Engineering excellence is a system of five interconnected competencies: building high-performing teams, setting actionable strategy, driving a relentless execution cadence, actively reducing organizational toil, and intentionally building a strong culture of values and behaviors.
To successfully integrate senior talent, implement a 'go slow to go fast' onboarding. Have new leaders spend their first 1-2 months as individual contributors. This allows them to absorb the culture, build relationships, and gain deep context before taking on management responsibilities, preventing 'organ rejection.'
When navigating steep technology curves like AI, the best knowledge resides with Individual Contributors (ICs), not managers. To accelerate, leaders must create parallel career tracks, elevate ICs into leadership roles, and use IC-only forums ('brain trusts') to bypass slow management layers.
To maintain speed in a large organization, ignore the formal org chart. Instead, structure work around small, empowered 'work chart' teams (squads). Give them a clear DRI, full context, agency, and a direct 24-hour escalation path to leadership to remove blockers.
To build a cohesive, fast-adapting organization, infrastructure teams must be as connected to the end customer as product teams are. This isn't optional. When the entire stack understands the 'why,' it builds purpose and dramatically reduces the delay in responding to market shifts.
The key benefits of vertical integration at extreme scale are speed and destiny control, not just cost. Owning the stack allows you to re-architect every layer simultaneously for a new product. This is impossible when relying on vendors who cater to the median customer, not your bleeding-edge needs.
Having experienced the dot-com bust at Akamai, Jay Parikh prioritized efficiency at Facebook from day one, even when capital was abundant. This proactive approach to preventing waste and building a frugal culture created massive capital allocation options later on, avoiding a painful course correction.
Success and scale naturally expand a company's 'learning loop,' making it take longer to learn. Leaders must constantly fight this by re-evaluating processes (e.g., meeting sizes), promoting clear ownership, and framing failures as learning opportunities to maintain speed.
Instead of spending weeks debating between two uncertain paths, execute both simultaneously. This strategy of running parallel bets may seem resource-intensive, but it doubles the speed of learning. You'll find the right answer in two months instead of the four it would take to try one, fail, then try the other.
To achieve discontinuous impact, 'big bets' must be cross-functional. A bet confined to a single team risks optimizing for a local maximum. By design, span bets across multiple teams (3-5) to force them to seek a new global maximum for the entire business, avoiding siloed thinking.