Such is the life of a pirate… morning is permanent and all good things will come to an end.

It’s always good to have alternatives and ensure you’re keeping up with them.

And then after 2 years they will no longer support it.

LLMs are okay at writing code, but not good at software architecture

This is very much the case. I have a local LLM that does a fairly good job of writing code for quick and easy apps that I might need - for instance, last week I wanted to create a simple floor plan of my house and couldn’t find exactly what I wanted online, so I had my machine create a web based app to do it. But, when I tried to use it to create a fairly large and complex text based adventure game when it reached a certain size my LLM just couldn’t handle it any longer. That project stalled for a bit until a couple of weeks ago when Fable 5 was released… as a test I pointed Fable at that code base and in about 45 minutes it had completed the entire game engine.

I really don’t like the idea of corporate owned LLMs, but I am very impressed with Fable and can’t wait until the Local LLM community is able to replicate its functionality.

Back when I used to collect comic books I’d always get excited by the ‘book value’ of something I owned, and my dad would always tell me “No matter what the price guide says, that comic is worth only what someone is willing to pay for it”.

Most people who are going to buy into SpaceX are likely going to do it out of love for space exploration and a desire to be a part of SpaceX.

Even if the stock dips soon after the IPO, getting in early on a huge IPO such as this historically pays off huge in the long run.

Last night I decided to try Little-Coder to work on a game that I’ve had in the ‘plannng’ phase for a while. I had a few issues getting Little-Coder to work with my system, but once it was working I pointed it to the project fact sheet in my Obsidian vault and told it to get familiar with the project and ask any questions. It had a few questions and then wanted to know if I wanted it to get started on a prototype for the gameplay. The game is a narrative driven politics game and I asked it to create the prototype as a brower game. It ran for about 40 minutes and then had the working prototype. There were a couple of bugs with some buttons not working, but those were easy to fix.

This is really impressive. I’m making this game for my daughter and I let her play the prototype (which is very repetitive at this point since the ‘narrative’ hasn’t really been written yet) and she loved it. She played it for about 15 minutes before realizing that it had started repeating. I’ve been working on the game all day today and she came in a bit ago and asked when she can play again.

And thanks for that video on Qwen3.6 27 dense… I might have to throw my system at it and see how it handles things.

I did some analysis on OpenCode, Little Coder, Aider, OpenClaude and ClawCode and each one had potential issues with my specific setup (OpenCode and OpenClaude have a looping bug at the moment, Little Coder is brand new and has potential, but I’m leary of new projects, Aider is promising but had some unknowns)… anyway, as I was evaluating these I realized that I currently use Claude Code, all I have to do is use that with my model rather than using their models. A few quick changes and now I have Qwen3.6 in Claude Code. I asked it to create a web-based Tetris game… 7m13s later I have a fully functional, minimally buggy Tetris game in my browser!

Thanks, I plan on checking out Little Coder later this weekend, it looks really cool, thanks for sharing!

I was trying that with Coder-Next but it would often just ignore those instructions. I didn’t change any of my instructions when I moved to Qwen3.6 and it has been following them solidly.

For ingesting I’m getting around 200 t/s, and for generation it is around 48 t/s.

I hadn’t heard of OcuLink before the Framework event yesterday… and now I’m intrigued. How does that connect to the Framework Desktop?

And as far as tools go, I tend to borrow ideas from others and then build them specific for my setup. For instance, a few months ago I came across a project called OpenBrain that uses a vector database for memory storage and retrieval. I leaned on Claude and asked it to evaluate the OpenBrain project and to let me know how we could use the concepts for my local system. Claude chugged away and then gave me its recommendations, which included spinning up PostgreSQL, creating an ‘Observer’ to watch my OpenClaw and Claude sessions to pull memories into the database, setting up an embedding LLM server to generate the embeddings, and then a process that runs nightly to remove duplicate memories, combine similar memories and archive old memories. If I recall almost all of these were components of OpenBrain, but I’m cautious about using other people’s code and appreciate that Claude is a good tool for using ideas from other projects without actually needing to use the whole project itself.

