TechLuddite

Gamer at heart. 30 years in IT: hardware→network→server→fleets→security. Learning software development in public. Systems design, pointed at better AI outcomes.

[TechLuddite @gh_TechLuddite]

Work has all of my extra cycles currently, as we're making some exciting changes.

I am making time to watch your feeds, and I'm tell you, light itself cannot keep up with the Omarchy community!

As I said before, I am cutting back on projects, massively. However, my focus on Omarchy and this community is now a staple for me.

That said, it might sounds a bit silly...that I'm launching ANOTHER NEW PROJECT!

This project is for me and my enjoyment, and for my love of @roadtovostok!

Everyone has been going crazy for Jev, but I did not have access. I love the idea of fast decision engines, so I was already excited to jump on the bandwagon.

Because I did not have access, I leaned on prior art from my own repos that had components designed to address similar needs. My first software project, ai-plays-as-ai-making-paperclips, and my first game repo (if you can even call it one), Universal-AI.

Well in the middle of the sparse time I could put into this new project, which I'm calling JevJev, I got access!

What was especially interesting to me is that getting JevJev to do anything in RTV took more work than the on-box decision engine I started with. However, JevJev is kinda rocking and rolling now! I'll be starting an AFK live stream here in a bit on my YT or X, but I'm on my lunch break from work so it's whatever gives me the least hassle.

I won't be around for the stream, but I thought others might enjoy see Jev in action, becoming thee Ironman as JevJev, on the Road to Vostok!

[TechLuddite @gh_TechLuddite]

@1280MHz Labor cost is certainly driving this forward, but I have a client that has, for years, extended hours because keeping the printers going near 24x7 has some serious benefits.

I'd love to hear from you about the event when you're ready. Safe travels!

[TechLuddite @gh_TechLuddite]

If you thought JevJev was cool with Jev under the hood, I cannot wait for you to meet Jeff, powered by this puppy: https://t.co/W4QCa7uigh

I burned my entire ($3) open-code budget, so I had to try something else. I am so stinky glad that happened!

The next stream should be much more entertaining. Jeff has a training arena in Road to Vostok now, and every time we iterate, a workflow plays out the new design over 1000 times headless at break-neck speed, while still giving me a viewport to see how Jeff is doing at 1x speed. All of it feeds back in to Jeff and then the next iteration is kicked off.

Image Jeff as a boss in RTV, just don't ask for his name please! Second AFK stream today, sometime this afternoon!

[TechLuddite @gh_TechLuddite]

Work has all of my extra cycles currently, as we're making some exciting changes.

I am making time to watch your feeds, and I'm tell you, light itself cannot keep up with the Omarchy community!

As I said before, I am cutting back on projects, massively. However, my focus on Omarchy and this community is now a staple for me.

That said, it might sounds a bit silly...that I'm launching ANOTHER NEW PROJECT!

This project is for me and my enjoyment, and for my love of @roadtovostok!

Everyone has been going crazy for Jev, but I did not have access. I love the idea of fast decision engines, so I was already excited to jump on the bandwagon.

Because I did not have access, I leaned on prior art from my own repos that had components designed to address similar needs. My first software project, ai-plays-as-ai-making-paperclips, and my first game repo (if you can even call it one), Universal-AI.

Well in the middle of the sparse time I could put into this new project, which I'm calling JevJev, I got access!

What was especially interesting to me is that getting JevJev to do anything in RTV took more work than the on-box decision engine I started with. However, JevJev is kinda rocking and rolling now! I'll be starting an AFK live stream here in a bit on my YT or X, but I'm on my lunch break from work so it's whatever gives me the least hassle.

I won't be around for the stream, but I thought others might enjoy see Jev in action, becoming thee Ironman as JevJev, on the Road to Vostok!

[TechLuddite @gh_TechLuddite]

@MichaelGannotti @autonomous_labs I haven't had much time for commenting, but I wanted to pop in here to let you know I enjoy and appreciate the infographics you put out!

Safe travels and happy computing!

[TechLuddite @gh_TechLuddite]

I'll leave this running as long as I can while I'm working:
https://t.co/VRs8Hh7CnW

[TechLuddite @gh_TechLuddite]

AFK Jev Test Stream | My JevJev is becoming Ironman, on the Road to Vostok! https://t.co/VRs8Hh7CnW

[TechLuddite @gh_TechLuddite]

Thank you, @btsouth!

I signed up for the $1/mo plan last night and I updated a small GH Pages site.

The model I used had an ongoing deal of 98% off. As I said, it was light work but I was using the Command Code harness, and I used it for the entire cycle from cloning the repo to a local folder all the way to watching the Pages site deploy.

My % usage didn't even move, so I used a more expensive model to audit the work, it found and made corrections, and it monitored the deployment. Surely that would eat my dollar!

I just wanted to see if the meter was broken I called it quits last night, because my usage % just was not moving. I am happy to report I was able to use 1% of my $1 before bed last night.

