Twitter Tea: Brewing Up ML Insights

Atharva,machine learningpythonaws

I came across a South Park Commons tweet saying Sam Altman is talking at SPC SF. Having read Paul Graham’s essays and followed YC’s journey religiously for the last 7 years, I decided I would attend. Not merely because it’s the guy with the most clout for productizing Gen AI to the world, but because he’s involved in quite a few industry defining companies. Its inspiring.

Turns out over the weekend there was a hackathon conducted by SPC along with OpenAI but the registrations for it were closed. Fortunately, will and james were looking for a teammate. We got on zoom and while some ideas like integrating Instacart with multimodal models didn’t seem feasible to develop over the weekend, twitter tea did.

Twitter Tea was born under the shared experience of having to keep up with the tech landscape everyday. Its a newsletter that would pull together the most relevant tech and machine learning insights from across the internet in the last 24 hours. You can check the one we tested here (opens in a new tab).

The frustration of trying to keep up with the constant flood of news and discussions online fueled our determination. We knew there had to be others out there like us, drowning in a sea of tweets and Reddit threads, craving a more organized and curated source of information.

So we got to work. First, we scoured Twitter, carefully hand-picking about 168 accounts of ML experts, researchers, and companies that we wanted to keep an eye on. Then we added Hacker News to the mix, since we knew that's where some of the juiciest tech conversations happen.

The real challenge was taking all that data and turning it into something coherent and useful. We leaned on some powerful tools like LangChain, Hugging Face, and Retrieval-Augmented Generation to help us sort through everything and find the most valuable insights. It was a lot of experimenting and fine-tuning, but eventually we landed on a format that felt just right.

As the hackathon went on, we found ourselves lost in a flurry of code, data, and endless revisions. Optimizing the algorithm, making sure the delivery would be reliable, hammering out all the little details - it was intense, but also strangely rewarding.

And when we finally got to present our creation, I couldn't help but feel a sense of pride. Twitter Tea wasn't just a technical exercise for us, it was a reflection of our own struggles to stay informed in this fast-moving industry. We'd taken a simple idea and turned it into something that could genuinely help people like us, who are trying to cut through the noise and find the most valuable insights.

As the hackathon wrapped up, I couldn't help but feel like we had finally arrived in the land of our dreams. For the last 48 hours, everyone around us was laser-focused on just one thing - building. No politics, no bureaucracy, just pure, unadulterated passion for creating something new. It was diet coke + code, a lifestyle I could honestly get used to. In that moment, surrounded by likeminded individuals who shared our sense of wonder and hunger for innovation, I felt a deep sense of belonging. This was our tribe, our people. I was home. It was a feeling I hadn't experienced in a long time - that electric combination of purpose, camaraderie, and the intoxicating high of turning an idea into reality. I couldn't wait to bottle up this experience and take it with me, to carry that energy and momentum forward.

Quite proud of what we built over the weekend. Here's to the next cup!