Voice profile — write toward something, not just away from tells
/humanize is the negative direction: it finds AI tells and says
what to remove. That leaves a draft that is merely less bad.
This is the positive direction: a written description of how you actually write, extracted from your own published work, so a draft can be measured against a target instead of a taboo list.
What this does not do. A voice profile makes prose sound like your prose. It does not make model-generated text stop reading as model-generated to a neural detector — nothing an LLM applies to its own output does. See
writing-with-ai.md. Use this to write well in your own register; write the load-bearing sentences yourself.
Building the profile
1. Assemble the corpus, and count it
Three to twelve of your own pieces where you were the primary writer. Published papers are best — they survived editing. Mix genres if you write in several (paper, referee report, grant, teaching notes); the profile should note where your register changes.
bashfind <corpus-dir> -maxdepth 1 \( -name '*.pdf' -o -name '*.tex' \) | wc -l
(find, not a glob — in zsh an unmatched glob aborts the whole command, which
reports 0 and defeats the count this step exists for.)
Count before starting. A corpus of eleven is a different task from four, and discovering that halfway through is how a session gets reset.
2. One subagent per document — never load the corpus into one context
Per pdf-processing.md: spawn one subagent per document
in a fresh context. Each reads only its own file, writes a ~300-word note to
notes/voice/<name>.md against the fixed schema below, and returns only the filename.
The main session then reads only the notes. Loading a whole corpus at once has repeatedly forced a session reset after partial work was already lost.
Per-document note schema — the same six headings every time, so the synthesis can compare:
## Lexicon words and phrases used repeatedly; words conspicuously avoided ## Rhythm typical sentence length; variance; where long sentences appear ## Openings how sections and paragraphs begin; how the paper opens ## Transitions the actual connectives used, verbatim, with rough frequency ## Hedging how uncertainty is expressed; how strong claims are made ## Quirks anything distinctive — punctuation habits, first person, humour, footnotes
3. Synthesize, and mark what is stable
Read only the notes. A trait belongs in the profile if it appears across most of the corpus, not because one paper did it once. Record frequencies where you can: "'note that' appears in 7 of 9 papers; 'delve' appears in none."
Write to voice-profile.md at the repo root (allowlisted in the repo-hygiene gate). Include:
- Signature vocabulary and the avoid list — words the author demonstrably does not use.
- Sentence rhythm, with a number: median length, and where the long ones land.
- Structural habits — how an introduction is built, where the contribution paragraph sits, how results are framed.
- Hedging register — the author's actual calibration language, which is usually narrower than a model's default.
- Deliberate quirks, labelled as deliberate. "Uses em-dashes frequently and on purpose"
stops
/humanizefrom flagging a habit as an AI tell. - Where the register shifts by genre.
4. Wire it in
/humanize reads voice-profile.md when present and respects documented preferences — a
quirk you have declared deliberate is no longer a finding. Point drafting work at the profile
before it writes, not after.
Auditing a draft
Pass --audit followed by a filename to compare an existing draft against the profile instead of building one:
/voice-profile --audit main.tex
Report, per section: distance from the profile, with concrete evidence — vocabulary outside your range, hedging denser than your baseline, transitions you do not use, sentence rhythm that has flattened. Every finding cites the profile line it violates, so it is a deduction rather than taste.
Read-only. Auto-rewriting prose degrades it and introduces new tells, and it cannot change what a detector sees. The report says where and why; the author edits.
Anti-patterns
- Profiling coauthored work you did not draft. You will extract someone else's voice.
- Treating the profile as a rulebook. It describes what you have done, not what you must do. Voices change; re-profile after a few new papers.
- Building it from AI-assisted drafts. The profile will encode the model's register as yours — the corpus must be work you wrote.
- Using it to pass a detector. It is a writing aid, not a laundering step. If a venue wants
an AI-use statement, make one (
/submission-disclosures).
Cross-references
writing-with-ai.md— readability vs provenance, and the human-readable standard/humanize— the negative direction; reads this profile when it exists/proofread— grammar and consistency, a separate lenspdf-processing.md— the one-subagent-per-document pattern
