[dr_dshiv]

Thanks for putting this together!

I’ve been thinking a lot about demoscene in a Vibecoding era. Demoscene was often very close to the metal; Vibecoding is often completely abstracted from it.

To explore this tension, I’m cohosting an 8bit game design workshop (pico8) in Amsterdam this summer. Just for fun.

Working with intense constraints can bring a lot of creativity. I really want to see how AI affects the workshop vibe.

[elpocko]

>very close to the metal

>(pico8)

But PICO-8 with its integrated tools and Lua programming is super high level, pretty far removed from "8bit". It's only 8 bit in aesthetics, entirely artificial and forced. Why not write games for the 2600, the C64 or the NES to experience real constraints of an actual 8 bit platform, the actual "metal."

[dr_dshiv]

I know, tell me about it, we’ve been debating this, tbh. We landed on pico8 to introduce the ideas of creativity within constraints as there is a very active community and nice tooling. We’ve done NES workshops before, but always with CS students and eproms. 6502 assembly isn’t that complicated, but can be scary.

[elpocko]

I've been thinking about making a fantasy computer/console that uses an emulated 65C02 CPU, maybe a bit simplified -- every instruction only taking 1 cycle, for example. You could use all existing tools for the 6502, like the amazing 64tass, and there's an interesting high level language called prog8 (https://github.com/irmen/prog8) that's relatively easy to use. Not sure if there's any interest for something like that.

[Lerc]

I was working on something using AVR assembly as a fantasy console.

I never quite completed it, but I managed a emulator/assembler IDE in the browser. Making my own assembler let me play around with some ideas for macros.

It could even load programs from gists.

https://k8.fingswotidun.com/static/ide/?gist=ad96329670965dc...

Reflecting on it now, I think one feature that could help a assembler on devices like this is the ability to inline compile assignment expressions that use values of only one type. It would be easy enough to emit a stream of instructions for

    x=32+(x*(y+2))

or even

    r15=32+(r15*(y+2))

using registers as expression values.

The result would usually be what an assembler writer would write themselves.

I think a macro assembler that did that would ease a lot of the tedium of assembly while maintaining the near absolute control over memory use and IO that you need for low level coding.

[ceronman]

Hi! That sounds very interesting, where can I find information about this workshop? I am interested in participating.

[dr_dshiv]

thanks! my contact info is in my bio. Drop me a line -- the website might be up later today... or next week haha. But several friends flying into Amsterdam to make this a reality.

[imiric]

Vibecoding is entirely antithetical to demos.

Demos aren't just pretty visuals. Their entire purpose is to showcase the skills of the programmer in accomplishing something remarkable within a constrained environment, or within artificially imposed constraints. Using exorbitant amounts of resources to run a machine learning model to generate a muddy mixture of existing demos goes completely against this spirit. The idea of the two together is, frankly, repulsive.

[dr_dshiv]

Be repulsed then! We are doing the workshop specifically to explore this tension. I’m well aware that times are weird.

This workshop is actually a forerunner to a workshop on demos with quantum computers, where the constraints are very real and creative participation is very lacking.

[esafak]

The game is to see who can prompt the LLM spit out the tightest code!! Sergey Brin says threatening the LLM works in a pinch!

[gneissguise]

It was my pleasure to work on and share, glad you liked it!