Martin

Programmer | Gamer | Game Dev
🇳🇴

[Martin @marvhus]

@packyM I found a few ways to do it using combinations of A/B and B/C, though it requires using the same tetramino more than once.

Using A, B, and C together doesn't seem to be possible.
Using D or E doesn't seem possible at all, as that would block off other squares completely.

[Martin @marvhus]

@Dashytwo Smalahove is delicious

[Martin @marvhus]

@dhh https://t.co/sEJD446waB

[Martin @marvhus]

@valigo I have had more audio issues on Windows than on Linux.

[Martin @marvhus]

@SHL0MS It's not about how good it looks.

Monet put thought into every stroke of paint.
The AI just made something resembling whatever you described.

Put a different way, it's interesting that AI can generate it, but what it generated isn't interesting... at all.

[Martin @marvhus]

@ThePrimeagen OotSS

[Martin @marvhus]

@_AlexHirsch @walterkirn It's impressive that it can generate things, but what it generates isn't impressive or interesting.

[Martin @marvhus]

@awesomekling The loud minorities are always annoying, regardless of their opinion.

[Martin @marvhus]

@crierlon @notch Voxel raytracing lets you skip having to generate a mesh, and doing raytracing on a fixed grid is really fast.

Here is a good video covering the topic:
https://t.co/PUlW90RSrL

[Martin @marvhus]

@ScarletLabyVT @YugoEnjoyer https://t.co/P8NHdQeDbk

[Martin @marvhus]

@YugoEnjoyer (There is also a more condensed version of it if you don't want to watch through all the streams: https://t.co/P8NHdQeDbk)

[Martin @marvhus]

@YugoEnjoyer It's not easy, but you can learn a lot of stuff without needing a wiki.
About Oliver proved that this is possible through his Minecraft Blind series: https://t.co/JCBf4Fl8sV

[Martin @marvhus]

@richterich_ @ThePrimeagen ffmpeg -i image.webp image.png

[Martin @marvhus]

@ID_AA_Carmack It can be a gift, but it still comes with expectations/requirements in the form of a license, which is stripped away in the process of training the LLM.

[Martin @marvhus]

@raysan5 That would favor those who are better at public speaking

[anton mikhailov @antovsky]

I have a hypothesis, that the framing of LLM coding as "a tool" is causing a lot of confusion. Framing it as "a tool" implies that good programmers should be learning just like any other tool in order to improve their craft.
But I think a better mental model for an LLM is actually something like "a delegate". Framing it that way, makes a lot of the feelings and affordances make much more sense.
When you use a better tool, you can get your work done faster or to a better quality than before.
When you delegate work, you can get your work done faster or to a better quality if the person is better (or a worse quality if they are just faster).
When you use a tool, you also learn get better at your craft.
When you delegate work, you don't learn much and don't get better at your craft.
When you use a tool, you enter flow and focus inward.
When you delegate, you focus on your communication outward.

So while on the surface the results of using a tool or a delegate are the same (or we can at least stipulate them to be so, with an infinitely powerful LLM), the effects on the person employing the methods are drastically different.

This to me explains why I'm perfectly happy to delegate some kinds of work to an LLM (just like I am perfectly happy hiring a tax person) but I want to keep tradcoding my own personal projects and other things that matter.

[Martin @marvhus]

@TeamYouTube @valigo bot

[Martin @marvhus]

@valigo 7) no data collection.