Two-Pizza Team Rule
Jeff Bezos's organizational heuristic that team size should be limited to roughly six to eight people—the number two pizzas can feed. The reasoning is grounded in communication dynamics: small teams communicate more effectively, decide faster, and produce more coherent output because communication overhead scales with the square of team size (a six-person team has fifteen channels, a twelve-person team has sixty-six) while productive capacity scales linearly. Conway's Law explains why this works: reducing team size reduces communication channels, which reduces the interfaces and potential misalignments that organizational communication introduces. Each small team produces a coherent component because internal communication is simple and effective. But the rule trades coherence within teams for fragmentation between teams—interfaces between components reflect inter-team communication, which is necessarily more formal and constrained.
In the AI Story
Amazon combined the two-pizza rule with the API mandate—requiring all teams to expose functionality through well-defined interfaces with no other communication form (shared databases, back-channels, informal agreements) permitted. This was the Inverse Conway Maneuver at scale: deliberately engineering inter-team communication structure to produce the desired inter-component architecture. The combination produced microservices—systems of many small, autonomous services communicating through well-defined interfaces. The architecture wasn't chosen then implemented; it was produced by organizational structure through Conway's Law's mechanism.
AI disrupts this equilibrium. When a single person can build what previously required a team, the two-pizza team becomes a one-person team with zero internal communication overhead. The communication structure is the individual's cognitive structure, which doesn't produce the inter-team fragmentation driving microservices architecture. One-person teams or very small AI-augmented teams are likely to produce more integrated architectures—not tangled codebases but tightly coordinated systems reflecting unified mental models rather than negotiated interface contracts. Evidence already visible: AI-native companies like Bolt.new (fifteen engineers, forty million dollars annual revenue) and Cursor (roughly three hundred employees, over a billion in revenue) build with radically small teams producing architectures reflecting the integration small-team communication enables.
The Infralovers analysis observed: "AI shrinks teams. Conway says: small teams build monoliths." Not monoliths in the pejorative sense (unmaintainable tangles) but in the structural sense: integrated systems rather than federations of autonomous services. This challenges the decade-long industry consensus that microservices are inherently superior architecture. The superiority depended on the organizational context—distributed teams requiring formal interfaces. When teams shrink to individuals or pairs, the architectural calculus changes. The coordination overhead that made monoliths unwieldy at scale may be manageable when the "team" is two people sitting together or a single person with AI.
Origin
The two-pizza rule emerged at Amazon in the early 2000s as Jeff Bezos formalized principles observed during the company's rapid scaling. Amazon had discovered that large teams produced slower decisions and more bureaucratic output—classic symptoms of communication overhead overwhelming productive capacity. The rule was never about pizza (feeding preferences vary, and not all teams eat together) but about the mathematical relationship between team size and communication complexity. Bezos and his technical leadership understood implicitly what Conway's Law predicted explicitly: small teams with simple communication produce coherent components. The rule spread widely beyond Amazon because it compressed a complex organizational insight into a memorable heuristic that practitioners could apply without needing to understand the underlying communication mathematics.
Key Ideas
Communication scales quadratically. A team of n people has n(n-1)/2 communication channels. Six people: fifteen channels. Twelve people: sixty-six channels. Each channel is a potential misalignment source.
Coherence within, fragmentation between. Small teams produce coherent components (simple internal communication) but fragmented systems (constrained inter-team communication). The architecture reflects both properties.
API mandate as Inverse Conway Maneuver. Forcing interface-based inter-team communication produces service-oriented architecture. The technical requirement shapes communication topology, which shapes system topology.
AI enables one-sandwich teams. When individuals can build team-scale systems, the two-pizza rule becomes the one-person rule. Communication overhead drops to zero within "teams" (it's cognitive, not organizational).
Integrated architectures return. Very small AI-augmented teams are likely to produce more integrated systems than large distributed teams produced, reversing the decade-long trend toward microservices—not regression but adjustment to new communication realities.
Appears in the Orange Pill Cycle
Further reading
- Steve Yegge, "Platform Rant" (Google+ post, 2011), on Amazon's API mandate
- Sam Newman, Building Microservices (O'Reilly, 2015)
- Fred Brooks, The Mythical Man-Month, on team size and communication
- Matthew Skelton and Manuel Pais, Team Topologies (IT Revolution Press, 2019)
- Werner Vogels, "Amazon's Architecture" (All Things Distributed blog, 2006)