Introduce Simillusion as an organization

README.md

@@ -7,11 +7,19 @@ sdk: static
 pinned: false
 ---
 
-# Simulusion
+# Simillusion
 
-**World models, many algorithms, one PyTorch API.**
+**Open tools for building, training and sharing world models.**
 
-[Simulusion](https://github.com/ParamThakkar123/simulusion) is an open-source PyTorch library for world models: neural networks that learn how an environment changes, so an agent can plan or learn by imagining inside them. It puts the major world-model families behind one interface:
+World models are neural networks that learn how an environment behaves: given what an agent sees and the action it takes, they predict what happens next. Agents can then plan, learn and explore by imagining inside them instead of the real world. They're one of the most promising directions in AI, from games and video generation to robotics.
+
+Building with them is still hard. Every paper ships its own codebase, training loop and data format, and most results need a cluster to reproduce. **Simillusion exists to change that.** We build open-source tools that make world models easier to train, compare and use, on hardware people actually have.
+
+## What we're building
+
+### 🧠 Simulusion: the library
+
+[Simulusion](https://github.com/ParamThakkar123/simulusion) is our open-source PyTorch library that puts the major world-model families behind one API:
 
 - **Models:** Dreamer (V1/V2), PlaNet, IRIS, DIAMOND, Genie, plus DiT and I-JEPA as building blocks, with more being added
 - **One `WorldModel` interface:** `encode` / `step` / `rollout` / `decode`, the same for every family
@@ -22,9 +30,17 @@ pinned: false
 pip install simulusion
 ```
 
+### 📦 Pretrained world models
+
+The models on this page, trained with Simulusion and ready to download. Each model card links its paper, reports what the run reached, and shows how to load it.
+
+### 🚀 An end-to-end platform for world models *(coming soon)*
+
+Everything from data to training to deployment, in one place. Follow this organization or visit [simulusion.vercel.app](https://simulusion.vercel.app) for updates.
+
 ## Demo models
 
-Every model here was trained from scratch on a **single 4 GB laptop GPU** (RTX 3050) in one to two hours, by the scripts in [`demos/`](https://github.com/ParamThakkar123/simulusion/tree/main/demos). They show what each model does on a small budget; they are not reproductions of the papers' results. Each model card links its paper, reports what the run reached, and shows how to load it.
+Every model below was trained from scratch on a **single 4 GB laptop GPU** (RTX 3050) in one to two hours, by the scripts in [`demos/`](https://github.com/ParamThakkar123/simulusion/tree/main/demos). They show what each model does on a small budget; they are not reproductions of the papers' results.
 
 <table>
 <tr>
@@ -52,11 +68,11 @@ from simulusion.world_model import load_world_model
 wm = load_world_model(snapshot_download("simillusion/iris-atari-boxing"), "iris")
 ```
 
-## Links
+## Get involved
+
+We'd love to hear from researchers and developers working with world models, model-based RL, robotics or video generation. What would make our tools useful in your work? Which models should we add next? Open an [issue on GitHub](https://github.com/ParamThakkar123/simulusion/issues), or start a discussion on any model here.
 
 - 🌐 Website: [simulusion.vercel.app](https://simulusion.vercel.app)
 - 💻 GitHub: [ParamThakkar123/simulusion](https://github.com/ParamThakkar123/simulusion)
 - 📖 Docs and video gallery: [paramthakkar123.github.io/simulusion](https://paramthakkar123.github.io/simulusion/)
 - 📦 PyPI: [simulusion](https://pypi.org/project/simulusion/)
-
-Feedback is very welcome: tell us what would make Simulusion useful in your work, or which models we should add next, in a [GitHub issue](https://github.com/ParamThakkar123/simulusion/issues).

