Amken LLC | Effective March 2026
At Amken, AI is a force multiplier for an experienced engineer, not a replacement for one. I am a solo founder with 20+ years of systems engineering experience, a graduate background in electronics engineering, and hands-on leadership of large-scale engineering programs. AI accelerates my execution. It does not drive it.
I am transparent about how I use AI because I believe the community deserves to know, and because I think my approach is worth explaining on its merits.
I use Claude by Anthropic as my primary AI assistant for software and firmware development.
AI assists me with code generation, scaffolding, and implementation of software tools, desktop applications, and firmware for microcontrollers. This includes:
- Generating boilerplate and scaffolding from a well-defined architecture
- Implementing modules where the design and interfaces have already been specified
- Writing and refining documentation and README files
- Debugging and iterating on errors during development
AI assists with drafting READMEs, API documentation, and technical write-ups. All documentation is reviewed and edited by me to ensure it accurately reflects the actual implementation and my intent.
AI-generated code does not ship without going through my engineering process. Here is how I actually work:
-
Architecture first. Before AI touches anything, I design the system architecture. This includes module boundaries, data flow, interfaces, state management, and constraints. AI does not make architectural decisions — though I may use it to surface examples, suggestions, and help identify blind spots.
-
Specification-driven prompting. I prompt AI with precise, well-scoped specifications. Vague prompts produce vague code. Every prompt reflects a decision I have deliberately made.
-
Review every line. AI-generated code is read, understood, and evaluated by me before it is accepted. I do not merge code I do not understand.
-
Test and verify. Generated code is compiled, run, and tested against real hardware or environments. I do not assume correctness.
-
Iterate with judgment. When AI produces errors or suboptimal solutions, I diagnose the root cause myself and guide the correction. I do not blindly retry prompts.
-
Domain knowledge governs. My embedded systems, motion control, hardware domain expertise, and systems engineering experience is what catches what AI gets wrong. That expertise is not replaceable and is not outsourced.
These are absolute. No exceptions.
| Domain | Policy |
|---|---|
| PCB & hardware design | AI is never used. Schematics, layout, component selection, signal integrity, and power design are done entirely by me. |
| Safety-critical firmware | Motion control limits, fault detection, and any code path that can damage hardware or cause physical harm is human-written and reviewed by me. |
| Hardware schematics | Never AI-generated or AI-modified. |
| Security & cryptographic code | Never delegated to AI. Any security-sensitive implementation is written and audited by me. |
Projects in the Amken GitHub organization that involve AI-assisted development carry a notice in their README stating that AI was used and in what capacity.
The presence of AI assistance does not diminish the engineering behind a project. My hardware products — the Crea8 motion controller, TinyFOC ESC, and pick-and-place machine — are designed entirely by me without the use of AI. The software tools built around them may use AI assistance in implementation, but the systems they support do not.
To sum it up in one word: transparency. I stand behind my products, my engineering, and my commitment to doing right by the communities I am part of. Publishing this policy is not a legal requirement — it is a reflection of who I am.
Amken LLC — Springfield, VA USA
github.com/amken3d