AI Use Statement

This workshop is about accountability, autonomy, and transparency in GenAI-augmented work. It would be inconsistent not to document how AI was used to build the very website that hosts these ideas.

The process

This website was built on March 30–31, 2026 by Hauke Sandhaus (web chair) using Claude Code (Anthropic’s AI coding assistant, model: Claude Sonnet 4.6), operating within the workshop proposal repository on GitHub.

The workflow was explicitly “vibe coding”—the category of practice this workshop critically examines. Hauke provided high-level direction: page structure, content requirements, design preferences, and transparency requirements. Claude Code generated the HTML, CSS, and page content. Hauke reviewed, guided, and refined the output at each step.

What AI contributed

• Translating the LaTeX workshop proposal into structured HTML content
• Generating all HTML pages (index, proposal, organizers, cfp, papers, this page)
• Designing the CSS design system (typography, color palette, layout)
• Drafting the arXiv LaTeX header template and submission instructions
• Structuring the AI Disclosure Statement framework for the CFP page

What Hauke contributed

• All intellectual content, decisions, and framing (the workshop itself)
• Direction on which pages to include and what they should contain
• Design brief: editorial/academic aesthetic, clean but distinctive
• Review and approval of all generated content
• The decision to document this process transparently on this page
• All meeting notes, organizer contacts, and workshop details

Documentation trail

This build is documented in three ways, consistent with good accountability practice:

• Git commit history in the website repository records every change. The initial commit reflects the AI-generated scaffold; subsequent commits reflect human revisions.
• AI-Input/ folder in the proposal repository contains meeting notes and context used as input to the AI.
• This page provides a plain-language account of the process.

“We asked participants to move beyond simple disclosure toward a deep reflection of agency, accountability, and co-thinking. We hold ourselves to the same standard.”

Reflection

Building this site with AI was fast—roughly two hours from blank directory to deployable website. The AI handled the mechanical work of translating structured content into HTML and CSS competently. The design decisions required direction; left to its own defaults, the AI would have produced a more generic result. The content required verification against the original proposal PDF.

The accountability gap here is real: a careless operator could deploy AI-generated content without reading it. We did not do that—but the risk exists, and it is exactly the kind of risk this workshop invites people to examine in their own practice.

Tools used

• Claude Code (Claude Sonnet 4.6) — code and content generation
• pbakaus/impeccable design skill — design guidelines (typography, color, layout anti-patterns)
• GitHub Pages — hosting
• Google Fonts (Cormorant Garamond, DM Sans) — typography

The workshop proposal itself notes that disclosure is a necessary first step but insufficient on its own. This page is our attempt to go further: not just stating that AI was used, but documenting how, what it did, and what we retained control over.

The meeting notes (March 23, 2026) record that the organizing committee discussed “vibe coding” the website, with Pooja raising the value of disclosure. This page is the follow-through on that conversation.

We invite workshop participants to hold us accountable to the same standards we ask of them.