What Remains Scarce When Code Becomes Cheap?

BEM began as a practical answer to modular frontend development. A block grouped one interface element's technologies and made the solution reusable across services. Tooling and the shared Lego component library grew around the methodology. Architectural rules, reusable components, and a working toolchain turned it into an organizational capability.

Adoption spread through free Yandex Subbotnik meetups and engineers explaining their work. External audiences sometimes accepted BEM faster than internal teams: criticism improved the explanation, visibility attracted people who already knew the approach, and their arrival strengthened internal adoption. That flywheel needs repeated communication and patience. Yandex's user-generated services provide a useful contrast: publishing at scale also creates responsibility for harmful advice or wrong answers. Public leadership depends on trust, not merely a technical showcase.

Sergey frames the organization as a matrix. Products change quickly, while a technical specialization persists, so a technical manager primarily mentors professional growth and a product supplies the setting where that growth produces outcomes. His practice uses a bounded number of direct reports, regular one-to-ones, long product assignments, and controlled movement between domains. Quarterly demos reconnect a technical discipline with product meaning. The staffing ranges and rotation rate discussed are local heuristics, not universal rules.

Senior individual contributors matter for cross-cutting work such as security, personal data, open source, accessibility, and education. A small Staff+ group cannot implement everything and has few direct commands available; it must align product teams, establish standards, and earn adoption. The matrix therefore carries a real trade-off. Technical perfection without a business goal and feature pressure without care for debt both damage the system. A development plan has to connect product outcomes, technical mastery, and contribution to the wider team.

AI is a meta-tool rather than only a code generator. It can improve writing and meetings, shape tickets, decompose work, explore options, and build development tools. Making the same five tickets faster is still a local optimization. Systemic value appears when a leader moves up one level: asking AI to design the workflow, distributing work across agents, and engineering a feedback loop. Middle management must translate investment from above and experiments from below into a redesigned process instead of inserting another service into the old one.

Education faces the same shift. Its value is not mastery of one framework but the ability to enter an unfamiliar domain, preserve critical thinking, and verify an answer. AI may become a personal tutor and adapt material and exercises, yet learning still requires cognitive effort. When content is nearly free, people pay for focus, commitment, practice, and feedback. The episode's answer is therefore an editorial synthesis: scarcity moves from production toward problem framing, judgment, trust, and accountability for the outcome.