Satvik Paramkusham

Anthropic just dropped Sonnet 5.5 and I think it's on par with Opus 5.5 at motion graphics. So I used my motion graphic skill and gave it one brief, "impress motion designers", and had it direct a small crew of subagents to make a 97-second film. Every frame is a WebGL2 shader. All the music is synthesised in WebAudio. No images, video files, samples or libraries. Skill link in replies btw 👇 How is your experience with Sonnet 5.5 so far?

You WON’T believe Fable 5.1 built this in HTML ONLY. 🤯 An epic Ramayana-inspired cinematic where a forgotten jungle temple awakens a colossal divine vanara guardian. It looks like a movie, but it’s all made with @threejs 100x better than what Gemini 3.8 Flash created with the same prompt. Fable is still the best model I’ve tested so far. Now testing the exact same prompt on GPT-6 Astra. 👀

Tested Ox Alpha head to head against Gemini 3.7 flash, Kimi K3 and Opus 5 on AAA fps shooter prompt. here is the result : honestly, I’m most surprised by Gemini 3.7 Flash. It built the entire thing in just 4 minutes while still delivering seriously impressive quality. Gemini 3.7 Flash is way more underrated than it gets credit for 👀 And the huge usage limits in Antigravity make it even better for these kinds of experiments. @GeminiApp 🔥

Qwen ai just dropped Qwen 3.8 27B, so we tested it head-to-head against Fable 5 on our @threejs Spider-Man game prompt. The results are honestly shocking: A 27B-parameter model that can run on an RTX 3060, with no censorship, API bill, or rate limits shipped a playable demo that was almost as good as the #1 frontier model. AI is getting crazy. Try it live : genaiexperiments.vercel.app Now Mom can finally say: “No, we have Spider-Man at home.” 😂📷

Here's our version inspired from this, a nostalgic railway journey playlist : live at : genaiexperiments.vercel.app/…

i gave claude fable one prompt and got back a working node-based image editor. café studio: you drop in a coffee or food photo, wire nodes together on a canvas, and gemini's nano banana models texturize it into an instagram story. three things that separate this from every "AI builds an app" video you've scrolled past: 1. it's a node graph, not a form. canvas UIs are exactly where one-shot generation usually breaks - hit detection, edge routing, zoom transforms. it got them right. 2. it's a real model integration, not a stub. nano banana is doing actual image work, not returning a placeholder. 3. it's live, Try it yourself. Link in replies👇 What's the most complex thing you've gotten out of a single prompt?

This Codex skill can create an entire ecommerce experience for you, in one prompt. Give it a product or company brief, and it can generate the creative direction, original AI product images, responsive storefront, 3D interactions and demo cart. I used it with GPT 5.6 sol with a simple instruction : “Build a crazy sneakers ecommerce website with 3D effects. Output a single HTML file.” This was the result👇

🚀 Unrot 2.0 is live on the Play Store! And we just crossed 10,000 downloads. Unrot is our microlearning app for Generative AI. 5 minutes a day. That's it. We built it for two kinds of people: - Those who want to learn AI but don't know where to start - Those who started but got left behind by how fast things move Open the app. Spend 5 minutes. Walk away knowing something new. Core concepts, latest news, interview prep - all in sessions that actually stick. Give Unrot 2.0 a try 👇

There are plenty of AI SDKs & frameworks to use but nobody's really knows about how they actually fit together - and that's where real AI systems get built. actually you don't have to pick one. you can layer them. SDKs → fast, direct model calls LangGraph → multi-step logic with memory CrewAI → multiple agents, different roles MCP → connects AI to your actual tools and data Here is the example of AI stack you can built 👇 what's your favorite AI SDK or framework? 💬

OpenAI just acquired @promptfoo - the open-source LLM eval tool used by 25%+ of Fortune 500. Still open-source. Still model-agnostic. Here's why every dev shipping LLM apps should care (7 use cases with GPT-5.4) 🧵

Everyone’s hyped about the @elonmusk × @nikhilkamathcio podcast. But learning? Not really. Using Educhain , you can turn the podcast into a quiz: • Drop the YouTube link • AI analyzes the full 1-hour conversation • Generates a complete MCQ quiz Full code 👇 #ElonMusk

Nano Banana Pro is wild. Take any research paper or long article → turn it into a detailed whiteboard photo. It's basically the greatest compression algorithm in human history. 🤯 Visual learning just changed forever. #AIart #Gemini3 #NanoBanana2 #TechTwitter #nanobanana

Gemini 3 is absolutely insane. People can't stop building things. 10 wild examples: 1. Remade a 90s theme park game in hours By @demishassabis

🚀 Build MCP Servers in <1 min — No Code Needed! With @getpostman's MCP Generator: - Pick from 100k+ APIs - Export server code - Drop keys in .env - Connect to your MCP client From idea → production-grade server instantly. Link in comments 👇

🚀 Meet Your New AI Search Assistant Powered By @mem0ai + @tavilyai - Remembers your preferences with smart memory - Finds exactly what you need with real-time search - Chats contextually & evolves with you Code Below👇

🚀 Turn ideas into Videos instantly! With our open-source @nextjs template, you get: 🎥 #Veo3 (text → video) 🖼️ Imagen 4 (image → video) ✂️ Timeline edits in-browser All powered by @Google Gemini API.

We created an open-source version of Gemini Storybook using @aisdk! 📖 – Uses 'gemini-2.5-flash' to generate story – 'qwen-image' for image generation via @FAL - 'page-flip' library for page flip effect Check out the repo & give a ⭐ -> github.com/buildfastwithai/s…

♟️ AI vs AI Chess Showdown! Built with @AgnoAgi , Where Gemini 2.0 and GLM 4.5 By @Zai_org Battle It Out On The ChessBoard 🤖 - Choose Your AI Models - Start A Match - Enjoy The Live Showdown Code Below👇