By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Product SchoolSnapchat SVP of Engineering on How Everyone Ships Code Without Breaking Quality | Saral Jain | E304
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
How Snap uses AI agents to ship faster without losing quality
- Saral Jain describes Snap’s shift toward cross-functional building, where designers, product managers, and engineers can all ship code as role boundaries blur.
- Snap organizes “startup squads” as lean, cross-functional teams aimed at zero-to-one initiatives, with faster iteration and less rigid role separation.
- To protect quality at Snap’s scale, Saral says AI agents like Code Pal handle the first pass of code review quickly, while humans remain accountable for the final code shipped.
- Saral explains that company-wide AI adoption came quickly, but business impact required anchoring AI work to a Jobs To Be Done framework and roadmaps to avoid throwaway prototypes.
- The episode outlines Snap’s AI strategy across three areas: user experiences (e.g., My AI and Gen AI Lenses), internal productivity (agents like Casper), and advertiser ROI (targeting, measurement, creatives, and AI Sponsored Snaps).
IDEAS WORTH REMEMBERING
5 ideasSnap is embracing cross-role code shipping because AI and platforms blur traditional boundaries.
Saral Jain said the boundaries between designers, product managers, and engineers are becoming less clear, so the focus shifts from job titles to what people actually do day to day. He described Snap as having designers and product managers shipping production code, enabled by underlying platform investments meant to limit quality and reliability risk.
Lean “startup squads” are Snap’s mechanism for zero-to-one bets inside a large company.
Saral described “startup squads” as lean, cross-functional teams (engineering, design, product, data science) that operate like startups inside a mature platform. He said role distinctions are less clear in these squads and noted Snap invested in nine squads, with most generating strong results quickly.
Snap uses an AI-first review pipeline, but retains human ownership for accountability.
Saral said quality is non-negotiable when shipping to nearly a billion people, so Snap relies on foundational platform work rather than relying solely on process. He highlighted Code Pal as an AI code review agent that performs the first pass and reviews over 90% of code within five minutes, while humans still remain accountable for what ships.
The key internal AI challenge wasn’t adoption—it was converting usage into measurable business impact.
Saral explained Snap initially saw broad AI adoption (e.g., enterprise AI tools) but struggled to convert it into business impact because many prototypes were built and discarded. He said leadership shifted to a Jobs To Be Done approach: define core jobs per function, align roadmaps to those jobs, and then apply AI to roadmap work rather than “throwaway” experiments.
Snap is building an internal agent ecosystem: broad agents for building, specialized agents for specific workflows.
Saral distinguished “horizontal” foundational agents (usable by anyone) from “vertical” function-specific agents. He described Casper as a horizontal “AI teammate” embedded in tools like Slack and Jira that can be invoked mid-conversation to build working prototypes/code, with Code Pal then reviewing the output in an agent-to-agent loop.
WORDS WORTH SAVING
5 quotesI have actually gotten very comfortable with that idea because it is a reality in the industry right now that the boundaries between designers, product managers, and engineers are actually changing in front of our eyes right now.
— Saral Jain
It reviews more than 90% of the code within five minutes of the code being published essentially, and it does a very, very good pass at identifying all of the major issues that can happen.
— Saral Jain
Adoption wasn't the problem. Translating that adoption to business impact was the problem.
— Saral Jain
Casper is listening into that conversation. And at one point, you can just invoke Casper saying, "Okay, you have all the context. You know everything of the history of the company essentially, and all of our knowledge base and, uh, literally all of our tens of millions of lines of code across all repos. Why don't you go ahead and build it?"
— Saral Jain
The bottleneck is not Figma mockups. The bottleneck is we just build the prototype and then show the prototype and the best idea wins.
— Saral Jain
SPEAKERS
Saral Jain
guestSenior Vice President of Engineering at Snap/Snapchat, leading major parts of Snapchat engineering.
Carlos González de Villaumbrosia
hostCEO at Product School and host of The Product Podcast.
High quality AI-generated summary created from speaker-labeled transcript.
