The Twenty Minute VCTomer Cohen: Why LinkedIn Stories Failed; How LinkedIn's Feed Was Born; AI Startups | E1019
CHAPTERS
- 0:00 – 0:58
Cold open: AI moving from remixing knowledge to creating new discoveries
Tomer frames today’s large models as masters of restructuring existing public knowledge—but points to the bigger frontier: models that can hypothesize and generate net-new knowledge. He highlights scientific mysteries (dark matter, Alzheimer’s, consciousness) as examples of where AI could eventually contribute original breakthroughs.
- •Today’s models largely recombine learned public information
- •The next leap is AI that can hypothesize and create new knowledge
- •Potential impact spans science and society, not just business
- •Examples: dark matter/energy, Alzheimer’s, quantum mechanics, the self
- 0:58 – 4:02
From engineering and the IDF to LinkedIn CPO (and meeting Reid Hoffman)
Tomer recounts an idiosyncratic “builder” journey: early exposure to cutting-edge tech during army service, then semiconductors, embedded systems, and a consumer startup. A Stanford talk in 2008 and meeting Reid Hoffman shaped his conviction about professional communities creating economic opportunity, eventually leading him to join LinkedIn in 2012 and become CPO in 2020.
- •Love of building across problem-solving, design, tech, and feedback loops
- •Army service and semiconductor work as formative technical experiences
- •2008 Stanford social networks lecture: Reid’s professional community thesis resonates
- •Joined LinkedIn to rebuild for mobile; became CPO in 2020
- 4:02 – 5:01
Reid Hoffman’s mentoring style: reframing problems to find “option orange”
Tomer describes Reid as a deeply philosophical thinker who reframes binary decisions into unexpected third paths. Their conversations focus less on product details and more on profound needs, complexity, and new ways of thinking that can unlock product inspiration.
- •Reid often reframes A vs. B into a surprising third approach
- •Discussions emphasize insight and needs over surface-level product debates
- •Tomer seeks Reid when facing complex, ambiguous problems
- •Mentorship as a repeatable source of product inspiration
- 5:01 – 7:40
Is product art or science? The craft foundation vs. the differentiating intuition
Tomer rejects a clean split: product requires both scientific craft (data, design, experimentation, business) and artistic judgment (vision, imagination, intuition). He argues that as product leaders grow, the expectation shifts toward anticipating needs and technology trajectories—not just executing best practices.
- •Science: best practices and craft—design, data, experimentation, business
- •Art: vision, creativity, intuition, judgment, imagination
- •Innovation requires uncovering needs users can’t articulate directly
- •Great product people bridge vision-to-execution and keep learning
- 7:40 – 10:05
What makes a true product builder: vision-to-grind, iteration from 0.6 to PMF
Discussing product archetypes (visionary/craftsman/operator), Tomer emphasizes he loves all sides—ideation through execution. He notes most launches start at “0 to 0.6,” and real product work is grinding from mediocre beginnings to product-market fit through iteration and feedback.
- •Archetypes are useful, but great builders span multiple modes
- •Execution matters: don’t stop at a whiteboard session
- •Most launches aren’t 0→1; they’re 0→0.6 and need iteration
- •Love of building (not just decision-making) is core to product identity
- 10:05 – 15:02
Roadmaps vs. customer input: Jobs-to-be-Done and the emotional layer of needs
Tomer explains that innovation starts with deep insight anchored in Jobs-to-be-Done, including emotional and social dimensions. He uses examples like B2B buying (stress, consensus-seeking) and LinkedIn creation (reputation and opportunity over dopamine) to show why functional requirements alone are insufficient.
- •Build conviction before evidence by grounding in deep customer insight
- •Move beyond audience segmentation to “humanize” needs (functional + emotional)
- •B2B buying is emotional: risk, stress, and consensus-building
- •LinkedIn creators optimize for reputation/opportunity; content is tied to identity
- 15:02 – 17:21
Founder-led product: when it’s essential, and how “founding moments” evolve
Responding to the claim that companies stop innovating once founders stop being CPO, Tomer argues context matters. Founders are uniquely valuable early, but mature companies can thrive with founders in guiding roles—while organizations may go through multiple “founding moments” that renew the original vision.
- •Founder involvement is critical early to preserve the founding insight
- •Later stages can work with founders as guides (board/chairman/mentor)
- •Successful companies can innovate post-founder-operation (examples cited)
- •Companies can experience multiple “founding moments” over time
- 17:21 – 25:37
Is LinkedIn’s UI outdated? Growth drivers, feed evolution, and social graph vs. recommendations
Tomer acknowledges UI constraints while pointing to strong business momentum and ongoing innovation across feed, messaging, and search—especially with genAI simplifying complexity. He explains LinkedIn’s enduring social graph value while also expanding topical discovery and clarifying “connect” vs “follow” as distinct relationship modes.
- •UI can constrain innovation; genAI may help simplify experiences
- •LinkedIn growth: revenue up, members up, record engagement
- •Social graph remains foundational; topical discovery expands beyond your network
- •Clarifying connect (mutual help) vs follow (learning) drives new growth
- 25:37 – 28:59
“Wrong, but not confused”: why LinkedIn Stories failed and what it taught about creators
Tomer unpacks the principle that clarity enables learning even when bets fail. LinkedIn Stories was launched to reduce the perceived risk of posting via ephemerality, but postmortems revealed creators wanted permanence—content as part of professional identity—leading LinkedIn to double down on reputation and opportunity as core creator motivations.
