A file lock waits out Windows removing the previous holder's lock; the site's header fits one row on a phone, the tools page draws its claims, and the README opens on the wordmark

- On Windows a lock directory being removed answers access denied, not already exists, so a
  second writer, such as two MCP servers signing in at once, failed instead of waiting.
- The header is Docs, Tools and the GitHub mark, one row down to 320px; /themes folds into the
  homepage's gallery and redirects there.
- The tools page has a terminal of measured rows beside its intro and a chart of whole-task cost
  against Luxe's 1x beside its table; Pi leaves the comparison.
- The README opens on a generated wordmark banner that carries its own colors.
This commit is contained in:
Evan Whittaker
2026-10-17 16:58:08 -05:00
parent 846a413a4d
commit 3c69a90b06
15 changed files with 534 additions and 420 deletions

View File

@@ -1,182 +0,0 @@
---
import Base from "../layouts/Base.astro";
import ThemePreview from "../components/ThemePreview.astro";
import themeCss from "../assets/generated/tokens.css?raw";
/**
* The palette, with its numbers attached.
*
* Swatches are read out of the generated stylesheet rather than typed here, so this page cannot
* disagree with the theme the terminal loads.
*/
const readTokens = (block: string) =>
[...block.matchAll(/--luxe-([a-z-]+):\s*(#[0-9a-f]{6});/g)].map(([, name, hex]) => ({ name, hex }));
const modes = [
{ label: "luxe", note: "warm graphite, champagne brass", tokens: readTokens(themeCss) },
];
/** Relative luminance, so each chip can print its own contrast rather than claim one. */
function contrast(a: string, b: string) {
const lum = (hex: string) => {
const parts = [1, 3, 5].map((i) => parseInt(hex.slice(i, i + 2), 16) / 255);
const [r, g, bl] = parts.map((c) => (c <= 0.04045 ? c / 12.92 : ((c + 0.055) / 1.055) ** 2.4));
return 0.2126 * r + 0.7152 * g + 0.0722 * bl;
};
const [hi, lo] = [lum(a), lum(b)].sort((x, y) => y - x);
return (hi + 0.05) / (lo + 0.05);
}
---
<Base title="Themes — Luxe" description="Luxe's palette, and the community favorites it ships with.">
<div class="page">
<section class="section section--head">
<p class="eyebrow">Themes</p>
<h1>Color that means something.</h1>
<div class="prose">
<p>
Luxe's own palette is warm graphite with a champagne-brass accent, and its hues come from a
set of paint colors rather than a terminal convention. Those are mid-luminance by nature,
so the hues were kept and the lightness re-derived against a contrast floor.
</p>
<p>
Three rules do the work, and each one is asserted in the test suite rather than eyeballed:
every background and gray shares one neutral hue, every accent sits on one lightness
plateau so none of them shouts, and nothing drops below 4.5:1 on its own background.
</p>
</div>
</section>
{
modes.map(({ label, note, tokens }) => {
const background = tokens.find((t) => t.name === "background")!.hex;
return (
<section class="section">
<div class="theme__head">
<h2 class="theme__name">{label}</h2>
<p class="muted theme__note">{note}</p>
</div>
<ul class="swatches">
{tokens
.filter(
(t) =>
!t.name.startsWith("wordmark") &&
!t.name.startsWith("capture") &&
t.name !== "tile-ink",
)
.map(({ name, hex }) => (
<li class="swatch">
<span class="swatch__chip" style={`background:${hex}`} />
<span class="swatch__name">{name}</span>
<span class="swatch__hex">{hex}</span>
<span class="swatch__ratio">
{name === "background" ? "—" : `${contrast(hex, background).toFixed(1)}:1`}
</span>
</li>
))}
</ul>
</section>
);
})
}
<section class="section section--last">
<h2>And plenty you may already use.</h2>
<div class="prose">
<p>
Ports that stay true to their originals. Luxe only steps in where a color is too faint to
read a word in, or where a theme's "red" isn't red, since a failed test shown in green
says the opposite of what happened.
</p>
<p>
Point at one to see it. That's how <code>/theme</code> works in the terminal too: the
cursor moves and the whole app repaints, because a palette isn't something you can judge
from its name.
</p>
</div>
<ThemePreview swatches />
<p class="muted">
Set one with <code>"theme": "nord"</code> in <code>~/.config/luxe/config.json</code>, or
override individual tokens with a <code>colors</code> map. On Omarchy, Luxe wears your
desktop's theme and changes with it when you switch.
