 6 answers for 2 blanks (2 correct), drag-drop interface.
Claude asks: quizzes 100% multiple choice? Structure: single sentence passage to blanks? Priority?
Uses MCP (Anthropic tool use) to create Linear issue STU-88: Fill in the blank questions with drag and drop interface. TLDR, current state (researched codebase), expected outcomes, context.
Feels like talking to engineer/engineering manager/CTO.
Days later, pick up: /exploration phase {Linear STU88} (tab for arg).
Fetches Linear ticket. Goal: deeply understand problem, current codebase state, files affected, best implementation. Claude reads files, asks clarifying questions.
"So it feels like it's talking to your engineering manager."
Uses ChatGPT prompt with CTO role, now integrated into Cursor as a single agent handling exploration, planning, and code execution. It's essentially Claude Code. System prompt in Claude.md defines workflow: challenge thinking in exploration phase.
For solo company, Linear issues generated by AI are high quality due to full context availability. Easy to spot misunderstandings. Suitable for side projects to kick off exploration, not fully ready to build. At larger companies like Meta, workflow would differ due to team coordination.
Claude analyzes StudyMate codebase, understands system, feature quest, key areas. Asks clarifying questions on scope, data model, UX/UI, validation, grading, AI system prompt changes. Prepares answers in advance.
"It's like such smart, sophisticated, important questions instead of just, 'Cool, here I go, I'm going to build it.'"
Difference between vibe coding and serious app building: extensive back-and-forth for understanding.
For difficult concepts: /learning opportunity primes Claude to explain using 80/20 rule for mid-level engineering knowledge (technical PM in making, understands architecture).
/create plan generates Markdown file from Twitter template: clear, minimal, concise steps; status trackers; TLDR; critical decisions; plan itself. Claude updates statuses as it progresses.
Split plans for different models: Cursor Composer (fast for non-complex); Gemini 3 (excellent for UI, frontend).
Tag plan file and say "execute" in Cursor with Composer for speed.
Lovable, Bolt, Replit, Base44, v0: opinionated, remove guesswork/hard decisions for ease, but less control. Models are the same (Claude underlying many), difference is the harness adding layers.
Cursor and Claude Code give full control, direct model access with tools.
"If you want the most cutting edge abilities of the models and you want to be able to make all the decisions on your own, it's probably best to be on one of these tools."
Base44 automates sign-in with Google, database, but no choices on database type or auth method.
Uses Gemini 3 for frontend, Claude for other tasks on regular MacBook. World changes weekly, no boundaries.
Time machine moments: Localized StudyMate Hebrew to English in 2 days (weeks for dev team); built personal site from no domain to live in 1.5 hours; three agents running in parallel, just wait.
"We live in the future."
"Best time to be alive... awesome time to be curious and optimistic and hardworking."
Run app locally post-Composer's minutes-long execution (days/weeks for human). Costs couple bucks in AI credits, views as tuition for learning.
Framer ad: Website builder for startups to Fortune 500s (DoorDash, Zapier, Perplexity, ElevenLabs). Real-time collaboration, CMS, SEO, A/B testing, live changes without engineering.
Manual QA first. Then /review for Claude to self-review. Uses multiple models:
- Codex (ChatGPT's Claude Code competitor).
- Cursor Composer.
/peer review command: Feed other models' reviews to Claude (dev lead). Claude explains or fixes, using full context.
Model Personalities
- Claude: Perfect CTO—communicative, smart, opinionated, collaborative.
- Codex (GPT model): Best coder in hoodie/sandals, dark room, fixes worst bugs silently.
- Gemini: Crazy scientist—artsy, talented at design, terrifying process (e.g., deletes dashboard then restores, considers DB edits unnecessarily), beautiful results.
Play to strengths, mitigate weaknesses via peer reviews. Models "fight it out"; Claude gets sassy on repeated non-issues.
Claude finds critical/high/medium bugs. Run reviews in Codex ("review all code in this branch"), Composer (/review). Copy-paste results into peer review as "dev lead one/two" until no issues.
Use /learning opportunity during reviews.
Constant postmortems: When bugs/mistakes occur, ask "What in your system prompt or tooling made you make this mistake?" Update Claude.md, documentation, tooling to prevent recurrence.
Even on success, reflect. Iterating prompts improves AI responses—key unlock distinguishing AI users.
