Adam Mosseri: AI is a tailwind for authenticity

Taste matters a ton; in a world where building is easy, the scarce skill is choosing what to build and knowing where AI helps and where it does not.

Today’s guest is Adam Mosseri, head of Instagram—used by over three billion people monthly—previously behind Facebook’s News Feed and ranking, now the public face of Instagram’s biggest shifts and debates; Adam, welcome.

Our product teams moved from big, specialist-heavy groups to small pods of six or seven, anchored by generalist engineers and a product-staff role that blends PM, design, research, and data, with senior specialists pulled in only when the work truly needs them; fewer handoffs, faster decisions.

Tools now automate a lot of mechanical data science, like funnel and drop-off analysis, so a product-staff can self-serve insights that used to require bespoke help.

Designers with real taste will thrive; many are crossing into product-staff because functional lines are blurring, while standout experts in areas like ML still matter deeply.

There’s anxiety everywhere, but functions won’t vanish—they’ll skew smaller, more senior, and more fluid; keep growing junior talent into future seniors, and avoid overconfidence in any forecast right now.

Across hiring, the constants still win: drive, fast learning, and self-awareness; what’s rising is curiosity and a willingness to ship, get corrected, and try again, while giant-org operators will be less in demand as teams shrink.

AI is reshaping who shines; coding is now steering models, and cross-functional walls are falling—designers code, engineers analyze, data scientists mock up flows—so people who know where AI is strong or weak, and can see near-term trajectories, will get outsized leverage.

On model costs, we killed obvious token bonfires and watch ROI; no hard caps yet, but I can imagine per-team or per-person budgets if compute burn rivals salary, even as competition should push prices down over time.

As AI eats more of the build cycle, human advantage concentrates in taste, judgment, and strategy—setting goals and constraints, calibrating autonomy, and giving crisp feedback without smothering creativity.

AI can aid strategy if you tightly bound the brief—technology realities, team talent, competitive and regulatory context, and brand—then have a back-and-forth with a model that pushes back.

Keep the human-on-top ‘centaur’ pattern; if AI dictates strategy and people just execute, motivation and quality will crater.

The best product leaders are curators of people, ideas, and paths; I care less where a strategy originates and more that it’s sharp, owned, and executed, with a leadership team whose skills and chemistry truly complement each other.

A common myth about Instagram’s ranking is that it semantically knows you in detail; most gains came from embeddings—opaque vectors—not tidy labels, though LLMs now help describe those regions and give people more agency.

We launched a ‘your algorithm’ view that surfaces inferred interests from your interaction map and lets you adjust them, with a future where you can steer for things like more friends or more fun, not just topics.

Pure chronological feeds sound fair but skew incentives toward high-volume posters and publishers; recency matters, yet relevance often matters more for long-term satisfaction, so the job is balancing agency with a system that serves the whole community.

AI content is a tailwind and a challenge; in a flood of synthetic media, people will seek distinct points of view, so we invest in creators broadly and judge work by merit and voice, not by the tools used to make it.

We’ll label AI where we can and may also label camera-captured content as a durable signal; at the account level, disclosure helps fight spammy AI personas pushing scams.

I admire TikTok’s exploration-heavy ranking that can surface new talent; we’ve been pushing exploration, originality, and recency so more fresh creators can break out, and we’re closing the gap.

I show up publicly because the debate will happen anyway; better to explain trade-offs and decisions, accept fair and unfair hits, and make the process more understandable over time.

On the ‘full-screen video feed’ uproar, several things got conflated: a small iOS test, a bigger push into Reels, more recommendations, and creator reach concerns; we likely moved too fast, but I’d rather take periodic backlash than stagnate.

At our scale, you must test, and tests will leak, so we pre-plan calm comms for controversial experiments, especially pricing or major UI changes.

Fail corner: Facebook Home was a spectacular miss that taught me operating-system realities and when to end a project; at Instagram, our first Reels built into Stories was the wrong foundation—had we shipped the standalone version a year earlier, TikTok might look different today.

With my kids, it’s boundaries and literacy: they earn weekend iPad time, we approve apps, flights are exceptions, and we’re learning to build with AI together so they grow critical thinking and modern tool fluency.

Adam, thank you—authentic, thoughtful, and generous with the real trade-offs behind the product we all use.

My ask: remember this work lives in gray areas with real trade-offs, and while you may disagree, we’re people doing our best—keep the critique coming.

Thanks for listening; subscribe, rate, and find past episodes at Lenny’s Podcast dot com—see you next time.