</p>
</section>
</div>
</Base>
<style>
.section--head {
border-top: 0;
padding-top: var(--space-6);
}
.section--last {
border-bottom: 0;
}
.prose {
max-width: var(--measure);
margin-top: var(--space-4);
}
.theme__head {
display: flex;
align-items: baseline;
gap: var(--space-3);
flex-wrap: wrap;
}
.theme__name {
font-family: var(--font-mono);
font-size: var(--step-1);
letter-spacing: 0;
}
.theme__note {
margin: 0;
font-size: var(--step--1);
}
.swatches {
list-style: none;
padding: 0;
margin: var(--space-4) 0 0;
display: grid;
gap: var(--space-2) var(--space-4);
grid-template-columns: repeat(auto-fill, minmax(17rem, 1fr));
}
.swatch {
display: grid;
grid-template-columns: 1.6rem 1fr auto auto;
align-items: center;
gap: var(--space-2);
font-size: var(--step--1);
padding-block: 0.3rem;
border-bottom: 1px solid var(--luxe-border-muted);
}
.swatch__chip {
width: 1.6rem;
height: 1.6rem;
box-shadow: inset 0 0 0 1px color-mix(in oklab, var(--luxe-text) 18%, transparent);
}
.swatch__name {
color: var(--luxe-text);
}
.swatch__hex,
.swatch__ratio {
color: var(--luxe-muted);
font-variant-numeric: tabular-nums;
}
.swatch__ratio {
min-width: 4ch;
text-align: right;
}
</style>

View File

@@ -223,7 +223,7 @@ const rows = new Map(
[...bench.repositories].reverse().flatMap((repo) => repo.rows.map((row) => [row.id, row])),
);
for (const row of primary.rows) rows.set(row.id, row);
/** The tightest of the four other agents' limits on the same output: Codex's. */
/** The tightest of the other agents' limits on the same output: Codex's. */
const tightest = (row: { forwarded: Record<string, number> }) =>
Math.min(...Object.values(row.forwarded));
@@ -240,9 +240,13 @@ const tokens = (bytes: number) => {
const count = bytes / 4;
if (count < 1000) return `${Math.max(1, Math.round(count / 10) * 10)} tokens`;
if (count < 10_000) return `${(count / 1000).toFixed(1)}k tokens`;
return `${Math.round(count / 1000)}k tokens`;
if (count < 1_000_000) return `${Math.round(count / 1000)}k tokens`;
return `${(count / 1_000_000).toFixed(1)}M tokens`;
};
/** The head's terminal: the commands whose output Luxe's tool cuts the most, measured like the rest. */
const terminal = ["git", "module_report", "glob"].flatMap((id) => rows.get(id) ?? []);
const times = (shell: number, tool: number) => {
const ratio = shell / tool;
if (ratio < 1.1) return "the same";
@@ -262,12 +266,24 @@ const versus = (cut: number, tool: number) => {
*/
type Other = { warm: number; cold: number; passed: number; runs: number };
type CostRow = { model: string; luxe: { passed: number; runs: number } } & Partial<
Record<"claude" | "codex" | "opencode" | "pi", Other>
Record<"claude" | "codex" | "opencode", Other>
>;
const costRows = agentCost.rows as CostRow[];
const vendorAgent = (row: CostRow) => row.claude ?? row.codex;
const multiple = (other?: Other) => (other ? `${other.cold.toFixed(2)}×` : "–");
/** The chart's scale: the costliest multiple, rounded up to a half, so 1× (Luxe) sits at a fixed mark. */
const chartScale =
Math.ceil(
Math.max(...costRows.flatMap((row) => [vendorAgent(row)?.cold ?? 0, row.opencode?.cold ?? 0])) * 2,
) / 2;
const chartWidth = (multiple: number) => `${(multiple / chartScale) * 100}%`;
const chartBars = (row: CostRow) =>
[
{ kind: "vendor", other: vendorAgent(row) },
{ kind: "opencode", other: row.opencode },
].flatMap(({ kind, other }) => (other ? [{ kind, cold: other.cold }] : []));
/** A bar's share of its row, never so thin it disappears. */
const share = (shell: number, tool: number) => `${Math.max((tool / shell) * 100, 0.6)}%`;
---
@@ -278,6 +294,8 @@ const share = (shell: number, tool: number) => `${Math.max((tool / shell) * 100,
>
<div class="page">
<section class="section section--head">
<div class="split">
<div>
<p class="eyebrow">Tools</p>
<h1>Better than the shell.</h1>
<div class="prose">
@@ -294,33 +312,61 @@ const share = (shell: number, tool: number) => `${Math.max((tool / shell) * 100,
Measured in a checkout of {primary.name} at <code>{primary.commit}</code>, the codebase
the harness eval runs in, with its virtualenv in place. The shell row is the command's
raw output; <em>tightest cut</em> is the least any other agent sends its model of it,
which is Codex's 10 KB, head and tail kept. Pi and OpenCode keep up to 50 KB and Claude
Code 30,000 characters. A token is about four bytes.
