Jan 18, 2026 · 1h 15m · lennys-podcast
How a Meta PM ships products without ever writing code | Zevi Arnovitz
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Meta Product Manager Zevi Arnovitz joins Lenny Rachitsky to break down his end-to-end workflow for building and shipping production software without writing code. Through live demonstrations of tools like Cursor, Claude Code, and multi-model peer reviews, Zevi explains how non-technical builders can architect robust applications and accelerate their product careers.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Lenny holds 28.1% of the talking time here. How this is scored →
Lenny as informed peer 3.9 Guest teaching 5.6 Guest disagreement 0.8 Lenny pushing back 1.0
speaking balance: gold is Lenny, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 54:25 Direct pushback against outsourcing thinking
Zevi forcefully rejects the critique that PMs using AI are merely outsourcing thinking or producing slop, calling it the worst perspective on product craftsmanship.
Hardest push from Lenny ▶ 53:43 Questioning craft decay and AI slopLenny directly confronts Zevi with the common fear that relying on AI tools causes PMs' analytical skills to atrophy and results in low-quality tickets and strategies.
Biggest teaching moment ▶ 40:20 Multi-model peer review architectureZevi educates Lenny on setting up an automated peer review pipeline where competing models challenge each other's code reviews rather than relying on human QA.
Lenny holds their own ▶ 1:00:38 Lenny on AI interview feedback loopsLenny injects his own research with Noam Segal to explain how post-interview transcript evaluation fills a systemic feedback void in tech hiring.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Lenny as informed peer | Guest teaching | Guest disagreement | Lenny pushing back | Why |
|---|---|---|---|---|---|---|
| Previewing the Episode and Vibe Coding Highlights | 4 | 3 | 1 | 0 | Lenny provides an enthusiastic introductory overview framing Zevi's vibe-coding methodology and setting expectations for non-technical product builders. | |
| Sponsor Segment: TenWeb and DX | 3 | 2 | 0 | 0 | Includes initial sponsor reads followed by a warm introduction where Lenny cites mutual praise from Tal Raviv and Zevi recounts his origin story in Japan. | |
| Designing a Virtual CTO and Overcoming AI Sycophancy | 4 | 6 | 1 | 1 | Zevi explains how he configured ChatGPT to act as a non-sycophantic virtual CTO and recommends gradual exposure therapy for non-technical PMs moving to code editors. | |
| Live Demonstration: Issue Creation via Linear and MCP | 3 | 7 | 0 | 1 | Zevi walks through his live Cursor environment, demonstrating custom slash commands, Linear integration via Anthropic MCP, and the architecture of his app StudyMate. | |
| Technical Exploration and Structured Implementation Planning | 4 | 7 | 1 | 2 | Lenny asks how well AI tickets hold up in practice, prompting Zevi to detail the exploration and markdown planning phases that precede actual execution. | |
| Fast Plan Execution and Tooling Trade-Offs | 5 | 6 | 2 | 2 | Lenny probes why platforms like Lovable or Bolt were insufficient, leading Zevi to compare opinionated harnesses with raw model orchestration and parallel time-machine workflows. | |
| Sponsor Segment: Framer Website Builder | 3 | 8 | 1 | 1 | Zevi delivers an insightful breakdown of multi-LLM peer reviews, personifying Claude as an articulate CTO, Codex as a silent specialist, and Gemini as an artistic mad scientist. | |
| Prompt Post-Mortems and Documentation Upgrades | 4 | 6 | 0 | 1 | Zevi explains the importance of conducting post-mortems on AI failures by prompting the model to diagnose prompt gaps and updating persistent system documentation. | |
| Scaling AI in Enterprises and Preserving PM Craft | 5 | 7 | 4 | 3 | Lenny challenges whether AI makes PM skills atrophy into slop; Zevi strongly disagrees, arguing that AI raises the ceiling and that skill degradation only stems from poor usage. | |
| Leveraging AI Models for High-Stakes Job Interviews | 5 | 6 | 0 | 1 | Zevi details how he automated interview prep for Meta using custom Claude projects and question-bank scrapers, while Lenny relates his own findings on interview feedback loops. | |
| Failure Corner: Shifting from 10x PM to 10x Learner at Wix | 4 | 5 | 0 | 0 | In Failure Corner, Zevi recounts bombing his first Wix product review and discovering that being a 10x learner and leveraging peer mentors outperformed trying to look like a 10x PM. | |
