build-vs-buy

Source Metadata

Original cangjie-skill frontmatter from the distillation run:

id: f16
title: "不在火车前冲刺 — AI自建vs购买决策框架"
book: "AI for Everyone"
author: "Andrew Ng"
source: p15, p17
tags: [build-vs-buy, strategy, resource-allocation, platform]
related_skills:
  - slug: ai-strategy-moat, relation: depends-on
  - slug: triple-due-diligence, relation: composes-with

R — Reading (Original Quote)

"Don't sprint in front of a train... if there is a company building an industry-standard solution, you should avoid constantly accelerating to stay ahead of the train... it is better to accept the industry-standard solution." (p15)

I — Interpretation

The "train" metaphor: when a technology component is becoming an industry standard (an open-source project, a startup's product, a major platform feature), you cannot invest more resources than the entire market into building it yourself. The train will eventually catch up and crush your proprietary solution.

Decision rule:

• Is becoming an industry standard → Buy/adopt. Don't try to outrun the train.
• Truly unique to your business and able to create a competitive advantage → Build. This is the differentiation work worth investing in.

Core principle: Focus your limited engineering resourcesonly on work that can bring differentiation. For components that are becoming standardized, accept the industry-standard solution and invest the saved effort into truly unique business value.

A1 — Past Application

Case 1: Building your own cloud computing infrastructure Many large companies once invested heavily in building their own data centers and cloud platforms, trying to compete with AWS/Azure/GCP. The result: these "trains" (the three major cloud providers) invested billions of dollars, which no single enterprise could match. In the end, these companies turned to adopting public cloud one after another — accepting the industry-standard solution.

Case 2: Building your own machine learning framework Some teams once developed their own deep learning frameworks, trying to catch up with TensorFlow/PyTorch. But behind these two frameworks is the continuous investment of the entire open-source community and giant companies, and self-built frameworks find it hard to keep up with the pace of iteration. Wise teams chose to adopt open-source frameworks and focus their energy on using the frameworks to solve business problems.

A2 — Future Trigger

Trigger scenarios:

• The team discusses "should this component be built or bought"
• In a technology selection meeting, people argue about "whether to reinvent the wheel"
• Evaluating an external vendor solution vs. internal R&D

Language signals:

• "build", "buy", "external solution", "vendor", "third party", "reinvent the wheel", "industry standard"

Key distinction:

• Is the componentbecoming standardized(multiple vendors/open-source projects converging) orstill unique(only you or a few others are doing it)?
• Is the component yourcore source of differentiationorinfrastructure？

E — Execution

1. Determine the industry-standard trend: Is this component becoming an industry standard? (Check: are startups, open-source projects, or large companies converging on the same solution?)
2. If yes → Adopt the standard: Don't build it yourself; choose the industry-leading solution and free up resources for differentiation work.
3. If no → Evaluate the differentiation value: Is this component truly unique to your business and capable of creating a competitive advantage? If so, invest resources to build it in-house.
4. Regular review: The industry-standard landscape changes; reassess the decision every 6-12 months.

Completion criteria: Complete a "build/buy" decision for each candidate component, and clearly record the decision rationale and review time.

B — Boundary

Not applicable scenarios:

• Core differentiation components (theseshouldbe built in-house; do not use this framework to reject them)
• Trivial everyday decisions (not worth analyzing with this framework)
• Early exploration stage (not yet enough information to judge the trend)

Failure modes:

• Mistaking a component that is "being standardized" for a differentiation opportunity and investing in building it in-house
• Mistaking a "truly unique" component for something to outsource, losing competitive advantage
• Over-reliance on external suppliers leading to vendor lock-in

Author's blind spots:

• "Industry standards" oftenonly in hindsightcan be clearly judged—when making decisions in real time, it is very difficult to determine whether a certain approach will become a standard. You need to combine market signals (funding, community activity, moves by major players) to make the best judgment, but remain humble: you may be wrong.

Related skills

• ai-strategy-moat — depends-on: Judging "what is worth building in-house" requires first understanding the theory of AI competitive moats; build decisions should serve the construction of lasting competitive advantage.
• triple-due-diligence — composes-with: Buy decisions require due diligence across three dimensions—technology, supplier, and compliance—and should be used in conjunction with the build/buy decision framework.