Aadit Sheth’s Post

Claude 3.5 Sonnet is the new favorite AI tool! Everyone is creating amazing things with it. 11 incredible examples: _____ 1. Making fun games with Claude 3.5 Sonnet is now very simple. 2. What once took hours to code can now be done in minutes. 3. Claude 3.5 Sonnet is better than GPT-4o, Gemini Pro, and Llama. 4. Claude 3.5 Sonnet now creates fun games to play while you work. 5. Claude 3.5 Sonnet is the best interactive model as compared to other LLM models 6. Claude Sonnet 3.5 is the best reasoning model so far. 7. Claude Sonnet 3.5 can generate a web app in just 25 seconds. 8. 3D solar system with collisions made using Claude Sonnet 3.5 9. Claude Sonnet 3.5 can accurately pick colors in the game. 10. Complex data analysis is now even simpler. 11. Claude 3.5 Sonnet can generate sounds _____ #ai #claude #sonnet

🚀 𝗢𝗽𝗲𝗻𝗔𝗜 𝗷𝘂𝘀𝘁 𝗱𝗿𝗼𝗽𝗽𝗲𝗱 𝗼𝟭, also known as 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗦𝘁𝗿𝗮𝘄𝗯𝗲𝗿𝗿𝘆/𝗤 — a groundbreaking AI model that can "think" and "reason" before responding! 🤖✨ Here’s what makes it next-level:    1. 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝘀 𝗩𝗶𝗱𝗲𝗼 𝗚𝗮𝗺𝗲𝘀 from just a prompt 🎮   2. Explains 𝗤𝘂𝗮𝗻𝘁𝘂𝗺 𝗣𝗵𝘆𝘀𝗶𝗰𝘀 like never before 🌌   3. Handles notoriously tricky 𝗔𝗜 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 with ease 🧠   4. Shows enhanced 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 and logic 🌱   5. Breaks down complex 𝗘𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 concepts 📊   6. Solves 𝗟𝗼𝗴𝗶𝗰 𝗣𝘂𝘇𝘇𝗹𝗲𝘀 in real-time 🧩   7. Advanced understanding of 𝗚𝗲𝗻𝗲𝘁𝗶𝗰𝘀 🧬   8. Perfectly 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗲𝘀 𝗖𝗼𝗿𝗿𝘂𝗽𝘁𝗲𝗱 𝗦𝗲𝗻𝘁𝗲𝗻𝗰𝗲𝘀 🌐   9. Tackles high-level 𝗠𝗮𝘁𝗵 𝗣𝗿𝗼𝗯𝗹𝗲𝗺𝘀 🧮   10. 𝗖𝗼𝗱𝗲𝘀 like a pro 💻  Check out the first video featuring AI's new 𝗰𝗼𝗱𝗶𝗻𝗴 𝗮𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀—a full video game built from scratch, just from a prompt! 🎮👾 Reference: https://lnkd.in/d7f3WsDN #OpenAI #ProjectStrawberry #AIRevolution #MachineLearning #AIReasoning #NextLevelAI #Tech #SoftwareEngineer #BestVideo #OpenAIModel

a"8-Puzzle Game using A* Algorithm" project: 🚀 Solving the 8-Puzzle Game with the A Algorithm!* 🧩 Excited to share my latest project where I implemented the A* (A-Star) algorithm to solve the classic 8-puzzle game. This project combines problem-solving with artificial intelligence to tackle a fun yet challenging sliding puzzle. 📍 Key Highlights: 🛠 Algorithm: Used the A* algorithm with heuristics like Manhattan distance for efficient pathfinding. 🧩 Puzzle Goal: Arrange tiles in a 3x3 grid to match the target configuration. 📊 Outputs: Visual step-by-step solution of the puzzle. Performance metrics (like explored nodes and execution time). 🌟 Applications: Great for understanding AI search strategies and heuristic optimization. 🔰 What I Learned: The power of heuristics in guiding search algorithms. Balancing efficiency and accuracy in problem-solving. The importance of debugging when implementing graph traversal techniques. 😅 #AI #ArtificialIntelligence #8Puzzle #AStarAlgorithm #Pathfinding #MachineLearning #CodingProjects

FLUX.1 is bringing out the same thrill I had when SD1.1 and Midjourney V3 was released. We are hitting new milestones fam! 🚀 Flux.1-dev Upscaler ControlNet just dropped, and it's something to definitely look forward to: +Low-res in, high-res out +Faster upscales +Depth and normal map integration This research was released by Jasper AI @heyjasperai Built on top of FLUX.1-dev by Black Forest Labs – we are seeing a whole new wave of visual content solutions with this open source model. ♻️ Repost if you belive this is true --- P.S. Want to turn your ideas into stunning AI-enhanced visuals in record time? Join the waitlist to our 3-week AI x Creative Accelerator: https://lnkd.in/dgXFcBbZ #ai #FLUX #midjourney #aiupscale

