Tom Lieber 2024-02-08 06:07:08 The future of coding may involve rendering histograms of random variables in Three.js, so here we are.
Justin Janes 2024-02-08 16:43:45 Don’t forget to toy with 3D - i recommend the glTF format in particular
Tom Lieber 2024-02-08 16:49:33 Thank you for the ref. It's been a while since I've done any game/interactive programming, so I don't know where the tech is!
I would love to experiment with that, but I'm reeeeally watching out for scope creep right now. I've decided that juiciness is fair game, but I'm going to try not to go outside my wheelhouse (where 3-D modeling is) for a while.
Tom Lieber 2024-02-08 16:51:59 It's all smoke and mirrors, but seeing the charts fill up was part of the UX that I wanted to test first, and… oh yeah, that's what I was hoping for. 👌
Lu Wilson 2024-02-08 22:55:18 Eli Mellen 2024-02-08 23:21:38 Makes me think of that older mathematical notation that used over bars in place of parentheses.
Mattia Fregola 2024-02-09 03:31:52 Also curious to know what was actually happening with the new variable there. Sparked some nice imagination for me!
Ivan Reese 2024-02-09 21:09:58 Mattia Fregola Holy crap! I was just talking to Lu (and the folks at Ink & Switch) about how I want it to be like this instead of having a separate blue bar and memory line. Nicely done!
Ivan Reese 2024-02-09 21:17:11 If you can see the contents of memory, you don't need algebra anymore…
Lu Wilson 2024-02-10 00:21:06 nice now this is getting closer to cellpond again
the Line is forming, i can see it!
Lu Wilson 2024-02-10 11:02:19 I love seeing these variations. the highlighting is great
there's one small consideration i had: i wanted the 'state of the memory line' to change between the lines of code. kind of in 'continuous' space. i think there's a chance of confusion from seeing the change happen when it's at the same time as a line of code. Like... does the change happen before the line of code? or after it?
Mattia Fregola 2024-02-10 11:21:48 yup, lazy me not doing transitions (and admittedly not really understanding the details of what goes on here)
Mattia Fregola 2024-02-10 11:23:32 also for more context, is this part of something or is it just this?
(so is there an overarching project / goal or just a way to visualise this )
Lu Wilson 2024-02-10 11:56:57 Mattia Fregola no no, lazy me for hoping for other people to do it instead 😆
this is part of some experimental work i'm doing for my residency at ink & switch. the project is called SEE IT. or SEET. or THE LINE. or something. still figuring it out. very very very early stages. i wanted to make a little animated demo that would ellicit some early reactions and responses. the actual planned project looks nothing like this, but I'm carving out some core concepts, and 'trying them out' with people.
the project is about being able to SEE your program's memory. I'm currently exploring the tension between seeing your program's actual memory VS seeing your mental model of the program's memory
Lu Wilson 2024-02-10 11:58:48 It's going weirdly, because I don't find the project very compelling so far. but responses have been strong (both positive and negative). so that's interesting to me. Why is there this difference? Why are other people's reactions different to my own? I'm developing my own understanding of it
Mattia Fregola 2024-02-10 23:09:47 sounds like a fun project Lu Wilson! hope you’ll discover a more intriguing facet to it other than the meta.
curious to see what comes out of it!
Maikel van de Lisdonk 2024-02-10 14:56:46 Hi, in this video (youtu.be/iFga2bfptBk) I show the progress on a new feature that I really wanted for quite some time: combining nodes into reusable compositions. This helps in making flows smaller. The compositions can be accessed from within the node-type selector so they can be added to the flow. One thing that I really like is how the inputs and outputs to compositions are created based on the connections which cross the bounds of the selected area. I've also finally added a way to copy & paste nodes which is a nice productivity improvement.
In the video I demonstrate my other flow engine instance with which you can create fragment shaders in webgl and here I've recreated a shader form shadertoy made by kishimisu (see shadertoy.com/view/mtyGWy ). I had to extend the node-types to make this possible. Currently I cant yet create nested compositions but once that is possible I hope it will help decrease the "node-spaghetti" further. Because I do think that the textual shadertoy code is still easier to read then my visual version. Which obviously I want to improve.
I am almost ready to send out a link to my visual editor but decided to work on it a little bit more for ar least this month. I want to extend and improve some features and create more example flows.
Maikel van de Lisdonk 2024-02-10 16:58:19 Unfortunately I currently only have bezier and quadratic curves and straight lines without an option to add extra control points
Maikel van de Lisdonk 2024-02-10 18:23:44 I was thinking a bit more about this, I think I can automate this.. when a node has multiple connections from a port to other nodes.. simply create overlapping svg paths to the end nodes. I already do something similar for nodes that connect to nodes that have a smaller X position. If possible I want to prevent extra user actions
Maikel van de Lisdonk 2024-02-11 09:16:09 Here's a link to a small video demonstrating the above: youtu.be/QrFnzhhHzD8
Connections that have the same output have their svg path calculated using the distance from the closest node to that output.
I've also changed the way value-nodes are handled by the shader : instead of recompiling the shader after changing the value, now direct uniform parameters into the shader are used which is a much smoother experience.
Tom Lieber 2024-02-10 15:26:46 I wanted to make a pit stop at this tug-of-war example from ProbMods before I moved on to working on the code editor because it's one of the most eye-opening in terms of what's possible in PPLs, and because this is the visual I wanted that I couldn't get in WebPPL, in two ways: 1) all the plots, arranged how I wanted them, and 2) always-on plots that let me see the samples as they come in, with the charts getting smoother and smoother the longer I'm willing to wait.