- •Clarity beats confusion: conviction enables clean learning loops
- •Stories hypothesis: ephemerality would unlock more posting
- •Reality: creators want permanence, visibility, and identity attachment
- •Insight: LinkedIn creation is reputational and opportunity-driven
- 28:59 – 33:49
First to market vs. first to product-market fit—and how to judge results after launch
Tomer argues the meaningful race is to product-market fit, not launch speed. He explains decision-making with data post-launch: define success metrics upfront, prioritize retention over trial adoption, and balance “evidence vs. conviction” as bets mature—illustrated by LinkedIn’s long-term investment in skills.
- •“First to launch” is a vanity metric; “first to PMF” matters
- •Define success criteria before building; avoid retrofitting narratives to data
- •Retention is the real test; adoption alone can be gamed by discovery
- •Big bets often start with conviction, then earn evidence over time (e.g., skills)
- 33:49 – 42:48
How LinkedIn runs product reviews: “product jams,” principles, and accountability mechanics
Tomer describes a rigorous cadence: quarterly big-rock reviews, then weekly sessions focused on feedback and improvement rather than judgment. He details the structure (problem, JTBD insight, principles with tradeoffs, demo), why in-person energy matters, and how follow-through is enforced via concise next steps and a “brief back” process with clear decision roles.
- •Renaming reviews to “product jams” shifts mindset to feedback as currency
- •Session structure: problem definition → JTBD → principles/tradeoffs → demo
- •In-person jams increase creative energy and discussion velocity
- •Post-jam: summarize 1–3 focus areas; team sends brief back with actions/ETAs
- •Clear roles (recommender/approver/decider) prevent diluted accountability
- 42:48 – 49:40
The controversial pivot: how LinkedIn’s modern feed was born (member-first, quality-first)
Tomer calls the feed a “minus one to one” turnaround story: shifting from an internal promotional surface to a member-owned space for professional conversations. He describes the internal innovator’s dilemma—many teams relied on the feed for discovery/revenue—and the later decision to curb shallow virality to protect trust and quality.
- •Early LinkedIn feed was activity/promotional; not centered on conversations
- •2015: assembled a dedicated feed team with a new purpose: professional dialogue
- •Controversial internal tradeoff: feed belongs to members, not org charts
- •As engagement grew, shallow spam/clickbait rose—prompting a quality/trust reset
- •Choosing trust over intoxicating growth protected long-term value
- 49:40 – 54:34
Balancing revenue vs. innovation—and the messaging challenge of “everything to everyone”
Tomer explains LinkedIn’s planning system: vision → 1–3 year strategy → annual revenue goals plus longer-term innovation, updated via continuous planning. He also addresses product marketing risk at scale by advocating narrow, specific Jobs-to-be-Done (e.g., distinct job-seeker segments) and highlights Gen Z/entry-level as a fast-growing focus area with unique needs.
- •Use vision as true north; translate into strategy with both revenue and innovation bets
- •Continuous planning adjusts as competition and strategy evolve
- •Competitors vary by job-to-be-done (jobs, recruiting, marketing, branding)
- •Avoid “everything to everyone” via narrow JTBD framing and segmentation by need
- •Gen Z/entry-level growth requires tailored matching and value props
- 54:34 – 1:27:06
AI’s impact on product orgs: new skills, bundling trends, value accrual, and responsible AI questions
Tomer argues AI is a larger revolution than mobile and product leaders must “take the pedals” by learning the tech deeply—down to data quality and algorithm objectives. The conversation spans unlearning deterministic control, enterprise bundling/unbundling dynamics, where startups vs incumbents may win, prompt craftsmanship as a critical interface, content scraping concerns, and the path toward regulation and responsible AI principles.
- •AI is the biggest tech shift yet; product leaders must learn and model usage
- •Integrate AI expertise into product teams (don’t treat it as purely horizontal)
- •Unlearn determinism: AI outputs are probabilistic; leaders must design for less control
- •Bundling wave: AI can collapse multi-step workflows into single intents/prompts
- •Value accrual: baseline public knowledge reduces some data advantage, but proprietary data + fine-tuning still matters
- •Prompting becomes a core human↔machine interface skill; best practices still emerging
- •Responsible AI: transparency, inclusivity, privacy; regulation and societal risks are real
- 1:27:06 – 1:36:24
Quick-fire: hiring questions, proudest launches, failures, CPO advice, fasting, and Microsoft’s AI strategy
In rapid Q&A, Tomer shares his go-to product interview prompts (complexity and growth mindset), reflects on major wins (mobile transformation, feed) and a failed bet (instant articles), and advises new leaders to learn the full LinkedIn ecosystem for exponential impact. He also discusses intermittent fasting habits and praises Microsoft’s OpenAI integration as standout strategy/execution.
- •Hiring: probe the most complex problem tackled + lessons from failure
- •Proudest: rebuilding LinkedIn for mobile; transforming the feed; responding during COVID
- •Failure example: instant articles—solid hypothesis, weak execution/timing and enterprise nuances
- •Advice: don’t stay in a swim lane; master the ecosystem to innovate cross-product
- •Personal: intermittent fasting approach and adaptation period
- •Most impressed: Microsoft’s OpenAI integration strategy and execution
