008 Alexander Zeitler - Running a Manufacturing Company on Omarchy & AI Agents

What happens when a manufacturing company stops accepting vendor lock-in and rebuilds its workflow around open tools, AI agents, and event sourcing?

Basti talks with Alexander Zeitler, a software veteran who runs a 10-person manufacturing business on Linux and Omarchy. After discovering he couldn't access his own data in a legacy ERP, he used Claude to extract it and started replacing siloed workflows with a system he actually controls.

In this episode:
• Why "owning" your data means nothing if you can't use it
• FreeCAD, OpenSCAD, Python and Rails in real manufacturing workflows
• Building products from parameters, not proprietary CAD
• Generating decision records, risk assessments and docs from reusable blocks
• GitHub issues as event streams, with agents that reason before they code
• A second AI review layer for every code change
• Why event sourcing matters more in an AI-assisted world

🔗🔗🔗 LINKS 🔗🔗🔗

Basti
https://basti.net
https://x.com/bastidotnet
 @bastidotnet 

Omarchy Stories
https://omarchystories.org
https://x.com/omarchystories
 @OmarchyStories 

Alexander Zeitler
https://www.waltergmbh.de/
https://x.com/lxztlr

⏱️⏱️⏱️ CHAPTERS ⏱️⏱️⏱️

00:00 Intro to Alexander Zeitler
00:32 From childhood programming to web development
02:20 Business software, not private computing
05:26 Why the company moved away from legacy software
08:20 Using Claude to extract data from an inaccessible database
10:24 Data ownership and control
12:18 Why Omarchy fits a changing engineering workflow
14:16 Rethinking CAD as parameters, math and scripting
17:10 The limits of proprietary collaboration tools
18:09 Building a full process from ideation to shipping
22:56 Documentation, compliance and safety requirements
24:46 From Markdown to a structured document system
26:10 Component-based documentation generation
27:10 Learning Rails while reviewing agent output
28:10 Why event sourcing beats CRUD
33:16 AI memory, speed and context loss
35:31 GitHub issues as a controlled agent workflow
39:10 A second agent for code review
40:27 What new engineers still need to learn
42:26 Domain-driven design and event modeling basics
45:29 End-to-end thinking and edge cases
47:54 Event sourcing the agent workflow itself
48:53 Lower stress, more parallelism, better reasoning
50:43 Closing thoughts

"If I can't access the data in my own company, I cannot make decisions." – Alexander Zeitler