Joey Pardella (@RichardDaffyAI) header
Joey Pardella (@RichardDaffyAI) profile picture

I’m just looking for a handle

Joined March 2025
Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)Photo by Joey Pardella (@RichardDaffyAI)
クリエイターの(Any-Reputation氏)が、わずか97KBのHTMLファイル1つで泳げる水中世界を制作。 Claude Codeでコードを記述し、画像や3Dモデル、外部ライブラリは不使用。海底や約200万枚の海草の葉をコードで生成し、WebGPUで描画。 魚群やマンタ、沈没船が広がる海を探索できます。
8
26
4
284
15,321
Joey Pardella retweeted
距離フォグはやっぱり「ゲームの中にいる」って感覚強いなあ。なんでだろう? オブジェクトとか空間を包み込んでる感じがしないから?
思った以上に反応があったので同じ画角で取り直しました 反応が良かったので、時間が許せばグラフィック設定で切り替えられるようにしたい🤔 ← 本当は採用したかった疑似多重散乱フォグ → 軽量化のために変更したシンプルな距離フォグ #VRChat
5
19
9
310
37,848
[vibe-]codevember 2026 - Day 13 - Labyrinth spite.github.io/codevember-2… #threejs
8
19
1
213
8,307
Joey Pardella retweeted
What happens when you click a spider web? The spiders think it’s dinner 🕷️ Day 14 of #codevember: Tension Real web physics, super accurate spiders. Lure them in, or pull too hard and snap the web. Open sourced. Links to play and the code in the comments. @threejs @mrdoob @akella @reactthreefiber #codevember #threejs #webgpu
30
32
4
469
17,817
Joey Pardella retweeted
RoboValue-Dataset is now on ModelScope: fine-grained training and evaluation data for robotic value models. 🤖 modelscope.ai/datasets/RoboV… 📦 Sim and real manipulation: 3,535 expert episodes across 15 simulation and 20 real-world tasks. 🧠 Evaluate execution, not just outcomes: Covers task state, progress, failure recovery, and value consistency. 🔍 Stress-test value judgments: Counterfactual instructions, repeated states, recovery trajectories, and alternative task orders expose unreliable feedback. 🔄 Test generalization: Includes cross-embodiment and cross-environment settings, with Full-Shot and One-Shot subsets in LeRobot v3.0 format. 📜 Code, scripts, and documentation are released under the MIT License.
3
29
2,295
Hey all! Oboro no Ichi is doing even better than I expected, and I couldn't be happier! I've been reading all your feedback and suggestions, and I'm so grateful for your support! I recently added this little feature to the game! :D #indiegame #cozygames #cozygaming #indiegames
10
28
5
249
8,594
Joey Pardella retweeted
Codevember day 9: Membrane 🔊 This prompt triggered immediate memories of Toshio Iwai's Hanenbow Electroplankton Nintendo DS game, I reinterpreted it as a modern version of that legendary interactive piece. codevember-2026.felixmartine… Physics and collisions are computed on the GPU so audio is also GPU generated :) #threejs #webgpu #webaudio
7
10
119
4,665
Joey Pardella retweeted
Flying down Mount Fuji in Japan with Opus 5.5! Still pushing my Earth simulator to its limits. Real-world terrain, photorealistic landscapes, dynamic weather, all running in real time directly in the browser I didn't think my GPU could handle flying this fast with all those details, but somehooow it did! :)
I pushed Opus 5.5 to make the whole planet Earth look photorealistic at night with no loading time between day/night cycles Here I'm still in Paris in that video and.. Yeah, really happy with the result! It looks pretty cool I think! But I want to 100x everything.. So now I'm thinking about moving the project to Unreal Engine because I'm starting to feel limited by the browser One issue is that I'm rebuilding the lighting system, shaders, collisions, etc. myself, and I don't have access to NVIDIA tools like DLSS. On the other hand, Unreal Engine has Nanite, makes better use of my GPU and also has many other useful tools I could probably rebuild a smaller version of all that, but I feel like it'd be a waste of time and tokens? Let's see! :)
7
4
94
5,922
Joey Pardella retweeted
まだまだ建てたいものとか畑とかはあるけど 当初の目標は達成したので散歩してみた スムーズにカメラ回転させるの難しい 最初は宿屋に改装した既存建物しかなかったのが すっかり村になって満足 #Enshrouded
8
5
1
167
9,143
Joey Pardella retweeted
We’re releasing new Qwen3.8-27B GGUFs with 10% higher accuracy. Unsloth Dynamic V3 outperforms others by >10% on Div-300, KLD & more benchmarks. We also release 1-bit quants that retain 77% accuracy. Run on 8GB RAM. Blog: unsloth.ai/docs/basics/dynam… GGUF: huggingface.co/unsloth/Qwen3…
Qwen3.8-27B can now be run locally! ✨ Run on 17GB RAM via Unsloth Dynamic GGUFs. Qwen3.8-27B is by far the strongest model for its size. We also uploaded NVFP4 quants. GGUF: huggingface.co/unsloth/Qwen3… Guide: unsloth.ai/docs/models/qwen3…
223
604
167
5,171
1,179,377
Joey Pardella retweeted
the $8 closed wrapper guy is mad that a free, open source agent ships with too many skills. skills that toggle off in two clicks, individually or by category, and you knew that going in. so "every user is forced to have a polymarket skill" was never true. it was a config preference. you could've toggled it in five seconds, or if you think the defaults are wrong, opened a PR, the code's right there. you did neither, you turned it into a public shit pile on to farm a thread. the funny part is you're calling an open, readable, PR-able agent "bad at feedback" from behind t3 chat, the closed $8 proxy that runs on other people's models, hides its own caps, and that you won't even open source, you ran a cloneathon so strangers would build the open version for you. in hermes agent i read the skill, switch it off, or fix it. in t3 i can't see a single line. which one's actually built for the user? you didn't review hermes agent. you reviewed yourself.
Hermes Agent comes with a truly absurd number of skills pre-enabled. Over 100 of them. This is roughly half. I get what they're going for - they want an agent that comes "ready out of the box". I just don't get why every user has to have a polymarket skill, 3 baoyu art skills (? never heard of this), a headless Pokemon skill, and Minecraft modpack server skills, all available the first time they run it. I guess Hermes Agent just isn't for me.
72
39
9
1,163
120,091