which is Codex's 10 KB, head and tail kept. OpenCode keeps up to 50 KB and Claude Code
30,000 characters. A token is about four bytes.
</p>
<p class="muted note">
Those are single calls, and across a whole task the bars shrink, because most calls are
small to begin with. What a whole task costs is <a href="#tasks">below</a>.
</p>
</div>
</div>
{/* The claim, shown: what the model reads back from a shell command, then from the tool */}
<figure class="terminal">
<div class="terminal__bar">
<span class="terminal__dots" aria-hidden="true"><i></i><i></i><i></i></span>
<span>{primary.name.toLowerCase()} @ {primary.commit}</span>
</div>
<dl class="terminal__body">
{
terminal.map((row) => (
<div class="terminal__pair">
<dt>
<span class="terminal__prompt">$</span>
{row.shell}
</dt>
<dd>{tokens(row.shellBytes)}</dd>
<dt class="terminal__luxe">
<span class="terminal__prompt">luxe</span>
{row.tool}
</dt>
<dd class="terminal__luxe">{tokens(row.toolBytes)}</dd>
</div>
))
}
</dl>
<figcaption class="terminal__caption">What the model reads back, measured.</figcaption>
</figure>
</div>
</section>
<section class="section group" id="tasks">
<h2 class="group__name">Whole tasks</h2>
<div class="prose">
<p>
Luxe, Claude Code, Codex, OpenCode and Pi each ran the first twelve Django tasks of
Luxe, Claude Code, Codex and OpenCode each ran the first twelve Django tasks of
SWE-bench Verified Mini, with the same model, the same prompt and the same keys, every
call metered at the network and priced from one list. Each figure is what the other agent
spent on the same task, as a multiple of what Luxe spent: above 1× costs more than Luxe.
</p>
</div>
<div class="split split--tasks">
<table class="costs">
<thead>
<tr>
<th scope="col">Model</th>
<th scope="col">Claude Code or Codex</th>
<th scope="col">OpenCode</th>
<th scope="col">Pi</th>
</tr>
</thead>
<tbody>
@@ -332,19 +378,40 @@ const share = (shell: number, tool: number) => `${Math.max((tool / shell) * 100,
</th>
<td>{multiple(vendorAgent(row))}</td>
<td>{multiple(row.opencode)}</td>
<td>{multiple(row.pi)}</td>
</tr>
))
}
</tbody>
</table>
{/* The table's numbers drawn; hidden from screen readers, which have the table */}
<figure class="chart" aria-hidden="true" style={`--one: ${chartWidth(1)}`}>
<figcaption class="chart__legend">
<span class="chart__key chart__key--vendor">Claude Code or Codex</span>
<span class="chart__key chart__key--opencode">OpenCode</span>
<span class="chart__key chart__key--luxe">Luxe, 1×</span>
</figcaption>
{
costRows.map((row) => (
<div class="chart__group">
<code class="chart__model">{row.model}</code>
{chartBars(row).map(({ kind, cold }) => (
<div class={`chart__row chart__row--${kind}`}>
<span class="chart__track">
<span class="chart__bar" style={`width: ${chartWidth(cold)}`} />
</span>
<span class="chart__value">{cold.toFixed(2)}×</span>
</div>
))}
</div>
))
}
</figure>
</div>
<div class="prose">
<p class="muted note">
Every run is charged as a cold start, the way a session begins after a break. Run side
by side on a warm cache, Claude Code and OpenCode cost up to 2.6× what Luxe does on
Claude models. Pass rates overlapped on every model. Pi is the cheapest by design: four
tools and a short prompt, with web search, MCP and the rest left to extensions. Graded
on one machine rather than in SWE-bench's own containers, so these are not leaderboard
Claude models. Pass rates overlapped on every model. Graded on one machine rather than in SWE-bench's own containers, so these are not leaderboard
scores. Measured at <code>{agentCost.commit}</code>.
