Data Over Code: A Short Hyperbole
summary
Argues that systems work better when code is split into data vs computation vs action and when executable behavior is turned into data. Shows routers implemented as vectors of tuples and interceptors replacing middleware, describes Onyx for dataflow clusters and Clojure spec + spec-tools for validation and generators, and demos EQL/DataScript pull APIs and Fulcro integration for server/client rendering. Claims reification yields optimizations (dynamic router benchmarks show 97.1% — ~3% slower than a C++ impl), and demonstrates these designs work for streaming, batching and introspection.
AI-written summary · about
chapters (16)
- 0:00Intro
- 8:44Side Effects
- 10:28What is Data
- 17:49What is Computation
- 22:31The 3 Stages of Programming
- 28:08Middleware
- 30:04Interceptors
- 33:05Compilers
- 35:45Onyx
- 37:54Des
- 40:03Macros
- 41:21Data Structures
- 43:17Pull API
- 45:03EQL
- 46:37React Component
- 50:41Conclusion
related talks
17mBring Your Own Data Structures to Datalog
30mRetooling for a Smaller Data Era
01h 01mLogic and Algebras for Distributed Computing: Applications and Open Questions
28mKnowledge Graph of Your IT Landscape: Code Is Graph Data
58mComposable Data Systems, the new frontier
42mA Practical Functional Relational Architecture
42mCategories for the Working Hacker
36mEmbrace the chaos: Composable data systems with fewer asterisks