Public artifacts · measured builds

Proof, not theatre. Open for inspection.

Ariadne is new. There are no client logos on this page, because there are no clients yet whose results I am able to publish. Inventing them would contradict everything else on this site.

What exists instead is inspectable. Below is public source code, an arXiv preprint, and a dated lab-audit summary for this website. As client engagements complete, each will appear here in the same structure: problem, baseline, system built, result, and the time window it took.

Public source code01

AriadneXAI: an ambiguous research question turned into a repeatable pipeline

Before

Did an AI agent’s stated reasoning actually matter to its answer? A qualitative question with no test.

After

A Python workflow: capture reasoning → apply controlled intervention → rerun → compare → score → export.

This is the same discipline a growth system needs: define the question, instrument the process, run it repeatably, and export results someone else can check. The domain was AI safety. The method is what you are buying.

  • Python
  • LangGraph
  • Local LLMs
  • Structured data
  • Automation
View the source Read the arXiv preprint
Previous-release self-audit · lab data, not field data02

A dated website baseline, published with its limitations

Before

Canonical, robots, and sitemap referenced a former domain. The inquiry form depended on the visitor’s email client. No named founder or inspectable proof on the page.

After

That release aligned crawl signals, added a validated same-site endpoint, and prerendered the core content. The site has expanded since, so the receipt is a historical baseline rather than a claim about this build.

This is a compact audit summary, not a raw Lighthouse report. It records Lighthouse 13.4.1 mobile lab results from a local production build on August 4, 2026, before the current site expansion. Lab scores vary by machine, network, content, and deployment; they are not field-user results or current-release validation.

  • Performance 98
  • Accessibility 100
  • Best practices 100
  • SEO 100
Open the audit receipt
Founding engagement option

Early client work can become the first measured case study.

Where scope, measurement access, and written publication permission make it appropriate, a founding engagement can include a reduced project fee in exchange for publishing the baseline, build, and measured outcome. Availability and terms are confirmed in the proposal; publication is never assumed.

Every case study here will follow the same structure: problem → baseline → system built → result → time window, with the measurement method and date stated, and anything the evidence does not support labelled as such.

How evidence is labelled

Every claim on this site carries its evidence class.

  • Public source code Anyone can read it and check what it does.
  • Self-audit — lab data Measured by me, in a lab environment, with tool version and date. Not field-user data.
  • Illustrative scenario A composite of common patterns. Not a client engagement, and not a result.
  • Client engagement — published with permission A real project, with a stated baseline, measurement method, and time window. None of these exist yet.

Your practical starting point

Interested in a measured founding engagement?

Start with an audit. If the scope can be measured properly and you are comfortable considering publication, the proposal can set out the option without making permission a condition of ordinary project work.

Request a growth system audit