How does a Programmer Think About Code?
summary
Research ties code comprehension to working memory ~7 chunks and shows semantic memory activates before working memory, favoring quick recognition. The talk applies that model to API design: prefer names that enable top-down vs bottom-up comprehension, limit method complexity so an abstraction has ≤4–5 subchunks, and avoid dense symbolic syntax. The Scala Toolkit uses this via Cognitive Dimensions scoring, picks libraries (HTTP, optics, os-lib, munit, Typelevel bundle), strict dependency CI, and Scala CLI/SBT integration.
AI-written summary · about
chapters (14)
- 0:00Programmer cognition
- 1:06Language design evidence
- 3:33Cognitive chunks in code
- 6:43Code comprehension styles
- 8:50Neuroscience of code reading
- 10:31Tips for readable code
- 11:53Code as reverse chunking
- 13:58Scala Toolkit project
- 17:29Low cognitive cost libraries
- 19:13Evaluating libraries
- 24:30Selecting libraries
- 25:54Contributing to Toolkit
- 26:37Toolkit future and release
- 27:43Closing and Q&A
related talks
53mWriting Code for Humans
59mBetter Code: Human Interface
01h 16mPL and HCI: Better Together
53mDocumentation Nobody Reads and UX in the Internal Dev Kitchen
18mAre You Vibing with Vibe Coding? A New Approach to Developer Productivity
46mMastering visual communication for software design
39mCognitive Science and Design
14mCode Style Sheets: CSS for Code