APIs are Illness and Cure: The Software Heterogeneity Problem in Web Programming
summary
Modern web apps produce vast heterogeneous graphs — companies average 300 APIs (5k–10k size) and many internal endpoints lack specs (only 35% rate documentation as above average). The claim is that networked APIs both cause and solve the heterogeneity problem by raising the abstraction to the API layer. Akita implements a blackbox dynamic analysis agent that listens to traffic, infers OpenAPI-style specs, supports HAR files and CI runs, performs diffing specs, posts GitHub comments, and demonstrates actionable, incremental checks that already work on real test and production traces.
AI-written summary · about
chapters (13)
- 0:00Introduction and background
- 2:34The promise and perils of APIs
- 4:57Defining the problem space
- 6:40From academia to entrepreneurship
- 11:23Challenges in API maintenance
- 15:30Why existing tools fall short
- 18:39The software heterogeneity problem
- 21:12Shifting the perspective
- 25:12API-based behavioral modeling
- 29:40Leveraging existing specifications
- 31:32Applying programming language magic
- 35:17Akita software in action
- 39:30Conclusion and takeaways
related talks
43mAPI Security By Design
23mDesigning APIs Both Humans and AI Love
35mBuilding Your API for Longevity
31mAPI testing journey through web protocols
43mDesigning REST APIs for the age of AI agents
01h 07mDesign and Build Great Web APIs
24mDesigning the APIs for an Internal Set of Services
07mThe Transition from Strict APIs to Flexible, Context-Aware Interfaces