I've always believed that the fastest way to learn something is to get in the arena and build. Ever since creating Plio, I realized how fulfilling it is to see an idea move from a visual in my head to a living product. AI has lowered the barrier to coding, so whenever I am consumed by an idea, I try to spin up a quick prototype. Here are some of my experiments:
In early 2023, I built a simple web app at the time of GPT 4's release. It let you upload documents and generate a shareable link so others could chat with that text.
It was a scrappy forerunner to the "context-aware agents" now taking off. I strongly believe memory and context are crucial to unlocking AI's deeper potential. There's a massive difference between a general AI model and one that's specifically tuned to your personal or professional knowledge. If we can harness our proprietary context - think your personal notes - and feed it into AI, we unlock a new layer of economic opportunity.
Imagine each person monetizing their unique knowledge set through an agent fine tuned on their proprietary personal knowledge.
In 2024 I built a voice-AI tool designed to help students prep for M7 business school interviews. I was blown away by the power of speech to text models and wanted to learn more. Talking out loud makes learning more fluid.
When you converse with AI in real time - like a practice partner that never tires - it's way more engaging.
Textbooks and text based LLMs are so passe :)
YouTube is my favorite product - hands down. It's the only marketplace I've seen that truly maximizes value for all participants: creators, consumers, brands, and the platform itself. It's also how billions of people, especially those in resource-strapped areas, learn new things.
But YouTube today is largely passive: click, watch, close. Inspired by my Plio days, I wanted to turn this browsing into active knowledge-building - let people talk to the video, take notes, build a knowledge graph of everything they watch.
I paused the project after NotebookLM launched video parsing, but one day I will get back to building the true co-pilot to YouTube.
In early 2023, I consulted for Hyperverge Academy and helped Aman and Gaytri design SensAI - an AI powered coding tutor to help students train and land software dev jobs.
Today, they have taken SensAI and built it into an open-source platform that educators can adapt for personalized, AI-powered learning of any subject. Insanely proud of where they have taken this.