Marco Gomez

How I created an OpenSource 3D game with @threejs and MML while live streaming the development of the entire game in a way that it could be watched inside the game itself 🤓 Time stamps: 00:00:00 - Intro 00:00:18 - The game engine 00:02:06 - The MML documents 00:04:06 - multi-player sync 00:07:58 - memory game example 00:09:20 - client-MML communication 00:10:50 - reusability and composability 00:12:06 - MML tags and attributes 00:13:32 - interoperability 00:14:11 - the platformer game 00:16:15 - more MML tags and attributes 00:18:16 - let’s write a game 00:24:52 - 3d objects with primitives 00:33:41 - creating interactive blocks 00:41:29 - creating click event handlers 00:44:30 - creating animation functions 00:51:25 - the game step function 00:59:22 - checking states for victory or draw 01:03:42 - creating display labels 01:15:16 - the reset game function 01:17:29 - the finished game 01:18:50 - final considerations If you liked the project, please also follow these legends: @MarcusLongmuir : the father of MML; @MorgeseSacha and @deej_io : my buddies at the MML Team; @RJFWhite and @HermanNarula : creating the craziest possible shit you can imagine (and some you can't even imagine yet) at @Improbableio

Eu investi um fim de semana ou dois olhando pra esse vespeiro, e um launcher é sim uma ideia viável e interessante. Dentre os problemas mais complexos que encontrei, a falta de padrões nos jogos daquela era é o pior deles. Um dos exemplos mais trabalhosos que eu tive que resolver que parecem triviais mas não são, foi lidar com o mouse PS/2 (as implementações da época lidavam com interrupções e incrementos de posição, enquanto nossos padrões atuais lidam com posições absolutas do ponteiro em relação à viewport). Resoluções diferentes, dois estados de cursor independentes, conversões assimétricas, etc. É um problema não apenas espacial, mas também temporal.

Generate your TTS with timings (you can do it on the cloud through ElevenLabs or through local engines like Kokoro, Chatterbox, etc.). Then I created an aligner (runs in a background worker). Every chunk of audio goes through it, and it uses a speech recognizer with a pronunciation dictionary to time each known phoneme like "m", "ah", "s", "o" (known means that I have a measured mouth shape for it). The mouth shapes' weights were measured with nvidia's Audio2Face, and I composed a pre-computed table. The player reads the audio clock and is responsible for lerping the mouth shapes' weights over time, overlaid on top of the current facial expressions (which are driven by tags between [square brackets] my agents generate for each sentence of their responses.

What really bothers me is that the X API was never that expensive for genuine use cases, and the spam-bot problem was never solved. In read-only mode, the API would be a fantastic way to keep up with current tech trends and news (which is hard to do these days), but their new pricing model makes it prohibitively expensive.

That wouldn't be a big problem, cause we all perceive things differently. What bothers me is that most of them tend to make things sound different from **what they clearly see they are**. Disingenuity matters the most.

People have a hard time believing me when I say GPT-5.4-mini is much better than Luna, but it is. Luna is 7.5 to 9X cheaper tho, but it's a shitty model.

webgpu+compute shaders and/or a razor blade. easy

Awn. Thanks, man! You should see the experiments with Metahuman models. I still haven't solved hair properly, so the nightmares they give me are just 😙🤌 impeccable

I know this is a particularly creepy character I'm using right now, but I'm having a lot of fun with lip-sync and giving agents a face.

Sem tempo irmão. To ocupado tentando encontrar novas formas de pagar mais imposto.

Qualquer um é menos comunista do que eu. To nesse exato momento lendo teu tweet e socando uma bronha, de olhos fechados, pensando sobre a propriedade coletiva e os meios de produção.

Quem é Ian Neves e por que eu deveria me importar?

Qwen 3.8 27B is amazing. But it absolutely needs fine-tuning for reasoning effort The `Actually,`s and `Hmm,`s make the model blow up second-guessing itself And yes, I'm aware of ThinkingCap, QwenPi, and Swift 1.5, and their respective approaches. I know it can be better tho.

This is so cute!

I have the same issue. I'm working on some "retro-revival games" in my free time, and I cannot finish them properly cause I waste too much time playing LOL Having big problems finishing this one 😅 mgz.dev/games/tankz/play

I absolutely fucking LOVE this! So much personality! I really love your work dude, but this is one of my favorites!

Treat and happy too, dude. Can't complain 🙂

Now that your coding skills are 100% offline, and I have your attention, tell me: how is life?

I understand your frustration, and I was genuinely upset chatting with a friend about it last week. What I finally asked to settle the direction was, albeit hyperbolic, "can this run on a fucking Playstation 5?" because "consumer hardware" seems to have lost meaning somehow.

Como se vendeu? É um episódio de podcast cara, não é um interrogatório. Político é liso. Tu queria que ele fizesse o que? Não deixasse o cara falar? Batesse na mesa? O Vilela botou papo na mesa (como o sigilo, por exemplo) que a maioria dos entrevistadores merdões do Brasil não botaria. Eu TAMBÉM queria que ele tivesse pegado mais pesado, mas julgar de fora quando não é tu fazendo a entrevista é macio.