AI Makerspace

What we’re 🏗️ building, 🚢 shipping, and 🚀 sharing this week: Cursor: An AI Engineer’s Guide 🧑💻 In 2025, the best AI Engineers are going beyond VS Code. They’re picking up Cursor: The AI Code Editor. But it’s not just for AI Engineers! 🧪 We’ve been testing Cursor inside The AI Engineering Bootcamp, working with experienced engineers, up-and-coming devs, and even AI Engineering leaders. 🏗️🚢🪨 Verdict? Cursor is now a best-practice tool that can help anyone get into building, shipping, and sharing AI applications. You’ll learn: 🧑💻To set up an interactive dev environment 📦 To set up uv for package management 🖱️ Explore Cursor and its functions 3️⃣ Levels of using Cursor: write code, chat, and agent 🫨How to vibe code and set up a simple application We’ll even leverage the new Claude 3.7 sonnet from Anthropic, voice interaction, and model context protocol. The future of AI engineering is here. Are you ready? 📅 RSVP now and step into the next era of coding.  https://lnkd.in/ge_BGizA #AIEngineering #Cursor

AI >> 3D worlds (Prompt: "A fantasy Tavern") Back to posting about my AI that works with 3D models! Imagine what would happen if you combined a huge 3D model database, LLMs, RL, and zero shot image classification (as an RL reward)? You get my approach to 3D world building using AI! We've got lots of great 3D assets out there already, or if you're a game studio, you're very likely to already have a lot of great assets on hand. The trick is teaching an AI to put them together in a way that looks visually meaningful. I've seen a ton of people talking about an AI that generates CUDA - this is a very similar technique where it explores the 3D model space using LLMs and reinforces the choices that generate the most reward from the vision model. #AIWorldBuilding #GameDev #3DModeling #VirtualEnvironments #MachineLearning AI Makerspace

Why fine-tune? Fine-tuning embeddings is one of the easiest and most straightforward ways to supercharge your retrieval pipeline, especially when your data uses language nuances that the original model missed. Learn how to customize your models to unlock better accuracy and performance for your specific datasets. Watch now and dive into the world of #Finetuning, #Embeddings, and smarter #Retrieval pipelines! #AI #MachineLearning #DeepLearning #DataScience #NLP #TechInnovation #ModelOptimization #CustomAI #TechTips #Innovation #DigitalTransformation #BigData #Coding #ML #ArtificialIntelligence AI Makerspace

I’m excited to share Seek3D—my AI Makerspace midterm project! 🎮 It’s a specialized 3D model search engine that helps video game-level designers quickly find high-quality, “good-enough” assets (no awkward scans, off-topic geometry, or bare-bones textures) that fit their game prototypes. Powered by GPT-4o-mini for text-to-model understanding and open_clip’s ViT-H/14-378 embeddings for accurate retrieval, Seek3D ranks and displays relevant 3D assets in a clean interface. Check it out on Hugging Face and let me know what you think! 🚀🔎 https://lnkd.in/gEQHnGFF #AI #MachineLearning #GameDev #3DModels #AIMakerspace #SearchEngine #HuggingFace

10,581 emails... ...that's the number of emails I received in the last 6 months. Roughly 1,700 per month. And that's excluding emails from clients. Most were either cold outreach (99% of which was irrelevant) or newsletters. Here is a list of newsletters I'm keeping as I return to work: 1. Biotech & Pharma news - BowTied Biotech - FirstWord Pharma & their HealthTech section - DIA - Fierce Biotech (and all their sections - the OG of biotech news) - Endpoints News (great but I would love more free articles) - TLDR Biotech (the only newsletter I read almost every day) 2. Professional development - Section (my favourite e-learning platform) - Maven (great free mini-courses) - Sales Gravy | Sell More (tried and true selling techniques) - Wes Kao's newsletter (do everything she says, and you won't regret it) 3. AI news & AI-Sales - AI Makerspace (very technical but I learn a lot) - Allie K. Miller's newsletter (you must subscribe) - Ruben Hassid's newsletter - ColdIQ's newsletter (need I say anything? :)) - The Clay Operator And here's what I'm ditching: - Harvard's program on negotiation (never really learned much, terrible UX) - 3 newsletters on AI that kept selling courses and provided very little value - Beauty-store newsletters I subscribed to to get a discount (sorry not sorry :)). What's the one newsletter you can't live without?

What we 🏗️ built, 🚢 shipped, and 🚀 shared last week: Reasoning in Continuous Latent Space 💡 Big Idea: We don’t need to use tokens directly for reasoning! We explored two approaches: 1️⃣ Chain Of CONtinUous Thought (COCONUT) 2️⃣ A Recurrent Depth Approach We learned: 🥥 COCONUT showed that we can reason in latent space 🕳️ Recurrent Depth hypothesizes that it might even be more efficient 🎦 https://lnkd.in/giwE_Dk3 Stay tuned for more Large Reasoning Model (LRM) content coming soon!

Thoughts from my degree #1: what makes usability testing important for not just products, but also educational content and even documentation? An expert designer has the skills and knowledge to cater to their users, but without usability testing, a piece will always be missing: the user's perspective as a novice. Why? Well... Although usability and inclusive design was one of my favorite Industrial Engineering subjects, the key which took my understanding of it to another level was actually a Psychology class on human memory. Human memory is reconstructive, meaning that the information we store is a condensed set of concepts rather than the whole picture. When we access a memory, it is reconstructed on the spot using our current knowledge and emotions to fill in the blanks, making a new version. Then, that new version is what gets condensed back into long-term memory. Designers and educators go through this process over and over to become experts in their field. This means that new users and new learners have an elusive, fleeting, and irreplaceable perspective. A design, tutorial, or piece of documentation can be theoretically perfect, but incomprehensible to a true newcomer. Usability testing is a way to capture that new perspective in data to turn it into better designs, better communication, and better education. One of the little things I like to do when I write code and learn new tools is to keep a "dev-log" in the documentation as I go. Writing documentation to bring a senior developer up to speed is easy enough, but there is only that one opportunity to capture what I really needed while I was learning.

Follow along as Chris "The Wiz 🪄" Alexiuk takes us on a tour of latent space reasoning with Coconut! #genai #reasoning #learnai See the full demo at https://lnkd.in/e58RRGKR

🕴️ PydanticAI: making it less painful to build production agent apps! The team behind Pydantic, the most widely used data validation library, just dropped PydanticAI—and now they’re officially in the agents and orchestration game! Everyone’s asking: If you already use Pydantic, does that mean PydanticAI will be easy to pick up? 🤔 We’re putting it to the test! In this event, we’ll break down: ✅ How the Pydantic team built PydanticAI ✅ Key similarities & differences between Pydantic and PydanticAI ✅ How it stacks up against LangGraph, LlamaIndex Workflows, CrewAI, AG2, Swarm, and more And, of course, we’ll build, ship, and share an agent application! 🔥 📅 Join us to see PydanticAI in action! https://lnkd.in/geadMf4v #Agents #Orchestration

Friendly reminder for all GPT-4o users: https://lnkd.in/gYHbbXGE