"Updating documentation and tooling is one of the biggest hacks for productivity."
Final steps: code reviews, update docs, testing, user testing before release.
Ships full features as non-technical PM using AI models, afraid of code initially.
You're making money with this product. We're so used to this now, but it's insane what is now possible. It's the best time to be alive. AI just makes it so much possible. My brother, who I'm building one of the apps with, is an entrepreneur. He has a beautiful business that helps old people and seniors understand to use technology and AI better. He's replaced all the tools he was paying for, like Zapier and Airtable, and built a full-fledged CRM system and automation system for his business completely alone. For people who are curious, optimistic, hardworking, this is the best time to be a builder.
The biggest barrier for a lot of people is how do I get started? What exactly do I do? I open up Cursor. It looks very intimidating. I don't know how to write code. I don't know how to build stuff. I don't know about databases. You're going to be sharing all these /commands and basically this whole workflow with the audience. Just start at GPT. Start on GPT, tell it what your idea is, tell it to explain to you what are even the first steps of thinking, what are the decisions you need to make? And just be inquisitive, learn. Don't rush things. It's very important to just dive in and really spend the time to learn. One of your /commands is learning opportunity, and it's how you learn a lot of these things. Just teach me this thing and how this database issue works.
Coming back to a question about how this might work at a larger company, say a thousand person company, 500 people. How much of this can you plug and play into a workflow as a PM at a larger company? What would be your advice for someone that may want to start trying to ship code, at least showing people what's possible? First making your code base AI native is a really important step, and this needs to be done by technical people. My codebase has a ton of just plain text in it. It will have a bunch of markdown files that explain to agents how to work in certain areas of the code base and high level structure so that the agents navigate through the code base easier. If this is set up in a really good way, PMs still shouldn't be shipping heavy database chain migrations or any big project, but contained UI projects, especially if you just build it, create the PR and send it to a dev to do the final finishes. That's definitely something that's possible. We're going to see that a lot in the next coming years. Everyone's going to become a builder.
As a PM, maybe don't go right to Cursor, start building, shipping, trying to ship features to production, especially complicated features. Do you think we'll get there? Do you think in a couple years, PMs will be doing this and it'll feel less scary and crazy? Titles are going to collapse and responsibilities are going to collapse and everyone's just going to be building. The models, the context window is getting bigger, the models are getting smarter. PMs or anyone from any background can be writing. At the moment, I wouldn't wait for that. Use this as a collaborative learning opportunity to work with your dev team. It's going to be difficult. A lot of developers are very skeptical about the current state, and it's going to be a lot of sales work on your end to convince, but if you're able to convince, teams that are really sold on this and want to take the time to work on their workflow about how can our team become more AI native, these teams are going to be a few years in the future and they're going to look back at the few weeks they spent setting this up as the best time they spent.
One of the biggest fears people have with these AI tools for PMs is you start to rely on these things, your skills start to atrophy, you're producing slop that looks great, cool, amazing strategy doc. No, it's actually not at all good. Are these Linear tickets or just products that are half-baked? How has this impacted your craft as a PM? Do you feel like this is weakening your skills because you're so reliant on these tools and how do you keep the quality of this stuff up and not just AI generated slop?
I have a very strong disagree. I've heard it a bunch. When I started using Tal Raviv's course on building a PM Copilot using projects, which is probably one of the best courses that you can take, people at work said, "Oh, so you're basically outsourcing your thinking." That's just the worst way to look at it. There's a misconception with a lot of PMs that the job is always having the right answers and being the smartest person in the room. How I was trained and how I believe the role of the PM is, it's the exact opposite. It's harnessing anything that can get us as quick as possible to delivering the right solution to users. AI is like that really smart person that has context or your mentor but is always available and doesn't judge you and can really help you.
If you're using it to just create your outputs and then putting them out there, yeah, that's AI slop, but it's also human error. Own your own outputs. If you put anything out there or show something in a product review and you say, "Oh, sorry, that was built by AI." That's your mistake. If you use these intentionally and really take the time to understand how to use AI in the correct way, it's one of the biggest game changers that will make you much better as a PM.
