The cure for AI slop is a 1986 aircraft manual/kit
README.md
2 KB·Raw on GitHub ↗
What is in the kit, and how to run the linter yourself
The cure for AI slop is a 1986 aircraft manual - the kit
Everything the episode promises "in the description": the distilled STE skill, the linter, and the full first-party test data. Self-contained - no external dependencies beyond Python 3.
Files#
| File | What it is |
|---|---|
ste-writing-skill.md | The distilled ASD-STE100 agent skill, two modes: strict (procedures, error messages) and flavored (prose, no dictionary lockdown) |
ste-lint.py | The heuristic anti-slop linter - the machine-checkable subset of STE. Deterministic; the score delta between two texts is the signal |
experiment-results.md | The cross-model test: 6 writing tasks x 4 conditions on Claude and gpt-5.5 |
experiment-results-openai.md | Per-category detail for the gpt-5.5 run |
before-after-samples.md | The real baseline-vs-STE outputs quoted in the episode |
run-openai.py | Reproduction script for the OpenAI side of the experiment |
Run the linter#
python3 ste-lint.py your-draft.mdScore is violations per 100 words - lower is cleaner. Lint a draft, apply the skill, then lint it again - the delta between the two scores is the signal.
The headline numbers#
| Condition | Claude sonnet | gpt-5.5 |
|---|---|---|
| baseline | 4.36 | 3.54 |
| banned-words list | 4.21 (-3%) | 2.14 (-40%) |
| Orwell's 6 rules | 2.48 (-43%) | 1.69 (-52%) |
| STE skill | 1.12 (-74%) | 1.76 (-50%) |
Give the model a writing system and slop drops by half or more, on every model tested. STE was best or tied-best. A banned-words list is the least reliable fix.
Not a certified STE checker. The judgment rules of ASD-STE100 need a human; this covers the mechanical subset - which is where the slop lives.
Spec: ASD-STE100 Issue 9, free at asd-ste100.org