Experiments in Reasoning -
summary
Experiments apply functional and logic techniques to real problems: finite state transitions become pure functions that return new states using currying and Option types, encouraging referential transparency and immutability and making onExit/transition/onEntry explicit. A reverse‑engineered credit score had ~700 branches, ~500 variables and ~5,000 references and collapses to a few significant factors when modelled as constraints. MiniZinc solves a knapsack‑style 1,505‑cent menu example (first solution uses 7 items, optimized to 4), and Google OR‑Tools solves Sudoku; results tighten understanding and reveal assumptions though some experiments remain learning‑only.
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 3:33Explaining what we do
- 8:55Understanding and applying knowledge
- 16:02Toolbox gaps in software development
- 20:23Functional programming experiments
- 36:43Logic programming in practice
- 48:19Lessons from experiments
- 53:58Final thoughts and Q&A
related talks
15mCurrent Practices for Building LLM-Powered Reasoning Tools Are Ad Hoc—and We Can Do Better
26mInfinite Possibilities
43mAn Imperishable Wonderland of Infinite Fun
01h 12mDefinitely not Java! A Hands-on Introduction to Efficient Functional Programming with REASON
09mOuvroir de Code Potentiel
41mDomain Experiments
06mLightning Talk: Design by Intuition
50mSoftware Is Details