Marsh Gardiner (openapi): What changes when everyone can code? Rethinking API product journeys in an LLM-world

Marsh Gardiner explores the evolution of APIs in the context of AI and large language models (LLMs). His experience includes significant contributions to the Swagger project, which evolved into the OpenAPI initiative, and ongoing involvement in platform engineering.

Key Takeaways

AI Metaphors

• Infinite Interns: AI as a vast network of assistants handling routine or laborious tasks, freeing up experts to focus on more critical work.
   
• Idea Dehydrator and Rehydrator: AI can condense complex ideas into simpler forms and then expand them back, similar to translating ideas across languages.
   
• Circus Bear: AI is likened to a trained circus bear—capable of impressive feats but still unpredictable and potentially dangerous if not managed properly.

Practical AI Use Cases

• AI-Assisted Documentation: AI can streamline processes like migrating websites or generating steps for technical tasks, showcasing its ability to enhance productivity.
   
• Embedding Vectors: Demonstrated by an example where AI handles abstract operations like “Germany plus Paris minus France” to yield “Berlin,” highlighting AI’s capacity for abstract reasoning and problem-solving.

Putting APIs into Practice

• Rubber Duck Debugging: Explaining code to an inanimate object or AI can clarify and solve problems. This technique helps refine requirements and discover insights through explanation.
   
• AI in API Documentation: Good documentation should not only describe technical details but also the semantics—why something is done, not just how. OpenAPI’s evolution aims to enhance this aspect by bridging technical descriptions with natural language.

Emerging Trends

• LLM Users as New API Consumers: With tools like ChatGPT lowering the barrier to coding, a broader range of people can become API users. This shift may lead to new user interfaces and interactions, potentially outside traditional API portals.
• Fostering Relationships: APIs must adapt to manage relationships with new kinds of users, including handling credentials, updating terms of use, and ensuring ongoing support.

Conclusion

Embrace AI’s role as an assistant and a tool, use its strengths in documentation and idea processing, and prepare for a future where API adoption and interaction may shift dramatically with the rise of LLMs.