index.html

@@ -3,14 +3,18 @@
 <head>
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
-<title>Simulusion</title>
+<title>Simillusion</title>
 <link rel="stylesheet" href="style.css">
 </head>
 <body>
 <main>
-<h1>Simulusion</h1>
-<p><strong>World models, many algorithms, one PyTorch API.</strong></p>
-<p><a href="https://github.com/ParamThakkar123/simulusion">Simulusion</a> is an open-source PyTorch library for world models: neural networks that learn how an environment changes, so an agent can plan or learn by imagining inside them. It puts the major world-model families behind one interface:</p>
+<h1>Simillusion</h1>
+<p><strong>Open tools for building, training and sharing world models.</strong></p>
+<p>World models are neural networks that learn how an environment behaves: given what an agent sees and the action it takes, they predict what happens next. Agents can then plan, learn and explore by imagining inside them instead of the real world. They're one of the most promising directions in AI, from games and video generation to robotics.</p>
+<p>Building with them is still hard. Every paper ships its own codebase, training loop and data format, and most results need a cluster to reproduce. <strong>Simillusion exists to change that.</strong> We build open-source tools that make world models easier to train, compare and use, on hardware people actually have.</p>
+<h2>What we're building</h2>
+<h3>🧠 Simulusion: the library</h3>
+<p><a href="https://github.com/ParamThakkar123/simulusion">Simulusion</a> is our open-source PyTorch library that puts the major world-model families behind one API:</p>
 <ul>
 <li><strong>Models:</strong> Dreamer (V1/V2), PlaNet, IRIS, DIAMOND, Genie, plus DiT and I-JEPA as building blocks, with more being added</li>
 <li><strong>One <code>WorldModel</code> interface:</strong> <code>encode</code> / <code>step</code> / <code>rollout</code> / <code>decode</code>, the same for every family</li>
@@ -19,8 +23,12 @@
 </ul>
 <pre><code class="language-bash">pip install simulusion
 </code></pre>
+<h3>📦 Pretrained world models</h3>
+<p>The models on this page, trained with Simulusion and ready to download. Each model card links its paper, reports what the run reached, and shows how to load it.</p>
+<h3>🚀 An end-to-end platform for world models <em>(coming soon)</em></h3>
+<p>Everything from data to training to deployment, in one place. Follow this organization or visit <a href="https://simulusion.vercel.app">simulusion.vercel.app</a> for updates.</p>
 <h2>Demo models</h2>
-<p>Every model here was trained from scratch on a <strong>single 4 GB laptop GPU</strong> (RTX 3050) in one to two hours, by the scripts in <a href="https://github.com/ParamThakkar123/simulusion/tree/main/demos"><code>demos/</code></a>. They show what each model does on a small budget; they are not reproductions of the papers' results. Each model card links its paper, reports what the run reached, and shows how to load it.</p>
+<p>Every model below was trained from scratch on a <strong>single 4 GB laptop GPU</strong> (RTX 3050) in one to two hours, by the scripts in <a href="https://github.com/ParamThakkar123/simulusion/tree/main/demos"><code>demos/</code></a>. They show what each model does on a small budget; they are not reproductions of the papers' results.</p>
 <table>
 <tr>
 <td width="33%" valign="top"><a href="https://huggingface.co/simillusion/dreamer-dmc-cartpole-swingup"><img src="https://huggingface.co/simillusion/dreamer-dmc-cartpole-swingup/resolve/main/dreamer_dream.gif" alt="Dreamer" width="100%"></a><br><b>Dreamer</b>: DMC cartpole swing-up</td>
@@ -83,14 +91,14 @@ from simulusion.world_model import load_world_model
 
 wm = load_world_model(snapshot_download(&quot;simillusion/iris-atari-boxing&quot;), &quot;iris&quot;)
 </code></pre>
-<h2>Links</h2>
+<h2>Get involved</h2>
+<p>We'd love to hear from researchers and developers working with world models, model-based RL, robotics or video generation. What would make our tools useful in your work? Which models should we add next? Open an <a href="https://github.com/ParamThakkar123/simulusion/issues">issue on GitHub</a>, or start a discussion on any model here.</p>
 <ul>
 <li>🌐 Website: <a href="https://simulusion.vercel.app">simulusion.vercel.app</a></li>
 <li>💻 GitHub: <a href="https://github.com/ParamThakkar123/simulusion">ParamThakkar123/simulusion</a></li>
 <li>📖 Docs and video gallery: <a href="https://paramthakkar123.github.io/simulusion/">paramthakkar123.github.io/simulusion</a></li>
 <li>📦 PyPI: <a href="https://pypi.org/project/simulusion/">simulusion</a></li>
 </ul>
-<p>Feedback is very welcome: tell us what would make Simulusion useful in your work, or which models we should add next, in a <a href="https://github.com/ParamThakkar123/simulusion/issues">GitHub issue</a>.</p>
 </main>
 </body>
 </html>