openai/codex scored 18/100 on Slopicide
Handcrafted
Artisanal, small-batch code. A human clearly sweated over this.
What stood out
How strongly each pattern showed up overall, and in how many of the files.
- Copy-paste repetition1/8 files · 23%
Near-duplicate blocks with small edits instead of a shared helper.
- Boilerplate padding1/8 files · 21%
More lines, wrappers and ceremony than the job needs.
Human fingerprints
- Domain-specific naming8/8 files · 68%
Names that show deep familiarity with the problem.
- Pragmatic shortcuts8/8 files · 44%
Terse idioms and deliberate omissions an experienced author takes.
- Lived-in code5/8 files · 39%
Idiosyncrasies and quirks of code that evolved over time.
- Comments explain why1/8 files · 21%
Comments about trade-offs, gotchas, bugs and workarounds.
Files we looked at
The files behind the score, sloppiest first.
- codex-rs/tui/src/app/recap_tests.rs33out of 100
Rust · 35 KB
- codex-rs/core/src/agent/control/budget.rs30out of 100
Rust · 1.2 KB
- Boilerplate padding 30%
Rust · 4.8 KB
Rust · 144 KB
Rust · 5.6 KB
- codex-rs/state/src/runtime/threads.rs11out of 100
Rust · 141 KB
- Copy-paste repetition 44%
- codex-rs/core-plugins/src/manager.rs5out of 100
Rust · 145 KB
Rust · 16 KB
Scanned Oct 30, 2026 from main. A playful gut check, not a verdict on the people behind the code. How scoring works. More recent reports.