AI Developer tool Hackathon: Experience dump

This weekend I participated in AI Devtool Hackathin in SF. We ended up building a “Always-on Debugger” that was judged best overall. Idea was simple — how can make debugging experience on terminal faster (and more fun).

The idea came from our personal experience. Samarth proposed this one on our way to attending the hackathon, and I was quickly I had plan to do something about this problem a couple of months back, and but couldn’t never get to it.

If you’re interested in idea, here are a few more details:

Repo: https://github.com/samarthaggarwal/always-on-debugger
Slides: https://docs.google.com/presentation/d/1u4KNxiYmSL67a6jMy85eplJH2V875MjlK17jhd0ttJY/edit?usp=sharing

Some takeways:

1. SF has unique network advantage. More than 50% of people knew each other from before and knew what products they were building! If you have something working here, it can just take off like wild fire.
2. Incredible how people were willing share the nuanced of product stacks, their approach to dealing with LLM challenges, business challenges … That’s typical characteristic of people believing in building and infinite games. This helped me learn so much.
3. Our idea was much simpler, less work. In fact, we were joking that we might be doing something wrong, as we didn’t have to code much. But that was a major revelation: people like simple, focussed product. One person told us they liked that our product does one thing and does it really well!
4. Many also told us that this should already exists or someone has built it. We did check and couldn’t find it in real form (outside of Cursor at least). It’s a typical feeling for interesting new product — “It seems obvious in retrospect, operates simply and (looks like) effectively.” Each line applies to us. It was obvious. It’s so easy to use (debug command). The npm install worked for Mac just 10 minutes before demo :)
5. A good MVP is always really focussed and simple. For a long time I have been trying to make something like this happen for Archie AI work, but it’s still work in progress. There was something unique about this hackathin environment that I would like to replicate that helps do that. I think there is something unique about being short-term focussed (just make something work in 2 days), no worry of funding/marketing etc just build, build something you’ll use (as no time for use research), and working with people with different skill set (npm was Amir’s idea. Shell scripting and original idea was Samarth. LLM logic and parsing files was mine.)

ps: after all the demos were done, just 10 minutes before award presentation ceremony was about to begin, Samarth and I were thinking of heading back to MTV. We had no hope of making it to any awards :) We found some of demos including product and delivery way better. But grateful, judges saw something in our idea. Glad we decided to stay!

Let me know if you liked the product :)