Making sense of frontend code with forensic techniques
summary
A concrete claim: forensic correlation of static, temporal and social metrics finds risky hotspots in real code. A 50,000‑line Phoenix example uses Code Cartel to map lines, cyclomatic complexity and the churn metric, revealing a hot business model file that changed constantly. AI reports (Claude) are unreliable — it actually shells out to grep and SQL queries — so rules and checks run with Dependency Cruiser and change analysis (Adam Tornhill’s change coupling) drive refactors that reduce nesting, add tests and make the team faster. Code Cartel has ~800 downloads/week.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:26Software archeology
- 6:16What is software forensics
- 8:22Visualizing code with 3D metrics
- 11:53Forensics on frontend components
- 15:23Detecting knowledge silos
- 17:08Revealing invisible coupling
- 19:32Enforcing rules with Dependency Cruiser
- 22:40Measuring delivery impact
- 23:41Getting started with Code Cartel
- 24:01Q&A
related talks
49mTreat Your Code as a Crime Scene
47mTreat Your Code as a Crime Scene
40mTreat Your Code as a Crime Scene
51mSoftware Archeology and The Code Of Doom
50mPsychotherapy for Code
49mIdentifying bad design in codebases, objectively…
45mCode Red: The Business Impact of Code Quality
32mPractical code archaeology