The Organizational Logic Behind Tech Stack Choices at Alibaba, Tencent, ByteDance and Bilibili

Forum topic · QianXun · 2025-11-24

Summary

This forum post analyzes why Alibaba, Tencent, ByteDance, and Bilibili chose different programming languages, arguing that tech stack selection reflects organizational structure rather than pure engineering merit. It proposes a framework based on the 'potential difference' between managers' and developers' technical expertise, yielding three models: Java as an industrialized pipeline for 'laypeople managing experts' (Alibaba), C++ as an elite special-forces model of 'experts managing experts' (early Tencent), and Go as a standardized cloud-native army for 'experts managing laypeople' (ByteDance, and Bilibili after a failed Java transition from PHP). The post traces Alibaba's Java empire to its e-commerce complexity, the 'de-IOE' movement, and in-house middleware like Dubbo, RocketMQ, and OceanBase; Tencent's C++ heritage to its telecom roots and horse-racing internal competition, followed by a shift to Go for scalability; ByteDance's Go adoption to cloud-native timing and rapid APP-factory iteration with frameworks like Kitex; and Bilibili's successful migration to Go via the Kratos framework. The core conclusion: tech stacks are extensions of managerial will—Go's restrictions beat C++'s freedom as talent density falls, while Java standardization remains a heavy but robust path for hierarchical organizations.

Tech Stack as an Organizational Code: The Balancing Act of Managerial Potential Difference

> Note: Tech stack choice is not an isolated engineering decision but a combined reflection of organizational culture, management structure, and talent supply. The "potential difference" between managers' technical ability (expert or layperson) and implementers' skill level determines the boundaries and direction of a tech stack—much like voltage difference drives the flow of current.

The post identifies three models:

1. Java: The industrial assembly line (laypeople managing experts) — Strong typing, verbose conventions, and a rich ecosystem (Spring, Dubbo) let architects design pipelines that ordinary developers simply "fill in." Suits large-scale, complex business scenarios by smoothing out individual skill differences. 2. C++: The elite special forces (experts managing experts) — Maximum freedom demands maximum expertise. Ideal for high performance, but requires extreme talent density; risk spikes when "experts" are scarce. 3. Go: The standardized cloud-native army (experts managing laypeople) — Simplicity, enforced formatting (gofmt), and native concurrency are designed for large, fast-growing teams. Go embodies managerial control.

Alibaba: The Industrial Rise of a Java Empire

E-commerce complexity: Taobao/Tmall involve long-chain, high-concurrency logic. Java's OOP modeling and Spring's IoC/AOP/transactions fit perfectly.
De-IOE movement: Replacing IBM minicomputers, Oracle databases, and EMC storage spurred self-developed Java middleware—HSF, Tair, OceanBase, Dubbo, RocketMQ, Nacos—built by top engineers so that thousands of P6/P7 developers only write business code in Spring containers.
Managerial fit: Java's standardized skills ease replacement, hiring, and error isolation—matching Alibaba's hierarchical bureaucracy.
SoftBank influence: Early investment brought in Yahoo-era architects experienced with Java, accelerating industrialization.

Tencent: C++ Elite Legacy and Transformation Pains

Telecom roots: QQ and WeChat are IM systems demanding extreme performance; founders and early staff came from Huawei and Runxun, fluent in TCP/IP optimization. C++'s epoll models and custom memory pools handled millions of concurrent connections.
Horse-racing mechanism: Internal competition and independent studios matched C++'s flexibility, spawning a wheel-reinvention culture and multiple RPC/storage/logging implementations.
Transition to Go: As Tencent moved from "connection" (IM, social) to "services" (cloud, gaming, video), C++'s long compile times, debugging difficulty, and talent scarcity forced adoption of Go in control-plane and web scenarios; the Tars microservice framework now supports Go.

ByteDance: The Brutalist Aesthetics of Go

Cloud-native timing: ByteDance rose when Kubernetes and microservices matured; Go's Goroutines and net/http fit high-performance microservices.
APP-factory expansion: Only 25 keywords and mandatory gofmt mean new hires push code within days; garbage collection and goroutines lower cognitive load; fast compilation matches the "brute force yields miracles" pace. Kitex and CloudWeWeaver-style infrastructure let business teams simply call APIs.
Managerial control: Go bans template metaprogramming and macros, reducing the risk of unmaintainable "show-off" code.

Bilibili: A Winding Path from PHP to Go

PHP era: Fast development and cheap hiring, but weak typing and the FPM model failed under high-concurrency danmaku (bullet comments) and live streaming.
Failed Java attempt: Steep learning curve (Spring's IoC/AOP), insufficient architect resources to build middleware foundations, and cultural mismatch with young engineers.
Successful Go migration: Led by Mao Jian, Bilibili built the Kratos microservice framework. Go's C/PHP-like procedural syntax eased the transition while providing static typing and Goroutine-based concurrency for low-latency, high-throughput systems. Kratos doubled as a tool of organizational control, unifying the engineering team.

Summary and Insights: The Tech Stack Is a Mirror of the Organization

Tech stack choice is fundamentally an organizational ROI calculation where the core variable is people:

Alibaba (Java): High fixed cost (middleware teams, complex frameworks) traded for low marginal cost (replaceable ordinary developers)—a heavy-industry model for hierarchical organizations.
Tencent (C++ → Go): From elite-driven artisan workshops to a modern standardized factory.
ByteDance (Go): Born in the cloud-native era, maximizing labor mobility and productivity under 996 pressure—an integrated-circuit model.
Bilibili (PHP → Go): Proof that Go is the best bridge for mid-sized teams modernizing their stack.

Core conclusion: The tech stack is an extension of managerial will. Java suits large organizations needing stability and scalability; C++ suits elite-driven high-performance scenarios; Go is optimal for rapid expansion and standardization. As organizations scale and talent density falls, restricting choices (Go) beats offering freedom (C++), while building standards (Java) is another heavy but robust survival strategy.

References

1. Conway, M. E. (1968). *How Do Committees Invention?* Datamation. 2. Alibaba Cloud Developer Community. (2020). *Alibaba's De-IOE Technical Evolution*. 3. Tencent Engineering. (2021). *The Evolution of the Tars Framework and Go Practices*. 4. ByteDance Tech Team. (2022). *Kitex: ByteDance's Go Microservice Framework*. 5. Bilibili Tech Team. (2021). *Kratos: The Path from PHP to Go*.