codazoda

Email Storyselling Playbook by Jim Hamilton

https://a.co/d/08zSx2LD

30 years ago a teacher told me of an art student applying to a school. The student included two art pieces for the submission but was meant to provide three. A note suggested that they look at the stamp for the third submission. It was a hand drawn counterfeit that was good enough to get it delivered.

Or so the story goes.

I have been building BASIC32, a ROM BASIC like the ones I grew up with.

The first one runs on an ESP32 that I plan to use as a watch. The software works, I just need to finish the 3D printed case.

The second runs on an RP2350 connected to my living room TV via an HDMI cord.

I have so much appreciation for these old machines I grew up with.

    BASIC32
    
    Ready.

I'm no fan of GitHub add-ons and I agree with the premise here but...

I can think of one other possibility. It's easier to write in a wiki via the browser. I can open that on my phone and edit docs. I can open it in my browser and edit docs.

On desktop it's a tiny bit more to pull the repo and open it in your editor (and you might already be there) but that tiny bit can be enough to stop you from writing documentation. For me, writing documentation must be totally painless so that I'll actually do it.

Why am I not a fan of the add-ons like PR's, wiki's, discussions, projects, and issues? Because they each introduce vendor lock-in to varying degrees.

I’ve been building a dark software factory. A system that takes stray thoughts and automatically plans and builds.

I’ve built two (successful) iterations and am now working on the third. They are all very minimal.

The main one works well and runs full time waiting for my next idea. The second targeted an open weights model and worked in limited cases.

Now I’m working on a more autonomous version. I give it only a goal and a strategy, it decides its own projects.

I’ve written a little about it here:

https://joeldare.com/creating-a-minimal-dark-factory

So far I’ve created a couple simple web based games and a BASIC32 interpreter that boots on an ESP32 and turns it into a 1980’s style computer.

I use Sonnit for my personal work and Opus for my professional work.

For my personal stuff, I'm on a small $20 plan, so I need to use tokens conservatively. I was very rarely exceeding limits until I built a Dark Software Factory. It's not as efficient at token use. So, I use Sonnit over Opus here.

At work I have a $100 plan that I rarely exceed so I use Opus. I have access to Fable too, and I did use it a lot while it was new, but I don't find it improves most of my work by too much. I do mostly bug fixing across several hundred repositories with hundreds of thousands of lines of code, mostly written by humans over the past 20-years. These projects interact with each other so I run claude from the root of my sandbox (I was nervous to try this but I'm not looking back now).

I also use Sol as a secondary for my personal work. I pay for it because I like to talk to ChatGPT on the web. Since I already have the subscription, I let Sol write plans for me. It does a better job at certain tasks and it saves me some Claude tokens. Maybe I should consider Terra for the task, but I don't run up against my usage limits for the little bit I use it.

I'm trying to use Gemma 4 12b for some workloads but I haven't mastered the model yet. It's still very experimental for me. I can get work from it but it takes a lot of hand-holding. For local models, however, it's all I have the RAM for.

That is almost to say that the only part of the project that matters is the code. But the code is only a tiny part. I take joy in describing the thing and bringing that vision to life.

This is timely for me.

Just an hour ago I made my most recent pen plotter generative art free on my website. I plan to send high resolution scans to everyone who joins my mailing list so you can print them at home.

In contrast with the author, I actually enjoy creating code with LLM’s. I’ve been on the technical management track, instead of boots on the ground coding, for quite a few years now. LLMs are letting me produce software again with limited time.

I still enjoy writing code by hand too and my generative art work is mostly that.

https://art.joeldare.com

On the game side, I recently created a Shut the Box game that runs in the web browser. This was created entirely by an LLM. In fact, it was the test project for a Dark Software Factory I’ve built.

I love that I can describe a project, from my phone, put it away, and have a prototype built a few minutes later.

The game is clearly unique and consistent with my own personal style. And, I’m enjoying playing it.

https://dark.joeldare.com/shut-the-box

I once did this by accident and without a bot farm. It was very lucrative.

What I did was create a game with an action button. Then I served ads right next to the action button. I didn’t realize people would miss click on the ads at a high rate, but they did. They fat fingered it, clicked the ad, I got paid.

I looked into it after a couple weeks because my click rate was crazy high.

My guess is the disks or drive are bad. I have a similar modern Chinese USB 3.5” floppy drive and a box of disks from the 80’s and most of them seem to work fine in that combination.

My research is a few months old, but at the time Vizio looked like the least bad option, largely because a lawsuit had pushed them to narrow what they collect.

Sounds pretty good. Reminds me of the Demo Scene and the types of music they were able to create.

Single html file, ~1000 lines. Love this kind of software.

I want a Qwen 3.8 27B hosted locally but I don't quite have the RAM for it. And, I don't want to buy the RAM until I prove I can use it.

Yesterday I did have success with Gemma-4-12b with 128k context. It fits in my RAM and it's relatively fast on my hardware.

I had to give it prompts that are quite a bit different from the way I use foundation models, but I did get it to work quite well. I feel like I could learn it's differences and get good at using it for real work.

Really an aside, but yesterday I got the Gemma-4-12b (128k context) to build it's first web app in the minimal Dark Software Factory I've been building for myself.

https://joeldare.com/a-local-open-weight-model-builds-its-fi...

Got it. But, they also charge the same for cached tokens, so that probably closes the gap on Foundation models quite a bit.

Do I understand their pricing correctly? This is $10 per month for a developer account PLUS you pay $1.49/M for output tokens and $0.99/M for input tokens on Qwen 3.8 27b with a 128k context?

EDIT: Or, maybe it's just token pricing, but $10 is the minimum? Maybe it's that.

https://www.cerebras.ai/pricing

I kinda assume it's because one went down and a large amount of work shifted to another.

I'm also aware that they have overlap in some areas on data centers.

I never tried ASM on it. Interestingly, the TRS-80 CoCo did not have a Z80, it had a 6809E @ 0.895 / 1.79 MHz. Mine had been upgraded to 64KB of RAM.

My identity is tied to it's competitor, the TRS-80 CoCo, released in 1980. I didn't own it until a few years later. My friends had the Commodore 64 and I was jealous.

Your art reminded me that some of my earliest code produced animated flying rockets:

          *
         * *
        *   *
        *   *
        * U *
        * S *
        * A *
        *   *
        *   *
        * * *
       ** * **
      *********
          *
         ***
         ***
          *
          *

In my experience they work well for some jobs.

I recently built a minimal Dark Software Factory out of an N150 Mini PC. It uses three models; Sonnit, Sol, and Gemma.

But, I have a LOT of instructions about how I prefer the software it builds. Gemma doesn’t handle all my instructions very well. But it’s close!

I’m running gemma-4-12b because I have limited RAM and larger models were too slow.

I do two types of jobs: planning and prototyping. It has done fine at some of my planning rounds.

I still consider it experimental and don’t use it a lot but I think we’re getting there.

> I have glued two small neodymium magnets to a 0.2mm fineliner pen

This sounds like an extremely simple way to attach a pen to a 3D printer. I'd love to see a picture of it.

I own an Ender 3 V2, which is a great little printer. I use it often and am considering a newer Ender because I want a larger print bed.

Anywayz, now I'm going to go see if I can easily throw some magnets on the head. Then I might attach those magnets to a pen or pencil.

I’m not sure I understand the correct answer. A public company has a fiduciary duty to the shareholders. This company opted to become public.

I was recently inspired by another article here to start my own. My skills are written and tested, the factory has built the first test project, and I’m setting up the final machine to run it. Here’s my initial post about my motivation and early plans plus some follow-ups, and there are more to come.

https://joeldare.com/creating-a-minimal-dark-factory

I really enjoy this format. Like reading a story from the authors life. Like you might read in an autobiography.

I don’t think I write this way very often, but I enjoyed reading it.

My takeaway is to write a bit like a journal more often.

I don’t want to read AI slop either, but it only seems to be a rule for comments.

And the keyboard is SO BAD. :P

Source: I still own a TRS-80 CoCo.

How are you doing this with Opus. Clearly I’m missing something. I always turn to ChatGPT when I need images because Opus typically refuses. I’ve tried Claude Code and Claude online in the past. I’m pretty sure neither created images for me and I thought this was because Anthropic was focused on code.

I guess I need to try harder. :)

Aside: The Typora website is horrible on mobile.