Does anyone know the current state of agent-assisted Smalltalk, Pharo or otherwise? Intuitively it seems like a match made in heaven, since the agent can directly manipulate the running program, but I also assume that there is very little training data available for Smalltalk.

▲

whacked_new 4 hours ago | parent | next [-]

I spent some time exploring Glamorous Toolkit [1] for agent-assisted custom UI development, in an emacs fashion, where you'd modify the interface you're working with on the fly. It was a pretty good experience and the built-in data exploration was top notch.

Ironically after watching glamorous toolkit videos on youtube, I was recommended a talk by Geoffrey Litt about dynamic documents [2] and at the end of his talk I learned about bb [3], which has much better agent integration compared to GT. Like @thelazydogsback said, with agents, the programming language matters less.

[1] https://gtoolkit.com/

[2] https://www.youtube.com/watch?v=MccJdr61xnc

[3] https://getbb.app/

▲

pmkary 2 hours ago | parent [-]

I was talking at around a year ago with Tudor Girba and asked him what is his opinion towards AI---I somehow believed it to be conflicting with his world view, but then he showed me some amazing demos of llms walking inside of the GT, making tools and molding it. It was way before these new agentic plugin-based IDEs.

I think Girba's really amazing idea of "language = ui = software = data" has a huge win in terms of AI that is not used anywhere else (maybe only in Chris Granger's Eve), but to work with data and build live on it, is something way better than BB.

	▲	whacked_new an hour ago \| parent [-]
		I agree, it's an awesome concept in principle (and Girba is graciously welcoming people in discord). Working with an agent in GT, the agent will just look at the data and figure out how to express it in the graphical framework. But in practicality, at least at this point, bb wins out by sheer expansiveness of ecosystem. I was working on a mishmash of an editor interface that combined obsidian with heptabase, directly in GT. It kind of worked. And the amazing thing is, everything has just a single representation: window positions displayed as text automatically have 2 way bindings across widgets, because they are the same data. But even with agents doing everything, 2 things creep up. First is in JS/TS, all these binding problems are solved 1000x already. You can fake shared memory with callbacks and the agent won't lose track. In GT you move the cursor around, and have to decide at what point do you convert a widget representation to source representation, and when to restore the widget: is it a delay? is it after changing lines? is it after you move off the text? before or after the whitespace? There's a package for that somewhere in npm that answered all these questions, and it has tables, and it's performant. It doesn't do the 2% stuff that GT does using its magical shared memory representation (which also makes TS->smalltalk porting not straightforward), but you can still tape over it with sheer optimized code. The second thing is that GT got really slow with all the data structure re-interpreting and rerendering. As "dirty" as bb is in reloading dynamic code, the runtimes are so optimized, I wanted the markdown nodes to be GPU-accelerated in bb, and it turns out they already were.

▲

Smalltalker-80 an hour ago | parent | prev | next [-]

SmallJS ( https://small-js.org ) is a file based Smalltalk with full agent support. Here's a vid of Claude converting a Counter app into a Tic-tac-toe game in a single shot: https://www.youtube.com/watch?v=4_ir6h2MIRg PS The context loading step is not even necessary anymore. The reason why this works so well, is that Claude knows its way around VSCode well. And all SmallJS projects (apps) have full unit tests and GUI tests that run in a couple of seconds. Claude also reads these tests and can make new ones, maintaining coverage. In tradtional live Smalltalks like Pharo, modifying stuff is easy, but running full tests not so much..

▲

kvgr 4 hours ago | parent | prev | next [-]

It got me interested so i ashed ChatGpt about the options andit found this repo: https://pharo-llm.github.io/pharo-mcp

Sounds pretty cool. Never heard about Pharo before, but i will try it in spare time.

▲

flavio81 4 hours ago | parent | prev | next [-]

>Intuitively it seems like a match made in heaven, since the agent can directly manipulate the running program

This should be a BIG benefit for agent-driven development, but sadly it seems that few people is taking advantage of it.

▲

xkriva11 4 hours ago | parent | prev [-]

It works well, I mostly use Pi with MCP.