Glavin Wiechert👨‍💻

Built a YouTube to fine-tuned LLM pipeline 🚀 🎯Goal: Startup Mentor AI 🎙️72hrs speech-to-text @DeepgramAI: $20 ❓550 questions w/ GPT3: $0.12 🧠Train LLaMA 13B w/ Axolotl & LLM Studio on @runpod 📉Quantize/GPTQ: 52->7GB 🚀Deploy @replicatehq w/ @LambdaAPI Questions?😀

Built #decentralized multiplayer #Physics for the web! Can be used for #gamedev & #Metaverse style #learning with interactive #VirtualReality scenes (eg. @BehindEng). Built with yjs.dev & @webrtc, #ammojs @pmndrs ‘s #r3f #threejs #ReactJS #React #webdeveloper

We drove 410% conversion lift on @Replit's enterprise funnel, 2x-ing the number of demo requests. All in a matter of months. "Coframe makes you feel like you have a team of 100 people. There's nothing on the market like it." ⧉ x ⠕

Introducing Alvea! An adaptive AI assistant creating unique user interfaces personalized to your tasks just in time. Your evolving tool, designed to enhance productivity and user experience 🚀 Won 🥇 @agihouse_org w/ @sockcymbal @narphorium @vhurgenes bit.ly/GitHub-Alvea-AI

Critically helpful context to know it is only for dedicated deployments, thanks! I interpreted the “Get in touch with our team for dedicated deployments of GLM-5.2, or get started testing the model today with our model API.” at the end of article meaning normal model API would offer the same capabilities.

@GlavinW @arjundivecha Hey Glavin, speed is going to depend on workload profile and overall system load. For an SLA at these speeds we would need to do a dedicated deployment

Automated #javascript coding teacher with real-time feedback: a guide to solving the Count Legal Drivers exercise. Try the demo of my Proof of Concept: bit.ly/3UIEznM #buildinpublic #education #javascript #web #automation

@arjundivecha @philipkiely Same here via their model API. I just signed up to test. Very interesting in using 250+ toks/s GLM-5.2! Has anyone else reproduced reported speed?

@philipkiely Respectfully: I can't reproduce the 250+ tok/s claim yet. Direct to Baseten's API (no proxy), using server-reported completion_tokens, I'm seeing ~90-115 tok/s with ~300ms TTFT on GLM-5.2. That's strong, but materially below 250. Happy to rerun with your exact config.

Wiring up a programmable & battery powered motor circuit Want to learn how? 👉 Join the waitlist: bit.ly/3EbAGBP @BehindEng #buildinpublic #electronics #hardware #engineering #threejs #r3f #javascript #React #webgl

@krandiash CARTESIA

We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today. New architectures enable new frontiers for speed and quality. We're now the only provider to have #1 models for both speaking and listening.

Love seeing how far the open models have come! For those looking to reproduce your setup, essentially 1. @lmstudio to generate batch of responses with models, 2. Claude opus 4.6 then to perform pair-wise comparison to pick which is best, 3. Custom UI which looks like LMArea to visualize?

Back in March, I tested whether Qwen3.5 4B was as good as GPT-4o for gen chat. The result was that they were ~similar. But lots of models have since come out - Qwen3.6, Gemma4 - how do they stack up? The answer is decisive: they beat GPT-4o handily for general LMArena-like chat.

@hybridherbst @SeloSlav @sea3dformat @threejs @NeedleTools This is so awesome! Is there any progress or ETA on native three.js integration? I want to add to my existing three.js projects.

@SeloSlav @sea3dformat @threejs 😄 wonderful model. You either grab the source from the @NeedleTools three.js fork and hack around with it, or you wait for a proper three.js integration into TSL, or you use an alternative like Needle Progressive 3D (not quite Nanite but best web streaming solution for glb)

Casually rendering ~a hundred million triangles, in the browser. All thanks to @sea3dformat's cool Nanite-style renderer for @threejs. I've added GLB-to-meshlet and PBR rendering to it, very exciting to see where this is going!

I'm vibe coding a new *GPU rigid body physics engine* from scratch for @levelsio 's #vibejam ! 💥 - Supports Rust & Web (via WASM+WebGPU) - 100% written by @anthropic 's Claude Code + Opus 4.6 - Performance optimized by @karpathy / @uditg 's autoresearch: github.com/uditgoenka/aut… - Uses Augmented Vertex Block Descent (AVBD) paper: graphics.cs.utah.edu/research/proje…

Startup Mentor AI v0.2 is a massive leap forward in quality & helpfulness! 🎉 ✅ 100% automated pipeline from raw text/YouTube to "textbook" high-quality synthetic dataset /w @langchain 📈 Output (right side in screenshot) appears much better to me! What do you think?

@InOneProjects Thank you so much!! I was looking forward to polishing this game, however, I ended up having to travel for work the last couple weeks of #vibejam so it is still far from finished. Really appreciate the kind words! Hopefully I’ll have time soon to revisit and polish

🏆 Vibe-Land @GlavinW: This game impressed me with it's very serious approach to being an FPS - the tutorial does a live "how does this feel?" calibration to help you dial in your aim. With some sound and improved visuals this is going to be seriously impressive.

Five more #vibejam mini reviews: I found 2 standout picks in this set, plus the usual mix of rough edges and good ideas. 👇

Learn how to train custom language models step-by-step! 🔗 New Repo: github.com/Glavin001/Expe… Currently includes my scripts to: 1️⃣ ✅ Bulk download YouTube videos 2️⃣ ✅ Convert audio to text with @DeepgramAI 3️⃣ 🚧 Convert unstructured text into chat format More soon 🚀

@ericclemmons Cloudflare artifacts is an awesome idea in the time of agents and massive scale code production. Any way to get beta access? My company is interested in using it for version control of our agents’ output code.

Couldn't sleep until I built this tonight 😅 $ 𝚐𝚒𝚝 𝚙𝚞𝚜𝚑 𝚌𝚕𝚘𝚞𝚍𝚏𝚕𝚊𝚛𝚎 ☁️ Host your git repo with Cloudflare Artifacts ➡️ Push to CI/CD app on your account 🔍 Agents/Humans can fix CI from the terminal

@dillon_mulroy @threepointone Cool! I’m also really excited about using Cloudflare Artifacts in our @coframe agent tech stack! Unfortunately, don’t have access yet. I’d love to replace our custom sandbox version control with git powered by artifacts

this is so freaking cool

@dangreenheck What is this?! 🤩

Gloop 🫠

@ps_starke @arnputz Would it be possible to process control inputs to return pose updates entirely in browser?

@arnputz the client handles rendering, user input, and sends control inputs to the server who returns pose updates from the single cpu inference

We now have new demos in AI4AnimationPy for stylized biped and quadruped locomotion control. Besides PC you can also run them in your browser or on your phone. Try it out here🎮: Web-Demo: paulstarke-ai4animationpy.hf.space GitHub: github.com/facebookresear… Video: youtube.com/watch?v=qifmLp…

Server is still going strong! 💪

My #vibejam game is finally taking form! 🎉 Join multiplayer at vibe-land.glavin.ca/play - 🚗 Race & jump vehicles - 🔫 Raygun Fights - 🌎 Open-world map for all players More updates coming soon!

Vibe designed this new WIP world using AI chat as editor! 🎮 Play it here: vibe-land-eosin.vercel.app/practice/share… #vibejam