Hi everyone! I'm Jen, a dance/music flavored person that uses computer to pay the bills, somewhere between site reliability, platform, and being an 8 foot tall unicorn doing tricks on a chair. I stumbled upon GToolkit, and was like, "ok cool, this is worth learning Smalltalk for", asked @Eli Mellen, "hey you ever hear of this thing?" and immediately got invited here (thanks!). I feel Malleable Computing is going to be super important in the LLM age, gotta figure out what the agents are doing somehow, especially if I'm responsible for the output and the system it produces. Big fan of graph views, 2d text editing, and visualization (think CodePod, IDA's graph view, Debugger Canvas, Max, Pd, stuff like that). Current side project is inspired by Codepod, but in Neovim. In broad strokes: Convert some source code to a call graph, layout the graph, then make all of method/function nodes floating text editor windows on a 'canvas', grouped by context. The floating windows work, spawning a bunch of 'em bounded around methods works, next up is drawing the edges between nodes 🙂

Hi Jen. Please do tell me about what you think about GT 🙂. And would you be interested to learn how working with LLMs feels like in it?

Hi Tudor. So far GT feels like a blend between Jupyter, RSW, and hypertext, but with panes and OOP. I like it! Going through the GT in 7' videos, and I've found them quite helpful. The documentation is impressive. I am still ramping up in Pharo, can't (yet) ask intelligent questions about GT, but I'll get there soon. Once I'm there I'll ask for more info about working with LLMs in GT. Definitely appreciate the composability of GT, I am a big fan of a bunch of single purpose small tools that get lego'd together and being able to extend them to my heart's content. I also dig the DevOps investigations example on the home page, wrestling CI pipelines is a big part of my 9-5.

Thanks for the details. I am glad you find the documentation useful. I think we still have a long way to go until we learn how to properly explain this though, which is why I encourage people to ask questions — even, or especially, if they find the questions to be too basic 🙂. There is also a dedicated Discord you might find useful. Regarding the LLMs, there is a page in the book that provides some examples: https://book.gtoolkit.com/playing-with-llms-e5f2uq8whwfmilp5k7jg4inc3

Ooooh and there's MCP tool + server support. Neat!

You might find you do not need that 🙂.

Hello both! This might look like I’m putting my nose into other people’s business. I started browsing around to get oriented and noticed that GToolkit is a thing here. Here is a basic question, as per Tudor’s encouragement: Does it make sense to import the system I’m currently working on and have the toolkit generate multiple representations of it? For example, could I just import the three folders of my current system (three repos for FE, BE, and ML) and start exploring? While watching the presentation of the Kit, I began thinking about how the legacy-system migration approach could be applied to our education system. It seems like a legacy, outdated system—not only here in Bulgaria, but everywhere. So perhaps by understanding it deeply and creating thousands of tools around it, we could find a way to migrate it.