AI Means Opportunity: Huge Improvement Gains in DIY Manufacturing

improvement

This weekend, I’ve created 4 web applications using Claude.ai that are going to help me realize serious improvement gains in the manufacturing of a 7-band HF SSB Single-Conversion Superhet. The design, build, and test aspects of each component is the fun part. The difficulty comes when fighting gaps in cash flow and also waiting for materials or PCBs. Once components are identified, I normally do with what I have and work on what makes the most sense at the time. These applications help me see the state of things in a more informed way. When I have available time for this project, I will always have something to do even if I’m waiting on parts or PCBs.

PartBin: Inventory Improvement and Awareness

PartBin

PartBin is the first application I created with the help of Claude.ai. I gave a clear paragraph of how I wanted an app that could record the individual parts I order and also track quantities (may be left over) and associate the parts with a specific modular component. This application uses local-storage (session) and also provided export/import of JSON files.

The application is visually beautiful and this was also part of my prompting of Claude.ai. AI creates a single page HTML you can download. The web app was a serious improvement over sifting through cardboard boxes from Mouser.com. I started adding parts and exported my first backup JSON almost immediately after sent the index.html from my Mac Mini back to my bench machine (Intel NUC w/Ubuntu). I was floored with the HTML Claude.ai produced in about 3 minutes (180 seconds). I told my wife how excited I was. “That’s nice.”, she said.

Component: Modular Architecture Improvement, Less the Diagram

improvement

When I began the Freedom7 project, I knew that the first radio wouldn’t be consumer-ready. It’s architecture would be like a component diagram where components were modular, i.e. the blocks defining the single-conversion SSB transceiver would need to stand on their own electronically. These components could be designed, built, and tested individually. The components could be improved over time. Enter the Component web app.

I first created a functional requirements text file with clear specifics. I also provided data points that were important. I did another paragraph-sized prompt, describing what I wanted and keeping the style like PartBin. I attached the requirements.txt file and let Claude.ai do its thing. About 5 min later, it showed me the results of its effort. Beautiful!

This was a second, immediate improvement. I now had a place for all components in a grid, visual status, and quick understanding of my work, very similar to a dashboard.

I now have this app loaded with components and details for each. What you see in the image, is dummy data for the initial HTML creation. Please remember that AI remembers working with me and what I’m actually doing. This is paramount in pleasing it’s consumers.

Projex: Task Management Improvement

improvement

Projex was the next to the last app I created, but my last app actually improves a dream and maybe not reality currently. Projex came about as something similar to Atlassian JIRA but my creation and one that I can refactor as I use it to track tasks across various projects.

Again, the information in the image was generated by Claude.ai on it’s knowledge of what I’m doing. I prompted the creation of a task tracker but wanted the ability to also manage projects. I need the project and task administration because my time is limited and no one else is currently working with me.

Projex is also my favorite application created this weekend. Like Component, I drafted a requirements.txt to go with the AI prompt. I have reference data in pull-downs but this data may change. Remember that the data can be backed up with an export of JSON.

KitLab: Improvement Toward the Possibility of Sharing Component Kits

improvement

KitLab is not for now, but more of a visualization of what could be. The Low Noise Amplifier you see is an AI generation done with ChatGPT, “Can you make me an image suitable for presentation of a low noise amplifier that I made, image attached, and have it on a semi-dark background?”. ChatGPT did a good job, didn’t it? I made the 20 meter bandpass filter too. The data shown was part of the HTML generation by Claude.ai with the original prompt.

KitLab helps me dream that I am closer to curating and selling RF component kits to people to learn about HF and why I did NOT choose the software-defined radio route. My work specifically chose the single-conversion superheterodyne radio for the fact that a DIY HAM radio enthusiast can actually make one of these at home with a little education.

Follow the Freedom7 Build

I’m building a real HF SSB transceiver from the ground up – no kits, no shortcuts, and no hiding the hard parts.

If you’re interested in how this system comes together over time, follow the full build here:
https://kr4bad.com/category/the-build

Not everything fits into a single build. For additional insights, experiments, and lessons learned:
https://kr4bad.com/field-notes/

Follow the build as it happens:
https://kr4bad.com/?subscribe=1

Looking for some background?
https://kr4bad.com/about/

Comments and discussion are always welcome:
david [at] kr4bad-dot-com

73,
KR4BAD David

Comments

Leave a Reply

Discover more from KR4BAD Amateur Operator

Subscribe now to keep reading and get access to the full archive.

Continue reading