The Twenty Minute VCTomer Cohen: Why LinkedIn Stories Failed; How LinkedIn's Feed Was Born; AI Startups | E1019
Harry Stebbings and Tomer Cohen on linkedIn’s CPO on AI’s Future, Product Craft, and Reinventing the Feed.
In this episode of The Twenty Minute VC, featuring Tomer Cohen and Harry Stebbings, Tomer Cohen: Why LinkedIn Stories Failed; How LinkedIn's Feed Was Born; AI Startups | E1019 explores linkedIn’s CPO on AI’s Future, Product Craft, and Reinventing the Feed LinkedIn CPO Tomer Cohen discusses his journey into product leadership, how he thinks about the craft of product as a blend of art, science, and execution, and the importance of clarity of vision and “jobs to be done.”
At a glance
WHAT IT’S REALLY ABOUT
LinkedIn’s CPO on AI’s Future, Product Craft, and Reinventing the Feed
- LinkedIn CPO Tomer Cohen discusses his journey into product leadership, how he thinks about the craft of product as a blend of art, science, and execution, and the importance of clarity of vision and “jobs to be done.”
- He explains how LinkedIn re-architected its feed from an internal promotion channel into a member-centric engine for professional conversations, and why controversial moves like killing Stories or throttling shallow growth were necessary.
- Cohen dives deep into AI’s impact on product building, organizational design, and competition between startups and incumbents, arguing that AI skills and prompt mastery are becoming core product competencies.
- He also explores where AI might go next—towards generating new scientific knowledge—and wrestles with questions of responsibility, regulation, data advantage, and how work, learning, and careers will change.
IDEAS WORTH REMEMBERING
7 ideasAnchor product strategy in “jobs to be done,” not just user segments.
Cohen emphasizes deeply understanding functional, social, and emotional needs—like the stress and consensus-seeking in B2B buying or creators wanting opportunity, not vanity metrics—to unlock real innovation and avoid misfires like LinkedIn Stories.
Clarity of thought beats hedging; be “wrong but not confused.”
He argues that strong, explicit hypotheses and principles—set before building—enable decisive execution and clear learning, even when bets fail, whereas confusion leaves outcomes to luck.
Treat feeds and discovery surfaces as member-owned, not org-owned.
Rebuilding LinkedIn’s feed from an internal promotion channel into a space for conversations between people you care about required overruling teams’ growth dependencies and re-optimizing for trust, quality, and member jobs-to-be-done.
AI literacy is now a core product skill, not a nice-to-have.
Cohen insists product leaders must understand ranking, data quality, prompting, and limits of models, and reorganize teams so AI is embedded (not a horizontal afterthought), because AI is becoming the main lever steering products.
Startups’ AI edge lies in rethinking problems, not thin wrappers.
While incumbents have compute, customers, and proprietary data, he believes startups can win by reimagining workflows and industries end-to-end with AI, especially when they bring truly unique datasets or specialized fine-tuning—rather than shallow layers on top of GPT.
Measure product bets by adoption plus retention, not launch or vanity growth.
LinkedIn defines success criteria pre-launch and looks for sustained, repeat usage; early Stories usage was mediocre and failed its underlying job (creators wanted permanence and identity, not ephemerality), prompting a strategic pivot.
AI will force unlearning of control-oriented product habits.
Because generative models are non-deterministic, product leaders must shift from micromanaging every UI detail to setting constraints, principles, and training data, then steering outcomes via prompts and feedback rather than direct control.
WORDS WORTH SAVING
5 quotesWe might be wrong, but we’re not confused.
— Tomer Cohen
The feed belongs to the member. It’s not an org chart.
— Tomer Cohen
Stories failed because we completely misunderstood the job to be done for creators on LinkedIn. They didn’t want things to disappear; they wanted them to last.
— Tomer Cohen
With AI, you don’t control the experience. It’s as if you’re the chef giving ingredients and philosophy, but the AI learns how to cook.
— Tomer Cohen
These models are amazing at restructuring existing knowledge. The real frontier is when they start to come up with new knowledge.
— Tomer Cohen
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsHow should product teams practically retrain and reorganize themselves to make AI a core capability rather than a bolt-on feature?
LinkedIn CPO Tomer Cohen discusses his journey into product leadership, how he thinks about the craft of product as a blend of art, science, and execution, and the importance of clarity of vision and “jobs to be done.”
What safeguards and governance structures does LinkedIn use internally when deciding to deploy powerful AI models in member-facing experiences?
He explains how LinkedIn re-architected its feed from an internal promotion channel into a member-centric engine for professional conversations, and why controversial moves like killing Stories or throttling shallow growth were necessary.
How will LinkedIn adapt its business model and content ecosystem if AI assistants increasingly answer professional questions without users visiting LinkedIn directly?
Cohen dives deep into AI’s impact on product building, organizational design, and competition between startups and incumbents, arguing that AI skills and prompt mastery are becoming core product competencies.
Where is the line between useful professional personalization and unsettling surveillance when AI can access rich behavioral and network data?
He also explores where AI might go next—towards generating new scientific knowledge—and wrestles with questions of responsibility, regulation, data advantage, and how work, learning, and careers will change.
If AI starts generating genuinely new scientific or strategic knowledge, how should ownership, attribution, and economic value of those discoveries be handled?
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome