🚀 Cut Latency, Ditch the Cloud: Our Hottest AI Tinkerers Demos - July 28
Issue #1 · Week of July 28
Recent live demos from London, DC and Tiruchirappalli: a monetized Remote MCP on Cloudflare Workers + Supabase + Stripe; Aura, a local Qwen2-VL + Mistral 7B gaming companion with sub-3s responses; and Survaize, vision/OCR -> JSON -> ODK. All ran live with repos; these are the demos we’re still thinking about.
Sourcebot: Agentic Codebase Understanding
Brendan Kellam from Sourcebot demo’d Sourcebot, a self-hosted tool to help you understand your codebase. Sourcebot provides flagship reasoning models the ability to search and navigate your codebase, allowing them to answer complex natural language questions. It summarizes these findings in a modern web UI that includes inline citations, allowing the answers to be grounded in navigable references into your codebase.
TECH STACK
|
Helpful Computer
Helpful Computer was a demo that showed off an interactive voice agent interacting with Excalidraw and Lexical, all built via Tauri. Ron, a seasoned tech leader from ITS, presented it and demonstrated how a voice interface could effectively collaborate with creative tools. People loved its practicality and clear demonstration of agent use, hinting at business potential in streamlining creative collaboration and everyday communication for both startups and larger organizations.
TECH STACK
|
Voice Trivia Bot
William showcased the Voice Trivia Bot demo, an app that read quiz bowl questions aloud from PDFs and let users buzz in with audio answers that were graded against the official responses. William, a staff engineer with a solid background in AI and healthcare at Maven Clinic, built the app using OpenAI APIs and Cursor with LLM coding assistants. People loved how it combined multimodal interactions with a playful twist, hinting at practical applications in interactive testing and self-paced learning.
TECH STACK
PROJECT LINKS
|
HackerNews.coffee: transparent fast personalization
Aleksi Tukiainen and Leo Browning from ClimateAligned presented HackerNews.coffee, a lean AI recommendation system that transparently personalized content by bootstrapping minimal user data. The demo showcased a practical tradeoff between fast and smart LLMs while keeping recommendations editable for users. It was a neat example of how you can ship a working system rapidly without sacrificing user control (people loved it). This practical approach offers valuable insights for builders facing similar personalization challenges.
PROJECT LINKS
|
Living Systems: AI Plant Automation
Aditya Prabhu presented his project, Living Systems, which built a smart plant automation system using LLMs and MQTT. The demo showcased a practical solution that controlled pumps, lights, and water levels through a React and Python interface. Drawing from his hands-on engineering background, Aditya impressed many with his down-to-earth approach. People loved how it tackled real indoor farming challenges, showing clear potential for a product that makes everyday grow shelves smarter and more connected.
TECH STACK
|
🚀 Cut Latency, Ditch the Cloud: Our Hottest AI Tinkerers Demos - July 28