[TechLuddite @gh_TechLuddite]

Kudos on the last few updates! Yesterday I was going to ask where mod support was at on the roadmap, because I wanted to overhaul the controls. Then I read the v6 and 7 changelog and you all are on top of it!

I streamed it running from Omarchy Linux on Twitch last night and it played perfectly! Keep up the great work!

[TechLuddite @gh_TechLuddite]

Another awesome one in the bag! Thank you, @ncfrontiersman & @NixFred!

@MuchmoreIT, last week I mentioned a project of mine i thought you might like, but a day or two later I saw DHH give a shout-out to a very mature MIT licensed project we're now evaluating it at work for the Linux & Macintosh systems we manage. We might even put it on the Windows machines, because it has features our closed-source, Windows-focused RMM and EDR platforms are lacking. I wanted to mention it, because it's checking a lot of boxes early on in our evaluation that may have some cross-over to your work. Here it is:
https://t.co/xHWc8ci4er
https://t.co/stxIA9sK6y

@cfaulkingham, I have a project you might be interested in, it's MIT licensed, so use anything you'd like. It has over 500 Omarchy related issue/fix records, with over 1600 sources cited, and a boatload of tokens and kWh's spent already on audits and testing against live models and machines, auto-provisioning VMs for testing and more. Feel free to DM me if you'd like any details.

Have a great week everyone and happy building!

[Christopher Weis @OnTheLiftAI]

Built a tiny Omarchy plugin last night.

Copy anything.
Hit SUPER + SHIFT + Q.
Scan the QR code with your phone.

That’s it.

No pairing.
No account.
No cloud.
No setup.

Omarchy Beam turns your clipboard into an instant QR code, right from the desktop.

Sometimes the best tools do one thing really well.

Open source. ⚡

https://t.co/e1jvcTkxU7

#Omarchy #Linux #OpenSource #Hyprland

[TechLuddite @gh_TechLuddite]

If you use Infomarchy, please go show it some love on the plugin site!

Congratulations, @NixFred, and thank you for putting out such awesome code!

A special thank you for turning it into an amazing community project! That's a layer on top of the code that adds effort and complexity for the maintainer, and I appreciate all the extra time you have given to that aspect. It means a great deal to me personally, and I cannot thank you enough!

[Nix 🔳 @NixFred]

Infomarchy the AI desktop dashboard has been approved on the official Omarchy Plugin Site!

Thank you to all the Contributors that helped make Infomarchy what it is today!

Find it here:
https://t.co/DuwNQtH93B

[TechLuddite @gh_TechLuddite]

@John4ZX4 Great quote, and a fantastic post! Thanks for sharing!

[TechLuddite @gh_TechLuddite]

@chriscompiles Currently not efficient or cost effective, and I spend a lot of personal time reading the docs.

However, it does find things we all are asserting in writing, with our names attached, but not backing up with real code.

[TechLuddite @gh_TechLuddite]

@chriscompiles Read those docs first though, compounding conficonfabulation seems to be a thing.

[TechLuddite @gh_TechLuddite]

@chriscompiles This has been my favorite thing about coding with AI. Self documentation.

I love to use those docs to then hit the repo with an adversarial review to check every single thing the in-repo docs say that maybe the code doesn't back up.

[chris @chriscompiles]

The best thing I’ve done for AI-assisted coding isn’t a better prompt, although that has helped.

It’s making the repo explain itself.

Good tests. Clear docs. Small commits. Explicit constraints.

All boring engineering habits.

Turns out making a codebase easier for an agent to understand also makes it easier for me to understand.

[James T @JamesTervit]

@CompleteSkeptic Jev making my local routing come alive. it routes based on need which has always been a slow blocker for me and the grading on output success is second to none. llama switching to vllm etc unreal.

The best next local lanes are already on disk:

- Ornith 1.5 35B A3B Abliterated on the M3, for creative coding and less constrained drafting.
- Qwen3.6 35B A3B MLX on the M3, for a faster Apple Silicon comparison lane.
- Qwen3.8 27B MLX on the M3, to compare MLX and llama.cpp using the same model class.
- DeepSeek R1 Distill Qwen 14B on GX10-002, through vLLM port 8113 after Minimax H3 drains.
- Nemotron 3.5 Lightning 30B MoE on GX10-002, the fastest measured local text lane at about 103 tok/s when media is not occupying the GPU.
- Nemotron Nano VL 8B on GX10-001 for vision triage.
- GLM 4.7 Flash and Qwen3.6 27B FP8 on GX10-001 as scheduled vLLM comparison lanes.

[TechLuddite @gh_TechLuddite]

@JamesTervit @CompleteSkeptic This sounds super cool, James! Please keep us posted on how it's going!

[TechLuddite @gh_TechLuddite]

@btsouth @CommandCodeAI I'm currently using 5 harnesses, I'd like to use only one, but my professional work benefits from experimentation in my personal work. So...let's make it 6 harnesses!

Thanks for sharing! I'll give Command Code a try tonight!