Especially for more junior PMs, it allows you to play at such a higher level than you would normally. At Wix, I wasn't thinking of what's the marketing strategy of the company and how will the onboarding be completely revamped within the whole product. But on my side product, I can just do whatever decisions I want and think of the strategy and marketing and the messaging. This is basically just getting me reps, which is one of the most important things at the beginning of your career. The upside is so much more valuable. The only way that AI makes you worse at your job is if you're using it wrong.
Anything you've learned about reducing the sloppiness of the output, a tip for keeping the quality high? Similar to people, setting up AI for success for the task at hand. If I just brought in a junior to write a deck and didn't give it any guideline, I just said, "Give a strategy deck." He would probably just go online and find top strategy deck and reproduce that, which is basically what AI is doing. It's fed all of the internet. Instead, guide it and give it context on what your style of writing is and what you're trying to solve. That's one of the biggest unlocks.
Cursor has a /command called deslop, which is basically going back over the code. Their founders have been talking about this on Twitter, so run after that to make sure that no slop is left behind.
You used AI to help you actually interview for the job that you got at Meta. A lot of people are struggling to find a job, reading about people using AI to help them interview. What did you use? What worked? People who have grown up now in their professional lives think AI first how to solve it. Meta reached out and said they'd like me to interview. Straight away, I opened up a project within Claude. I started looking online for all the best information out there, things that resonated with me. I took a ton of frameworks from Ben Erez who has written a guest post, one of the best minds out there right now. I created a project which was my coach, which I would come and consult what to do at each phase. I would mock interview with it, and this was amazing.
I created a game in Base44, which helped me. I was really struggling with segmentation within the product questions, so thinking of the correct segments. I created a quiz game which creates questions and different segmentations and I have to choose. This was a web app that I would play sometimes when I was on the bus to work.
Creating a project and feeding it with all the best information on the internet and then mocking a bunch. The biggest game changer was doing human mocks. Cold outreaching to people on LinkedIn and having them do actual mocks. For Meta PM prep, which is super competitive and difficult, there's no way to get around that.
Ben shares prompts you can feed ChatGPT to help you prepare for interviews, do mocks online. Those take you to a point, but it's actually better to use humans. People use AI to get feedback. They record the interview and it gives them feedback. Here's where you could have done better. Here's what you missed because the feedback loop is so missing. No one ever tells you what you did badly in this interview, and AI can do that.
I mocked with AI. There's a question bank online free by Louis Lynn, an always updating bank of questions that people are asked in real interviews. I used Comet, which is Perplexity's browser. I had the agent run all kinds of analyses on what the most asked questions are. That's how I knew how to prioritize what questions I would mock. At the end of these mocks, I would tell Claude within the project, "You're my coach and I don't want you to make me feel good. I want you to make me as ready as possible for these interviews. So give me feedback." For some questions where I didn't have time to mock, I would ask Claude to play the candidate, and then it would just give me a really good answer. I could learn from that, learning from someone who does a perfect answer.
People in this generation default is, "I have something I need to do. Let's go to AI immediately and help me prepare for this thing."
It's not that you will be replaced by AI at least for a long time. It's you'll be replaced by someone who's better at using AI than you.
Failure Corner: what's the story of a time you failed in your career and what did you learn? When I started at Wix as a student program, straight out of the student program, you get put into a certain team. I was in the editor, which is the core product of Wix. The other PMs were the best PMs almost at Wix. There were four other people with much more experience than me and they were ridiculously good. My first product review, I thought I'm going to blow these people's socks off. I'm going to show how good of a PM I am. I didn't really share what I was thinking. I would work tons of hours alone. I ended up failing miserably. My product review was not good. It wasn't the format they expected. They had a ton of questions that I missed and I felt awful.
Everyone was like, "All right, cool. Yeah, just come back in two weeks and we'll keep getting at this." I understood that they had zero expectation of me being a 10X PM, but the expectation was being a 10X learner. The second I understood that, my whole mindset shifted. The best tip I give now to junior PMs is be the best learner you can be at the beginning. No one expects you to know all the answers and no one expects you to be good.
I took each person on the PM team, there were four other PMs, and assessed what their strength is and used them as a mentor for that. Neri, still my mentor till today, has the best product sense of anyone I've met. Oya is a methodology expert. She just thinks in frameworks. Yahra, who is the head of product, can look at a product and instantly understand the third and fourth order effects, the system thinking. Every time I had an issue with one of these areas, I would come to one of them and consult them. This does two things. First, I learned a ton. Second, when the next product review came, my success felt to them like their success because it wasn't this kid trying to show us up. It was our mentee making us all proud. I really excelled through this.