| The Golden Age of Junior Builders and Lightning Round | 3 | 4 | 0 | 0 | Zevi encourages junior builders to embrace AI leverage, completes the lightning round, and tells a story about running a high school thermal clothing venture. |
Statements from this episode (15)
Arnovitz: AI coding agents jumping straight to code causes severe bugs
“Planning is really important when you're implementing something technical and let's say you're implementing payments or something that's going to be a change to your database. If the coding agent is just like, all right, I got it and just starts writing code. …”
Arnovitz: Default ChatGPT makes the worst CTO due to sycophancy
“I think chat GPT would probably be. The worst CTO because it's such a people pleaser and it's so sycophantic”
Arnovitz: Non-technical builders should progress gradually from ChatGPT to Cursor
“I think if you see this where I'm working like in Claude or in cursor, you might be excited to start using those, but I would really recommend starting slow with a GPT project. Beautiful UI, super simple, then maybe graduate to like a bolt or a lovable. And th…”
Arnovitz: Updating Documentation Enables AI Coding Agents to Write Better Code
“And at the end we update the docs. So this is updating documentation and everything so that agents can write better code later on.”
Arnovitz: Serious AI development requires deep exploration unlike superficial vibe coding
“And I think this is the big difference between like just vibe coding and going along with the vibes and really building serious apps. I spend a lot of time going back and forth and understanding.”
Arnovitz: Gemini 3 is unbelievable at frontend and UI development
“Gemini three that just came out is unbelievable at UI. So a lot of times I'll split the plan into backend and front end, and then I'll have Gemini just read the plan and do the front end.”
Arnovitz: AI coding tools differ by harness, not underlying model
“The main difference between all these tools is basically the harness. So the models are all the same models. You know, I I'll run Claude within cursor. I'll run it within Claude code. And it's also the models that Claude is also the model that is underlying Bo…”
Arnovitz: Localized StudyMate in two days using parallel AI agents
“I was fully localizing study mate from Hebrew to English, which I did in two days, which would probably take a dev team weeks. And I was building a personal site, which went from no domain, no nothing to live on a domain within an hour and a half. And I was do…”
Rachitsky: The main challenge with AI coding is code review, not writing
“Cause this is one of the things that comes up a lot in this podcast is writing code is now so easy. The main challenge people have is reviewing the code that AI has written.”
Arnovitz: Multi-model peer reviews mitigate individual AI coding weaknesses
“So I think that using all these models and basically playing to their strengths and mitigating their weaknesses by using other models is, is a game changer for me. So I'll do peer review a bunch of times and I'll have other models review other models code and …”
Arnovitz: Prompt post-mortems separate mediocre AI users from proficient builders
“Going back to your prompts, understanding what was not good enough, iterating on them, and then seeing how AI's responses get better. I think that's probably one of the most important things. And one of the things that divides between people who are like okay …”
Arnovitz predicts job titles will collapse as AI makes everyone a builder
“I think titles are going to collapse and responsibilities are going to collapse and everyone's just going to be building.”
Arnovitz: Human mock interviews remain indispensable for Meta PM prep
“The biggest game changer for me was doing human mocks. So cold outreaching to people on LinkedIn and having them do actual mocks for me. I think that at the end of the day, especially for the meta PM prep, which is super competitive and difficult. I think ther…”
Arnovitz: Communicative junior builders using AI can outvalue 20-year veterans
“If any listener is listening to this and you're a curious person, you're a hardworking person. I wanna say kind, I'm not sure, but 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 …”
Arnovitz: Open-Source Tool Cap Is a Well-Crafted Alternative to Loom
“I was kind of disappointed with loom. They were taking so much money and the product I dunno, I just didn't love it. And there's an open source alternative called cap, which is just really well crafted. You can see that the person was like really sweating the …”