docs: apply the Metis design system

README.md

@@ -1,10 +1,14 @@
+<p align="center">
+  <img src="docs/metis-benchmark-hero.png" alt="Metis / Benchmark — can a model estimate software effort from public data?" width="100%">
+</p>
+
 # metis-benchmark
 
 **Can a model estimate software effort from public data? A pre-registered blind test — with an honest negative result.**
 
 [![CI](https://github.com/NaCode-Studios/metis-benchmark/actions/workflows/ci.yml/badge.svg)](https://github.com/NaCode-Studios/metis-benchmark/actions/workflows/ci.yml)
-[![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
-[![Python 3.11–3.13](https://img.shields.io/badge/python-3.11%E2%80%933.13-3776ab.svg)](pyproject.toml)
+[![License: MIT](https://img.shields.io/badge/License-MIT-5C6567?labelColor=14181A)](LICENSE)
+[![Python 3.11–3.13](https://img.shields.io/badge/python-3.11%E2%80%933.13-3776AB?logo=python&logoColor=white&labelColor=14181A)](pyproject.toml)
 
 Reproducible backtest of the **Metis** software effort-estimation
 engine on open datasets. This repository is the Phase 0 gate (G0) of the
@@ -104,5 +108,8 @@ Conformalized Quantile Regression (Romano, Patterson, Candès 2019) via the
 MAPIE library. Cite the original sources when reusing the data.
 
 ## License
-
 Code: [MIT](LICENSE). Datasets: see their respective licenses.
+
+---
+
+Brand assets — the tetradrachm mark, the wordmark, and the colour and type tokens — are in [`docs/brand`](docs/brand).

docs/brand/README.md

@@ -0,0 +1,42 @@
+# Metis brand assets
+
+The mark is a tetradrachm: a filled field, a beaded rim, and the mu of ΜΗΤΙΣ punched into it. Athenian
+silver is the reference, currency being the first instrument anyone agreed to trust because it was
+verifiably struck to a standard. The wordmark sits beside it at cap height, and each repository names
+itself after it: `METIS / BENCHMARK`, `METIS / API`, and so on.
+
+| File | Use |
+| --- | --- |
+| `metis-mark.svg` | the mark at 24px and up, where the rim reads |
+| `metis-mark-small.svg` | below 24px: the rim is dropped, the disc and the mu carry it alone |
+| `metis-mark-bronze.svg` | the bronze field |
+| `metis-mark-outline.svg` | one keyline on paper |
+| `metis-mark-reversed.svg` | on an ink ground |
+| `metis-lockup.svg` / `-reversed.svg` / `-bronze.svg` | mark, wordmark and the line beneath it |
+| `metis-favicon-512.png` | favicon, avatar, app icon |
+| `metis-tokens.css` | colour, type and tracking tokens |
+
+Every SVG here is self-contained: the letterforms are outlined, so the mark renders the same whether or
+not Bodoni Moda and Cormorant Garamond are installed.
+
+**Clear space** equals the rim inset, r/6, on every side. **Minimum size** is 24px for the full mark;
+below that use `metis-mark-small.svg`.
+
+**Type.** Bodoni Moda for titles and verdict statements, whose high-contrast strokes echo the punched
+mu. Cormorant Garamond for the wordmark only, always letterspaced 0.16em, never body copy. IBM Plex Sans
+for body and UI, IBM Plex Mono for every number, metric, dataset and command.
+
+**Colour.** Ink `#14181A` and a cool neutral ground carry almost every surface. Bronze `#2C5449` marks
+one thing per view: the reading that matters. The verdict colours, brass `#9A7B3F` for under-powered
+evidence and oxblood `#7B2B33` for a fired stop rule, are reserved for pre-registered outcomes and are
+never used decoratively.
+
+**Don't.** No metallic gradient, no emboss, no drop shadow: it is a flat strike, not a rendering of
+silver. Do not set the wordmark without its tracking, and do not use the mu on its own as a letter.
+
+## One note on the mu
+
+Bodoni Moda ships no U+039C, so a live Greek mu falls back to Georgia and the mark loses the
+high-contrast strokes the type spec describes. Greek capital mu and Latin M are the same letterform, so
+the outlines here are drawn from Bodoni's M at a size that reproduces the cap height of the original,
+which keeps the intended typeface and the intended proportions at once.