To tie back, if people walk away thinking, "Zevi's so cool." Then I've failed. It's the best time to be alive. It's the best time to be a junior contrary to what a lot of people are saying how there's no more junior roles out there. When else in history could you get out of school and just build a startup on your own with a couple of friends completely bootstrapped. I saw more and more people building their own stuff with AI. It's the best time to be a junior. It's the best time to be a learner. If you're a curious person, you're a hardworking person, if you're a kind person and a good communicator, you have such an unfair advantage and you can give more value to companies than most people who have 20 years of experience.
Lightning round: two or three books that you find yourself recommending most? Fiction: The Fountainhead by Ayn Rand, really makes you think, really makes you feel. Business books: Shoe Dog, the Nike story. Psychology side: Mindset by Carol Dweck, who coined the term growth mindset. It sounds like a self-help book, but it's completely psychological and based on research. That book completely changed my life. I was always with a fixed mindset, and after reading that, I understood that it was something holding me back. Since then, I've been really trying to cultivate a growth mindset.
Favorite recent movie or TV show? The Pitt was amazing. First recommendation: if you haven't seen Severance, run to see it, one of my favorite shows.
Favorite product recently discovered? A Loom alternative called Cap, open source, really well-crafted. You can see the person was sweating the details. Also Supercut, another Loom alternative.
Favorite life motto? "You can just do things." And the second one I stole from my brother: "Nobody knows what the fuck they're doing." It makes you take life more lightly. People see companies on the outside and think they've got all figured out. On the inside, you're like, how is this staying on the rails?
You've had a long entrepreneurial thread. You did a thermal clothing business and a hummus delivery thing. Pick one. Thermal clothing: in high school, 10th grade, I was selling thermal clothes, packs of thermal clothing, shirt and pants. I grew up in Jerusalem, so it's a bit chillier there. They were $20, $25 a piece and I was making like $4 a sale, sixth or seventh down the line, crazy margins. During the summer I thought I should go straight to the importer. I called the importer and at first he was really mad. "No, you have to work for me for years." I said, listen, I'm finishing school soon. This is not going to be my career. Either do it or not. We negotiated throughout the whole summer. He would throw out something, "Oh, the import tax has gone up." I'd search Google, Import Tax Israel, and stall. I ended up getting $12.50 a piece, 100% profit. I spread throughout a bunch of different schools. Each school, I had the coolest people in school selling for me.
Our basketball team would be 30 points up within the first half, got boring for the crowd. I wrote a basketball chant about Thermal Clothes that has my number in it. The end was if you join in now, we'll give you a discount. It was with drums and everything. Still when I go to Jerusalem, people know my number by the tune. Sometimes people stop me and say, "Hey, it's Thermal Zevi."
Where can folks find you? Reach out on LinkedIn or on X. I'd really love to help whoever I can. How can listeners be useful to you? I've been helped throughout my whole career a ton, so I love helping any way I can.
Zevi discusses how listeners can be useful to him by trying his products and providing feedback.
- If you're a student, try StudyMate and tell me what you think.
- If you're in Israel and not using dictation yet, try Dibur2text. Tell me what you think.
The host praises Zevi: > Amazing. I just love how much you're giving away and how useful that's going to be to so many people.
The host notes that links to these products will be in the show notes and thanks Zevi: > Zevi, you're awesome. Thank you so much for being here. Thank you so much for sharing so much. This is going to help, I think, a lot of people and I think it's going to help people get over the hump on. Okay, I see all these people doing cool stuff. Here's how I can actually do this stuff.
Host repeats thanks: > Thank you so much for being here and for sharing so much.
Zevi responds: > Thank you for having me. And if you build something cool with some stuff that I learned here, hit me up, send me. I'd love to see.
Host concludes: > Amazing. Zevi, thank you so much for being here.
Zevi: > Thank you. Bye everyone.
Thank you so much for listening. If you found this valuable, you can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app. Also, please consider giving us a rating or leaving a review as that really helps other listeners find the podcast. You can find all past episodes or learn more about the show at lennyspodcast.com. See you in the next episode.
Keep Lenny's Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
