Building AI product sense, part 2

Original: Lenny Rachitsky · 10/02/2026

Summary

Why Meta added “Product Sense with AI” to its PM interview loop

Key Insights

“Why Meta added “Product Sense with AI” to its PM interview loop” — Highlighting the importance of AI product sense in product management roles at Meta.

“What minimum viable quality (MVQ) means and how to define three critical thresholds” — Explaining the concept of MVQ and its significance in AI product development.

“How to design guardrails that protect users from model shortcomings” — Discussing strategies for safeguarding users against potential failures of AI models.

Topics

Full Article

[![](https://substackcdn.com/image/fetch/$s_!hXEy!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fabf081d7-a01f-4c69-abf8-20cd69125385_3016x3016.png)](https://substackcdn.com/image/fetch/$s_!hXEy!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fabf081d7-a01f-4c69-abf8-20cd69125385_3016x3016.png)

If you’re a premium subscriber Add the private feed to your podcast app at add.lennysreads.com Dr. Marily Nika, longtime AI PM at Google and Meta, shares a simple weekly ritual that rapidly builds AI product sense – the ability to translate probabilistic model behavior into products people can trust. In this episode, Marily walks through the framework for uncovering failure modes before users do.

[Subscribe now](https://www.lennysnewsletter.com/subscribe?)

Listen now: YouTube | Apple | Spotify In this episode, you’ll learn:

* Why Meta added “Product Sense with AI” to its PM interview loop
* The rituals that surface hidden failure modes
* Why generative models confidently invent structure when confronted with mess
* What minimum viable quality (MVQ) means and how to define three critical thresholds
* Five strategic context factors that raise or lower your quality bar
* Why you need to estimate your AI feature’s cost envelope early
* How to design guardrails that protect users from model shortcomings
* Four patterns that cover most real-world failure cases

Referenced:

* [AI PM B…](https://maven.com/marily-nika/ai-pm-bootcamp)

[Read more](https://www.lennysnewsletter.com/p/building-ai-product-sense-part-2-0aa)

How to do AI analysis you can actually trust

Lenny Rachitsky · how-to · 77% similar

Lenny Rachitsky · tutorial · 75% similar

The rise of the professional vibe coder (a new AI-era job) | Lazar Jovanovic (Professional Vibe Coder)

Lenny Rachitsky · explanation · 74% similar