Nuxsol
Start a conversation

Worknuxsol.eu/work/this-site/

This site.

Nuxsol’s own site. A site for a new company that had two kinds of readers from day one.

The home page of this site: the headline, the terminal with llms.txt and the reader switch.
The home page, as a person reads it.
Sector
Nuxsol’s own site
Year
2026
Project
Nuxsol
Client
Nuxsol
Design and development
Nuxsol

This site was built to prove the studio’s own claim: that one page can serve a person and a machine with the same words.

In the project

  • Every page rendered on the server
  • A Markdown copy of every page
  • llms.txt, robots.txt and a sitemap
  • Structured data on every page
  • Search palette and machine view

The case

The challenge

A new company has two kinds of readers from day one: people deciding whether to call, and the AI systems those people ask first. Most agency sites are built only for the first. They reach a crawler as a script and an empty page.

A studio that sells readable websites cannot have one of those. Every claim on the service pages had to be something a visitor could check on the site itself.

What we built

Every page is rendered on the server, so a crawler receives the same words a person reads. The crawler policy names each search and answer engine. The company, its services and its articles are described in structured data. Every page has a Markdown copy at a predictable address, and llms.txt points to them.

The machine layer is part of the design. The reader switch in the header shows the same content both ways.

What it runs on

Astro renders the pages on the server. GSAP and Lenis handle motion, on transform and opacity only. The site was built with Claude.

The machine view of this site: the same page as Markdown.
The same page as a machine receives it: Markdown, at a predictable address.

Who it is for

  • The person deciding whether to callNeeds to see what the studio does, and how well, in one visit.
  • The crawlerReceives HTML and nothing else. Needs the words to be in it.
  • The assistant answering a buyerNeeds a short description it can quote, and a clean copy of each page.

What we did

By service: what was used in this project, and the steps of execution.

Digital design

Every page has a job.

Websites and digital experiences built to convert. We design with a purpose, and every journey leads a visitor towards becoming a customer.

  • Web design
  1. 01

    Set the two readers

    Every page is designed for a person and for a machine.

  2. 02

    Design the scenes

    Hero, case studies, services, approach, offers and notes.

  3. 03

    Draw the visuals in code

    Interface drawings and case artwork, with no stock imagery.

Development

The design, built exactly.

Clean code that brings the design to life: fast, responsive and exact on every device, with a back end you can rely on.

  • Front-end
  • Astro
  • Marketing sites
  • Performance
  1. 01

    Render on the server

    Astro builds every page as HTML.

  2. 02

    Generate the machine copies

    Markdown, llms.txt, the sitemap and structured data come from the same data as the pages.

  3. 03

    Add motion last

    GSAP and Lenis, on transform and opacity only, with a reduced-motion state.

  4. 04

    Measure

    Speed, accessibility and layout are checked on the production build.

Marketing & growth

Creative that is measured by what it moves.

Strategic creative that drives results, from campaigns to pitch decks. It includes being found: SEO, GEO, AIO and websites that agents and machines can read.

  • SEO, GEO and AIO
  • Agent-ready websites
  • Machine-readable websites
  1. 01

    Name the crawlers

    Each search and answer engine has its own rule.

  2. 02

    Describe the company

    Structured data on every page.

  3. 03

    Publish the copies

    A Markdown address for every page, listed in llms.txt.

Design decisions

Each with the reason for it, and what we chose not to do.

  1. 01

    Every page is rendered on the server.

    A crawler that does not run scripts still receives the full text. The page a person sees and the response a machine gets hold the same words.

    Instead ofAn app rendered in the browser, with a prerender service in front.

  2. 02

    A Markdown copy of every page.

    Each page has an address ending in .md, generated from the same data as the designed page, so the two cannot drift apart.

    Instead ofA separate text version written by hand.

  3. 03

    The machine layer is shown, not hidden.

    The reader switch in the header and the terminal in the hero put the Markdown, the schema and the crawler rules in front of visitors.

    Instead ofKeeping structured data and llms.txt as invisible plumbing.

  4. 04

    Counted, not claimed.

    The figures on the home page are counted each time the site is built: pages, Markdown copies, schema types and named crawlers.

    Instead ofRounded marketing numbers.

The case study cards on the home page of this site.
The case studies as large cards.
The services section of the home page of this site.
Five services as rows, each with an interface drawn in code.

The system

Four tones, each with one job. Contrast is measured, not judged by eye.

  • Ground#050505Near-black: every scene.
  • Ink#FFFFFFWhite: all text.
  • Signal#E7FE65Lime: the mark, one word in the headline, arrow boxes and status dots.
  • Soft#101010A lifted black: panels and cards.
Contrast ratios of the palette
PairContrastPasses
Ink on ground20.4 : 1AAA
Ink on soft19.0 : 1AAA
Label on signal16.0 : 1AAA
Signal on ground18.2 : 1AAA
Type
Geist for everything people read. Geist Mono for addresses, file names and Markdown.
Voice
Plain sentences. Numbers with a source and a date.

How it works

  1. 01

    Rendered on the server

    A crawler receives the same words a person reads.

  2. 02

    Named crawlers

    The crawler policy names each search and answer engine.

  3. 03

    Structured data

    The company, its services and its articles are described in a form machines read.

  4. 04

    Markdown copies

    Every page has one at a predictable address, and llms.txt points to them.

The principle

The same words for a person and for a machine.

A service page of this site with its Markdown strip.
An inner page: the title for people, and the first lines of its Markdown beside it.

What it is built to change

  • Build countsPages rendered on the server, Markdown copies, schema types and named crawlers, counted at every build and shown on the home page.
  • The same words both waysOpen any page, switch to the machine view, and compare.
Next case A quoting desk for a freight forwarder.

Show us the page you want machines to understand.

Send us one URL. On the first call we tell you, for free, what a person and an agent each find there, and roughly what it would take to fix.