Am I a human dreaming to be an LLM? Or am I an LLM dreaming to be a human?

Sunnyvale, CA
Joined February 2011
so I made a thing: github.com/hxy9243/chorale a small skill for reading and creating music🎵 sheets. supports musicxml import. feel free to play it around with your agent 🤖
3
4
516
Kevin Hu retweeted
这个也太猛了! @LexnLin 让 Opus 5.5 用 three.js 从零搭了一座古希腊城,全程不用任何外部素材,网格、贴图、着色器、树、雕像、瀑布全是代码生成的。仓库叫 Athenaion。 做法是先生成 9 张 AI 参考图,然后把构图、比例、材质、颜色尽量对准它们。他给 Claude Code 立了规矩:弱核显也能跑,质量靠离线渲染;瀑布和水要真实;远看也得够细。接着引擎、气氛、阴影先写好,再开 12 个并行专员分别管地形、水、建筑、雕像、树这类模块,对照参考图一轮轮修。 大概 45 小时墙钟时间,27 小时在搭建迭代,55 个子代理,约 45k 行 JS。最后在集显 AMD Radeon 780M 上逐帧渲了 4K 成片,渲了十几小时。 提示词和代码已开源。
32
72
9
538
45,840
Gemini Argon is ok in following instructions, but still lags Claude in taste.
1
52
Kevin Hu retweeted
花费16天,我用 Opus 5.5 将小王子的故事制作成了3D剧情游戏 结合Tripo+Blender,Opus 5.5 将小王子故事中的每一个星球变成了可探索的3D场景 希望大家能支持一手,后续会开源整个游戏,以及详细制作流程 Welcome to The Little Prince: Across the Stars✨:littleprince.dev/
1,389
1,714
858
16,807
3,730,754
this is Opus 4.5 a year ago what difference a year can make in AI
Argon has this amazing natural understanding of the world. It can do so many things from first principles. Here's a pelican riding a bike, using fourier epicycles. Video made by Argon. Time-synced Lyria 3.5 music prompted by Argon.
45
so that’s why memory is so expensive right now
我听说有个新模型,2分钱1秒,就能搞出这个效果。 是不是改编大时代要来了???
1
105
“If everyone is busy making everything, how can anyone perfect anything?” fells incredibly relevant today youtu.be/LcGPI2tV2yY?is=QtAh…
1
44
tried it with sol-6.1 looks pretty good with one prompt and a few iterations
开源四天,huashu-art-motion已经3000+stars🌟了。 今天,我耗尽Claude Code token,对它进行了全方位的迭代和更新。主要是为了让Codex和国产开源模型也能做出类似的效果。 为了说明迭代后的skill的适配能力,这期项目更新的视频是让DeepSeek V4.1 flash全权操盘,自主制作的,欢迎项目股东们检验效果~
1
293
有点不敢想象这一代长大的AI原生孩子会是怎样
Amazing!这个视频再次刷新了我的认知!! 这是一个国外网友用纯 opus5.5 制作出来的!!没用任何视频生产模型参与。 他提供了手绘作品和参考视频,让 Opus 5.5 编写程序,约 1 小时 20 分钟完成作品 最近一直在深度玩纯模型做视频,停不下来。 等不及的朋友直接去仓库: github.com/agharsallah/opus-…
1
81
yeah
Meta CEO, Mark Zuckerberg: "within 12-18 months, most of the code is written by AI" It won’t just be autocomplete. AI agents will set goals, run tests, find problems, and write better code than top engineers.
43
there’s also an openclaw i now have more agents than i have tasks
1
86
We’re living in incredible times!
Computer made this animated scene from concept art to final edit. It planned shots in Blender, generated reference images with Nano Banana, and animated and scored it with Seedance 2.5.
31
17
1
237
38,144
Kevin Hu retweeted
梵高场景生成器更新,这次搭建了相对完整的《星夜》小镇。 假期在家尝试用工具中现成的生成器制作小镇,查漏补缺了很多很多组件,还做了天气系统。 最开心的是终于把心心念念的星空流场做出来了,实时的笔触,而且星星的大小和位置会影响附近的流动轨迹(当年在 Blender 里试了很多次都没做出实时效果) ▶ 主力模型: GPT-Astra Extra High ▶ 辅助模型:Opus 5.5 High 目前我的迭代思路是:先把最小单位的生成器做出来,然后用这些做出来的生成器亲手搭建一次,每次都能修复不少bug,并且让这个生成器越来越好用。 油画材质球 → 最小单位的生成器(花/草/树/木/云/建筑)→ 简单场景 → 复杂场景 还没准备好上线,目前暴露的参数面板只有我能看懂,我还得收拾收拾……后续应该会逐个场景发布,下一个场景准备试试黄房子,感觉把完整小镇做完,建筑生成器就足够好用了,祝我好运~ #threejs
39
66
10
658
31,875
Kevin Hu retweeted
小西天,看着像视频,但其实是我们在现场实地拍摄 3,811 张 206 GB 的照片后建模的。FUNES 把《黑神话:悟空》里「既见未来,为何不拜」满天神佛的原型,来自自山西临汾隰县的小西天,做成了一个可漫游的 3DGS 数字存档。 完全实地拍摄,每天清晨一开门就冲上山去,趁着没人的时候拍。然后通过 Gaussian Splatting 重建,没有手工建模,尽量保留真实悬塑和圆塑的极其密集的金色空间、细节和光感。不同的材质在这里交织成了无法分辨的一个天国世界。这种半空中的小塑像是「悬塑」,它们大多出现在十六世纪到十七世纪。 在现场如果要看清小西天的所有细节,我想大概需要三天时间。但是有了模型,我们可以在屏幕前慢慢看。在相当长的时间里,学术界并没有特别重视小西天这样的悬塑——因为在只有学术图录的年代,平面印刷无法展示出悬塑的震撼。而随着技术的进步,我们终于可以在远方一窥明代悬塑的璀璨。 重轻特意为这个模型做了配乐,大家可以打开慢慢欣赏。 推荐电脑访问:funes.world/apps/the-hanging…
600
749
90
3,517
929,631
the price to create software is collapsing now
This is the doom I predicted a few days ago, coming for Photoshop. A clean-room open-source reimplementation. No prizes for guessing that they decompiled Photoshop to source code, processed that to some kind of non-code specification language, then fed the spec to an LLM with an instruction to generate Rust. Adobe just got nuked. And closed source is dead, dead, dead. github.com/storytold/photocr…
40
u literally did the meme
What the f*ck is this drug trip I told Opus 5.5 to make a cow that runs. This is what it gave me for $6.80. It runs on its f*cking udder. Four teats as legs. Then it jumps a fence in slow motion and lands like a pancake.
70
👀
Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.
22
教 AI 練習書法中
1
1
92
更多書法:
1
76
more calligraphy animation work:
20
The Bitter Lesson explained with a music video. Extraordinary.
13
39
11
162
18,232