🚀 𝗢𝗽𝗲𝗻𝗔𝗜 𝗷𝘂𝘀𝘁 𝗱𝗿𝗼𝗽𝗽𝗲𝗱 𝗼𝟭, also known as 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗦𝘁𝗿𝗮𝘄𝗯𝗲𝗿𝗿𝘆/𝗤 — a groundbreaking AI model that can "think" and "reason" before responding! 🤖✨ Here’s what makes it next-level:    1. 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝘀 𝗩𝗶𝗱𝗲𝗼 𝗚𝗮𝗺𝗲𝘀 from just a prompt 🎮   2. Explains 𝗤𝘂𝗮𝗻𝘁𝘂𝗺 𝗣𝗵𝘆𝘀𝗶𝗰𝘀 like never before 🌌   3. Handles notoriously tricky 𝗔𝗜 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 with ease 🧠   4. Shows enhanced 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 and logic 🌱   5. Breaks down complex 𝗘𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 concepts 📊   6. Solves 𝗟𝗼𝗴𝗶𝗰 𝗣𝘂𝘇𝘇𝗹𝗲𝘀 in real-time 🧩   7. Advanced understanding of 𝗚𝗲𝗻𝗲𝘁𝗶𝗰𝘀 🧬   8. Perfectly 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗲𝘀 𝗖𝗼𝗿𝗿𝘂𝗽𝘁𝗲𝗱 𝗦𝗲𝗻𝘁𝗲𝗻𝗰𝗲𝘀 🌐   9. Tackles high-level 𝗠𝗮𝘁𝗵 𝗣𝗿𝗼𝗯𝗹𝗲𝗺𝘀 🧮   10. 𝗖𝗼𝗱𝗲𝘀 like a pro 💻  Check out the first video featuring AI's new 𝗰𝗼𝗱𝗶𝗻𝗴 𝗮𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀—a full video game built from scratch, just from a prompt! 🎮👾 Reference (Twiiter): https://lnkd.in/dDW3rj7C #OpenAI #ProjectStrawberry #AIRevolution #MachineLearning #AIReasoning #NextLevelAI #Tech #Softwiz #SoftwareEngeering

Inspired, after attending the Arize AI conference at SHACK15 where leading experts discussed how to program LLMs among other advances in LLM tech. Here's a summary of key approaches for those working on implementing AI Agents. 1. DSPy - Similar to PyTorch, a powerful framework for programming LLMs. Zenbase AI (YC S24) is a promising startup by Cyrus Nouroozi and Amir Mehr that uses DSPy to Automate Prompt Engineering. I think we'll see this become a bigger opportunity. 2. Naologic founder Gabriel P. shared an alternative approach where there is a router that breaks the prompt into sub prompts to specialized AI assistants that you can configure to work in serial or parallel. His approach hits home because it addressed several obvious implementation challenges. 3. High density conversation (and picture) with Chi Wang the Founder of Auto-gen from Microsoft. The release of Auto-gen Studio is going to be super helpful in more people being able to do numerous iterations. Check out Auto-gen 'Nested Loops' that will answer several open questions. 4. Also, there is the co-pilot approach where an agent has assigned multiple tasks and is a different implementation than having multiple AI agents that interact with each other. Thanks Manuel Del Verme for sharing your insights. 5. Finally, awesome discussion with Vipul Gagrani on implementing based on query intent to get complete control for your use case and not use 'heavy' agentic frameworks. Thanks for explaining how You.com can be used to provide additional context to the agents. There are a few other approaches crewAI, Langbase, LangChain that are relevant for anyone looking to implement AI Agents. Humanic AI system is well-positioned to implement a "production grade" agentic architecture to solve challenges in Marketing Automation. Happy to chat about how to build an agentic architecture for your use case. Interested? DM me! #AI #ArtificialIntelligence #LLMs #Conference #Programming

A new image generation model called “red_panda” is making waves, outperforming popular models like Midjourney, Black Forest Labs' Flux1.1 Pro, and OpenAI’s image tools on the crowdsourced Artificial Analysis benchmark. With a lead of 40 Elo points over Flux1.1 Pro, red_panda is currently ranked highest in this text-to-image competition, where models are tested head-to-head on user-selected prompts. Artificial Analysis, similar to Chatbot Arena’s approach, uses a system where users evaluate images based on how well they match the given prompts. The model’s standout performance has sparked curiosity about its origins and underlying technology. #AIArt #ImageGeneration #RedPandaModel #ArtificialIntelligence #CreativeTech https://lnkd.in/gpdh2Rk3

There's a world in which AI overtakes game development... ...a world in which will non-technical teams know how to clearly express their technical needs. Suffice to say, I think devs are safe for the 50 years to come. + there's no certainty regarding the objectivity of AI models, as they've been known to replicate the bias in their training data... (By the way, we're already able to express our technical needs to a machine for it to compute a certain way. That's called code.) Thoughts? #GameDev #AIinGames

🌈 Exciting Developments in AI! 🌐 I'm thrilled to share my latest project using TensorFlow.js with the COCO model for real-time object detection on the web! 🎉 🔍 What is it? TensorFlow.js is a powerful library that allows us to run machine learning models directly in the browser. The COCO (Common Objects in Context) model is pre-trained to recognize 80 different object categories, making it incredibly versatile for various applications. 💻 Why TensorFlow.js? Accessibility: No backend server needed everything runs client-side! Real-time Processing: Instant feedback and interaction for users. Cross-Platform: Works seamlessly on any device with a browser. 🌟 Use Cases: Smart image tagging Interactive web applications Real-time video analysis 🔗 the complete guide https://lnkd.in/gBTV3e_J #TensorFlowJS #MachineLearning #ObjectDetection #WebDevelopment #AI #Innovation

A short introduction to AI Web Explorer -- one of many research projects we have been working on at profiq. Our goal is to create a very smart web scraper that can do much more than just follow links. https://lnkd.in/etMu27zv As a side note, creating videos like is harder than many people might think. Big respect to all creators doing this full time. #ai #llms #webscraping