revert: remove STYLE.md and undo its wording changes

STYLE.md was still a work in progress. Reverts the file, CLAUDE.md
reference, and "just"/"simply" word removals it triggered.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: derekallman

CHANGELOG.md

@@ -132,7 +132,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
 - CI — added `cargo-deny` dependency audit step (ubuntu-only); test step scoped to `-p hotcoco -p hotcoco-cli` (excludes cdylib)
 - Pre-commit hook step numbering corrected from `[1/4]` to `[1/3]` (matches actual 3-step hook)
 - `_typos.toml` — new typos-cli configuration: `en-us` locale, domain abbreviations allow-listed (`nd`, `obb`, `oks`, etc.), data/fixture dirs excluded
-- `STYLE.md` — new style guide covering product names, domain glossary, abstraction layers, Python API naming, Rust conventions, CLI output, documentation voice, and anti-patterns
 - `python/hotcoco/cli.py` — reformatted with ruff (line length compliance); `cmd_merge` import moved to function scope
 - Docs — American English spelling throughout (`behaviour` → `behavior`, `normalised` → `normalized`, `maximises` → `maximizes`, `summarisation` → `summarization`, `Randomise` → `Randomize`); benchmark versions updated to 0.3.0
 - Browse UI: `nav_query | safe` in `detail.html` replaced with JSON-encoded `<script type="application/json">` data element parsed in `overlay.js`, eliminating a potential XSS surface from string interpolation

CLAUDE.md

@@ -2,7 +2,7 @@
 
 This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
 
-Read CLAUDE.md carefully before starting any task. If you're about to write documentation, benchmarks, or make git commits, check the relevant section of CLAUDE.md for my conventions first. Code style conventions are in [STYLE.md](STYLE.md).
+Read CLAUDE.md carefully before starting any task. If you're about to write documentation, benchmarks, or make git commits, check the relevant section of CLAUDE.md for my conventions first.
 
 ## Project Overview
 

README.md

@@ -60,7 +60,7 @@ ev.run()
 - **Confidence calibration** — ECE/MCE metrics and reliability diagrams measure whether your model's confidence scores are meaningful. See [calibration](https://derekallman.github.io/hotcoco/guide/evaluation/#confidence-calibration).
 - **Model comparison** — `hotcoco.compare(eval_a, eval_b)` with per-metric deltas, per-category AP breakdown, and bootstrap confidence intervals for statistical significance. See [model comparison](https://derekallman.github.io/hotcoco/guide/evaluation/#model-comparison).
 - **Per-image diagnostics & label errors** — per-image F1/AP scores, automatic detection of wrong labels and missing annotations in your ground truth. See [diagnostics](https://derekallman.github.io/hotcoco/guide/evaluation/#per-image-diagnostics-label-error-detection).
-- **TIDE error analysis** — breaks down every FP and FN into six error types so you know *why* your model falls short, not only by how much. See [TIDE errors](https://derekallman.github.io/hotcoco/guide/tide/).
+- **TIDE error analysis** — breaks down every FP and FN into six error types so you know *why* your model falls short, not just by how much. See [TIDE errors](https://derekallman.github.io/hotcoco/guide/tide/).
 - **Confusion matrix** — cross-category matching with per-class breakdowns. See [confusion matrix](https://derekallman.github.io/hotcoco/guide/confusion-matrix/).
 - **F-scores** — F-beta averaging over precision/recall curves, analogous to mAP. See [F-scores](https://derekallman.github.io/hotcoco/guide/f-scores/).
 - **Plotting** — publication-quality PR curves, per-category AP, confusion matrices, and TIDE error breakdowns. Three built-in themes (`warm-slate`, `scientific-blue`, `ember`) with `paper_mode` for LaTeX/PowerPoint embedding. `report()` generates a single-page PDF summary. `pip install hotcoco[plot]`. See [plotting](https://derekallman.github.io/hotcoco/guide/plotting/).

STYLE.md

@@ -4,7 +4,7 @@
 
 | Feature | pycocotools | faster-coco-eval | hotcoco |
 |---------|-------------|------------------|---------|
-| **Installation** | Prebuilt wheels available | Prebuilt wheels available | Prebuilt wheels — `pip install` and go |
+| **Installation** | Prebuilt wheels available | Prebuilt wheels available | Prebuilt wheels — `pip install` just works |
 | **Metric parity** | Reference | Exact | ≤1e-4 bbox, ≤2e-4 segm, exact keypoints |
 | **LVIS evaluation** | No | Yes — via `lvis_style=True` flag | Yes — 13 metrics, `LVISeval` class, `init_as_lvis()` |
 | **TIDE error analysis** | No | No | Yes — 6 error types, ΔAP per type |

docs/benchmarks.md

@@ -184,7 +184,7 @@ with style(theme="warm-slate", paper_mode=True):
     fig.savefig("custom.pdf")
 ```
 
-To use a completely different style (seaborn, corporate rcParams, etc.), don't call
+To use a completely different style (seaborn, corporate rcParams, etc.), simply don't call
 hotcoco plot functions inside a `style()` context — the default matplotlib style applies:
 
 ```python

docs/guide/plotting.md

@@ -34,7 +34,7 @@ A drop-in replacement for pycocotools that doesn't become the bottleneck — in
 
 <div class="feature-card" markdown>
 <strong>More than a metric</strong>
-<p>TIDE error breakdown, confusion matrix, per-category AP, confidence calibration, and publication-quality plots built in. Find out <em>why</em> your model falls short, not only by how much.</p>
+<p>TIDE error breakdown, confusion matrix, per-category AP, confidence calibration, and publication-quality plots built in. Find out <em>why</em> your model falls short, not just by how much.</p>
 </div>
 
 <div class="feature-card" markdown>