</p>
</div>
@@ -537,6 +604,190 @@ const share = (shell: number, tool: number) => `${Math.max((tool / shell) * 100,
margin-bottom: var(--space-2) !important;
}
/* The words, and beside them what they claim, drawn; one column once there is no room. Shares
rather than a fixed text column: the prose keeps its own measure, and the drawing gets room for
its longest command just above the breakpoint */
.split {
display: grid;
grid-template-columns: minmax(0, 1.2fr) minmax(0, 1fr);
gap: var(--space-5) var(--space-6);
align-items: center;
}
.split--tasks {
grid-template-columns: auto minmax(0, 1fr);
align-items: start;
margin-top: var(--space-5);
}
.split--tasks .costs {
margin-top: 0;
}
/* === The head's terminal === */
.terminal {
margin: 0;
background: var(--luxe-user-msg-bg);
border: var(--border);
font-size: var(--step--1);
}
.terminal__bar {
display: flex;
align-items: center;
gap: var(--space-3);
padding: var(--space-2) var(--space-3);
border-bottom: var(--border);
color: var(--luxe-dim);
}
.terminal__dots {
display: inline-flex;
gap: 0.4em;
}
.terminal__dots i {
display: inline-block;
width: 0.6em;
height: 0.6em;
border-radius: 50%;
background: var(--luxe-dim);
}
.terminal__body {
margin: 0;
padding: var(--space-3);
display: grid;
gap: var(--space-3);
}
.terminal__pair {
display: grid;
grid-template-columns: minmax(0, 1fr) auto;
gap: var(--space-1) var(--space-3);
}
.terminal__pair dt {
color: var(--luxe-muted);
overflow: hidden;
text-overflow: ellipsis;
white-space: nowrap;
}
.terminal__pair dd {
margin: 0;
text-align: right;
color: var(--luxe-muted);
font-variant-numeric: tabular-nums;
}
.terminal__prompt {
color: var(--luxe-dim);
margin-right: 0.6em;
}
.terminal__pair .terminal__luxe {
color: var(--luxe-accent);
}
.terminal__caption {
padding: 0 var(--space-3) var(--space-3);
color: var(--luxe-dim);
font-size: var(--step--2);
}
/* === The whole-task chart === */
.chart {
margin: 0;
display: grid;
gap: var(--space-3);
font-size: var(--step--1);
}
.chart__legend {
display: flex;
flex-wrap: wrap;
gap: var(--space-2) var(--space-4);
color: var(--luxe-muted);
}
.chart__key::before {
content: "";
display: inline-block;
width: 0.8em;
height: 0.8em;
margin-right: 0.5em;
vertical-align: -0.05em;
}
.chart__key--vendor::before {
background: var(--luxe-dim);
}
.chart__key--opencode::before {
background: var(--luxe-muted);
}
.chart__key--luxe::before {
width: 2px;
height: 1em;
background: var(--luxe-accent);
}
.chart__group {
display: grid;
gap: 0.3em;
}
.chart__model {
color: var(--luxe-accent);
}
.chart__row {
display: grid;
grid-template-columns: minmax(0, 1fr) 4rem;
align-items: center;
gap: var(--space-3);
}
.chart__track {
position: relative;
height: 0.9rem;
background: var(--luxe-user-msg-bg);
}
/* Luxe's 1×, drawn through every row: each bar is read against it */
.chart__track::after {
content: "";
position: absolute;
top: -0.2rem;
bottom: -0.2rem;
left: var(--one);
width: 2px;
background: var(--luxe-accent);
}
.chart__bar {
display: block;
height: 100%;
}
.chart__row--vendor .chart__bar {
background: var(--luxe-dim);
}
.chart__row--opencode .chart__bar {
background: var(--luxe-muted);
}
.chart__value {
text-align: right;
color: var(--luxe-text);
font-variant-numeric: tabular-nums;
}
/* === Measured === */
.bench {
@@ -678,7 +929,9 @@ const share = (shell: number, tool: number) => `${Math.max((tool / shell) * 100,
}
@media (max-width: 68rem) {
.tool {
.tool,
.split,
.split--tasks {
grid-template-columns: minmax(0, 1fr);
}
}