One instance recently where I’m glad I did this was Milla Jovovich’s MemPalace… since I already had a memory system in place I didn’t need to use her project, but one thing I did find very intriguing was the AAAK, which she had described as a lossless compression language for AI agents. I asked Claude to evaluate all of MemPalace and see if there was anything that would be an improvement to my setup, and Claude also got excited about AAAK. Claude said that everything else we were doing was as good or better than MemPalace, but if we could add an AAAK version of each memory that would help lower the token usage (but we kept the full prose memory to help with search). We implemented it and I asked Claude to pull some random memories from the db… and it was immediately clear that there was no way AAAK was lossless because a lot of my memories had to do with system details, including IP addresses, and the AAAK didn’t include any IP addresses. Of the 1800 memories that I had stored, Claude was only able to find 3 where enough meaning had been preserved with AAAK to be usable. It was easy to remove from my system - and I just checked and the project no longer refers to AAAK as lossless.

Using this method I also built a filesystem based context system that allows any LLM on my system to use the same skills, context, agents, projects, and memories (although the memory folder just has instructions on how to access the memory database). This was another project I saw from someone else and I used a lot of their ideas, but tweaked things to fit my needs.

Other tools where I am using other’s projects are ‘Faster-Whisper’ for speech-to-text, ‘Piper TTS’ for text-to-speech, ‘Moondream2’ for image analysis and QMD Search for indexing my Obsidian Vault and putting it into the memory system.

This is very exciting to hear! I might see about creating a dedicated coding agent to test it out.

Creating dedicated agents isn’t something I’ve really done much of yet, I’ve mostly just kept it as a single agent, but this might be a good use case for it.

I went with the decked out Framework Desktop: AMD Strix Halo APU (Zen 5 + RDNA 3.5 iGPU), 128GB shared system RAM, running Nobara Linux.

The only thing I skimped on was ssd. I only have 2TD. I was going to go with 4TB but I figured I should hold off because ‘storage prices only go down’.

Same, I believe we’ll eventually get there with open models, but right now it’s hard to beat the frontier models for coding.

No, I feel fortunate that I was able to get the Framework Desktop before the tariffs and chip shortages hit and so I have the computational power that I need without needing to rent it. I’m sure if I did the math I would probably come out ahead by renting but I don’t like monthly bills, especially ones that fluctuate.

I have mine setup to only use thinking for complex tasks, and that can take significantly longer (2-8 minutes for thinking versus <30s for not thinking) but when I look at its thought process it is impressive the logic steps it goes through to solve an issue.

It doesn’t have to be unhealthy. They do have salads and grilled chicken, and even grilled chicken salads. Of course, the healthy items are quite a bit more expensive than the unhealthy items ($3 for large fries vs $4.20 for kale salad).

It works in Europe too. It uses LoRa (A Long Range radio protocol) to be able to send messages out to other nodes, which can bounce them out to further nodes. A node can be configured to relay through the Internet to reach people in other areas.

I ordered the radio shown below from a kit on Amazon (it’s a Heltec v4 and came with a battery that isn’t pictured) and it took about 5 minutes to setup. Attaching the antenna to the board was the hardest part.

Great points, and there’s some amazing discussions going on here!

One thing I’d like to add is EVERYONE needs to start setting up some meshtastic nodes. It’s really easy to setup (just hook up a USB cable from your computer to a esp32 board, visit a website to get the configuration, and that’s pretty much it), it’s cheap (as little as $30) and it is secure. Build 2 nodes (one to leave at home, and another for your backpack). This way you’ll be able to communicate should the Internet become unavailable or unsafe. You can also use this at a protest so that you still have a means of communication without needing to bring your phone that the Feds will be able to track.

This is what I did… I tried to ‘just move on’ without blocking them, but they had commented several times in a thread I was trying to read and it was such a distraction, so I blocked them and only ever think of them when I see posts like this. It’s a shame too because the person I blocked did seem to have worthwhile comments, they were just too annoying to try to read.

But… that costs money.