Blue Lumen / Company & practical work
From a question
to a checked result.
We bring hands-on AI workflow experience to our training and consulting. Here are concrete examples of the inputs, outputs and review steps in that work.
01 / Company
Blue Lumen AB.
A Swedish company with two equally important core areas: AI and AI-agent training and consulting, and patent/IP consulting.
- Legal name
- Blue Lumen AB
- Organisation no.
- 559584-6212
- Incorporated
- 5 May 2026
- Based in
- Stockholm, Sweden
- Co-founder
- Yulin Cui
Public company record: Blue Lumen AB on Eniro ↗. The company registration certificate is available for an agreed verification process.
02 / Practical examples
Workflows you
can follow.
Our co-founder built and actively uses Personal AI OS for engineering, research and day-to-day business tasks. The examples below show that experience using our own website and public documentation.
Engineering / Completed 7 October 2026
Blue Lumen website delivery
- Input
- Update our website to describe AI and IP as equally important core areas and show practical AI workflow experience.
- Process
- Read the confirmed business brief
- Make scoped website changes with an AI agent
- Review content and build the release
- Publish and inspect the public website
- Output
- A published website with two service areas, three workflow summaries and consistent company identity.
- Review
- Content was checked against company facts; public files, links, and desktop/mobile views were verified.
An extract from the published output
Applied AI and AI-agent training and consulting.
Patent and IP consulting.
Two equally important core areas.
View the website ↗Research / Demonstration prepared 8 October 2026
A source-linked technical brief
- Input
- Read the official documentation and produce a short deployment note: how static files are served, where custom response headers apply, and how the deployed site should be checked.
- Process
- Read official sources, connect each factual statement to its source, and review the resulting brief.
- Output
- The short, source-linked note below.
- Review
- Claims were checked against the linked documentation; the review decision is labelled separately from source facts.
Demonstration output
Documented fact · S1A matching static asset can be served directly without invoking Worker code.
Documented fact · S2An _headers file defines custom headers for static asset responses. Worker-generated responses need their own headers.
Review decision · Workflow reviewCheck the actual public response after deployment, including the page content and expected headers.
Read the structured example data ↗ · Prepared 2026-10-08. The technical brief was prepared with Codex assistance and checked against the linked sources. These examples describe the founder’s working system and Blue Lumen’s own website delivery.
03 / Development direction
Evaluate, then
apply with care.
We plan to evaluate the Claude API for non-confidential document analysis and structured research, then develop reusable demonstrations for our B2B training and consulting.
The evaluation will cover source grounding, output quality and cost, with human review. This describes our planned Claude evaluation; the examples above show our existing workflow experience.
Discuss an AI workflow ↗