Episode Summary
Executive Summary: LinkedIn CPO Toma Cohen argues product leadership blends science and art: clear hypotheses, data, and experimentation matter, but intuition, creativity, and vision distinguish great builders. The conversation spans LinkedIn’s evolution in feed, mobile, and creator products; why product-market fit matters more than being first; how AI changes product design, team skills, and competitive moats; and why responsible AI, growth mindset, and cross-functional product reviews are essential.
Main Topics: Product as a blend of science and art (Priority: 5/5): Cohen rejects a strict split between art and science, arguing strong product work requires applied best practices plus imagination, judgment, and the ability to uncover unspoken customer needs. LinkedIn’s product evolution and controversial decisions (Priority: 5/5): He describes major shifts at LinkedIn, including moving from desktop to mobile, reworking the feed to serve members rather than internal teams, and learning from failed bets like Stories and Instant Articles. Product-market fit over first-to-launch (Priority: 4/5): Cohen argues that being first is less important than reaching product-market fit first, since that creates durable insights, momentum, and retention. AI reshapes product building and team skills (Priority: 5/5): He says AI changes how products are designed, reducing deterministic control, increasing the need for prompts and AI fluency, and requiring teams to develop new technical capabilities. Data, fine-tuning, and where AI value accrues (Priority: 5/5): Cohen argues public data advantages are shrinking because foundation models are trained on public information; differentiation increasingly comes from proprietary data, fine-tuning, and specialized applications. Responsible AI, regulation, and future knowledge creation (Priority: 4/5): He emphasizes transparency, privacy, inclusivity, and factual accuracy, and predicts a future where AI moves beyond recombining knowledge into generating new scientific knowledge and more autonomous systems. Product leadership operating model at LinkedIn (Priority: 4/5): He explains LinkedIn’s quarterly product reviews, in-person product jams, feedback-heavy decision-making, and the importance of company-wide ecosystem thinking rather than narrow feature ownership.
Key Arguments: Great product leadership is not purely analytical or purely creative; it requires both rigorous method and intuition to see unmet needs and anticipate market shifts. Before launch, teams should define success metrics clearly; post-launch, adoption and retention matter more than short-term usage spikes. LinkedIn’s feed needed to belong to the member, not act as an internal promotional channel for different teams. Launching Stories showed that ephemeral sharing was not the real job-to-be-done on LinkedIn; users wanted content to endure and reinforce professional identity. Being first to launch is less important than being first to product-market fit, because PMF yields real momentum and learning. AI makes product experiences less deterministic, so product leaders must shift from fully controlling the experience to guiding principles, inputs, and guardrails. The prompt is becoming a critical human-machine interface and an important skill for product and AI teams. Public-data advantages are diminishing because foundation models have already ingested much of the public internet; proprietary data still matters, but mostly when specialized through fine-tuning. Startups can still win by rethinking workflows from scratch and exploiting niche, proprietary, or high-quality domain data. Responsible AI must be built in through transparency, inclusivity, privacy, and factual accuracy rather than retrofitted later. The next frontier is AI-generated new knowledge, not just better synthesis of existing knowledge, which could transform science, business, and society.
Data Points: LinkedIn revenue growth: More than tripled - Cohen cites this as part of LinkedIn’s recent product and business evolution. LinkedIn member base: More than doubled - He references growth in the LinkedIn community during his tenure. LinkedIn engagement: At record levels - He says engagement is at all-time highs. LinkedIn growth rate: Fastest it has ever grown - Cohen describes current company growth as the fastest in LinkedIn history. Company scale on LinkedIn: 900 million members - He cites LinkedIn’s consumer platform reach. Companies on LinkedIn: 60 million - He references the number of companies represented on the platform. Skill-set change over 5 years: 25% different today - Cohen says the required skill set has shifted materially in five years. Skill-set change over next 5 years: 50% different - He predicts accelerating change in job skills. AI code generation share on GitHub: 41% of new code creators - Harry cites this statistic during the discussion on AI productivity. Intermittent fasting schedule: 18:6 to 24:4 - Cohen describes his personal fasting routine. Fasting frequency: Three days a week for a month - Harry shares his own trial period when discussing fasting. Visual processing: 90% of information processed by the brain is visual - Mentioned in the Canva ad read. Canva adoption: 85% of Fortune 500 companies - Mentioned in the Canva ad read.
Pivotal Quotes: "I think there isn't a notion of enough data, is that the hypothesis you have isn't being validated." — Toma Cohen: On deciding when post-launch data is sufficient and whether a product hypothesis is working. "If you're first to product market fit that is amazing. If you're first to product market fit, you build an amazing leg up in terms of both insights and momentum and speed." — Toma Cohen: On why product-market fit matters more than being first to launch. "With AI, you don't control the experience. AI is not deterministic." — Toma Cohen: On how AI changes product leadership and the need to give up some control.
Implications: Product teams should optimize for clear hypotheses, retention, and ecosystem thinking while building AI fluency. Durable advantage will come less from generic data and more from proprietary workflows, fine-tuning, and responsible deployment.