What nobody tells engineers about becoming a CEO

• Engineer→CEO transition requires almost exactly opposite skillsets.
  • Engineering: knowable right/wrong answers. CEO: unknowable, high-consequence fog.
  • Target 80% of each exec’s domain knowledge — enough to recognize quality.
• Founder optimism is overrated; tenacity through low-probability bets wins.
  • “20% odds? We’re not quitting until we’ve lost all hope.”
• Kafka open source sat dormant until one blog post explained why real-time data matters.
  • Years of engineering impact < one well-framed narrative with 25 weird pictures.
• Must-do vs. can-do: market forces override internal resource debates.
  • Enterprise cloud demanded managed service; “no one in public cloud wants licensed software.”
  • Half of Confluent thought the cloud product was the dumbest idea ever.
• Cloud product was embarrassing for years while software hit $100M+ revenue.
  • Progress was “a little bit less shitty every day” — no dramatic inflection moment.
  • Early enterprise customers quit; Kreps pushed through with blunt-force focus.
• Centralization happens unconsciously at scale — actively preserve pockets of excellence.
  • Units with revenue targets + autonomy outperform centralized orgs.
  • Best hiring managers bypass formal pipelines; protect that.

Jay Kreps (Co-founder & CEO, Confluent), interviewed by Brett Berson (First Round Capital) · 2026-03-26 · Read on review.firstround.com