A long overdue update

It's been a while since my last post. We had to move from Seattle to rural Mississippi, right in the heart of the Mississippi Delta. I'm still unemployed, but we've moved into a generational family home, so we're doing okay.

The job market for software engineering professionals is tough. On top of that, rural Mississippi doesn't have much demand for my particular background. The economy here is centered around agriculture, and practical trade skills are especially valuable. After more than two decades in software, I've discovered that my most marketable technology skill around here might be connecting an iPad to a printer.

I've been focusing on updating a workshop I inherited from my father a few years back.

Since taking that photo, I've cleaned all that stuff out. My father hoarded just about everything. I donated a lot of equipment to a local charity and burned several truckloads of scrap wood. I've also had to run off a few squatters:

That clean-up took me about two months! Since then, I've had A/C installed, and I've learned how to run electrical so I could install a bunch of new woodworking equipment that runs on 240V. Oh, and I now operate a tractor several times a week.

What about Mnemonic?

I haven't forgotten about my baby, Mnemonic. I have, though, put off the work to get it running in the cloud for now. It's a cost issue for me, still being unemployed and all.

But I've started actively using Mnemonic again to design some custom software to run my woodshop. I've been using Codex CLI for this instead of Claude Code. I've found Codex to be much better at helping me flesh out ideas, create designs, and capture my requirements. That said, I think I'm going to write it by hand for now. I don't want to lose my "edge". I'd use Claude Code, but I had to cancel my subscription for now. Claude is just too expensive while I'm unemployed.

I'll use Codex to help with test writing and code reviews. I'm also thinking about updating gralph to be configurable so it can use Codex, or any other AI coding assistant for that matter. I'd really love to use gralph with a local LLM, like Qwen 3.8 running on an Nvidia DGX Spark using Mnemonic. But, alas, that costs a LOT of money, so I'll need gainful employment before that fantasy is made real.

What does the future hold?

I don't know. AI has flipped the script for software development. Is it a bubble that's going to burst? Yeah, sooner or later. Will the issues like context rot, the lost-in-the-middle problem, and hallucinations ever be resolved? Yeah, probably. Eventually. Right now, we're flush with AI slop everywhere, but that isn't AI's fault. That's a user issue. AI is a shiny new hammer and right now everything is a nail. Will software development ever go back to the way it was? No, I don't think so, but I'm okay with that. I know how—and when—to use that AI hammer. And when I swing it, I'll strike that nail squarely on the head 100% of the time!