It’s been quite a long while since I updated my personal website. Last I did it was 2021, about 3.5 years ago.

TL;DR: A lot has happened. I joined Google, got a bit bored of the work, ChatGPT happened, I made a viral app called GitaGPT, left Google and started up for a while, consulted for startups to make some cash, joined Entrepreneur First as a Founder-in-Residence, went through the startup founding journey.

🚀 trying to startup

In July 2024, I joined this really cool startup accelerator called Entrepreneur First (EF henceforth) in Bangalore. It was quite a thrilling experience, I made good friends, tried out many co-founder pairings, had a good idea that I tried to get customers for as well.

Entrepreneur First has quite the interesting model. They collect highly talented individuals in a cohort, and then encourage them to get paired into optimal teams. The expected result of the three-four month sprint is to end up with a venture funding worthy business idea, that has early traction with the two co-founders being the right people to build the startup. My cohort saw quite the interesting bunch of startups, from space tech (shoutout to Aule Space), to developer tools (shoutout vluna) and AI agent startups from a general AI agent company, to one verticalised on Salesforce administration.

Meeting the cohort was an absolute blessing, and I am quite grateful that I met these wonderful ambitious smart folks, who I can build so much with. I grew a lot as a person during the last 6 months.

🦉 OmniGrep: new age code search and understanding

I got an idea about re-inventing code search and understanding in the developer tools domain, then did a lot of user research to validate it, got it validated as well.

My idea was quite simple: understanding and navigating codebases from simple stuff like just classes, to larger questions like “Why does this microservice use Firebase for authentication and not the in house authentication module, and when was this decision made?” , all answered by a simple chatbot that is trained on your company’s codebase, technical documentation, issue tracker and the QnA channels within the company.

I called it OmniGrep. Here’s the thesis I wrote about it, feel free to email me on feedback!

However, on attempting to get design partners, things did not go well. It was hard to establish trust as a 25 year old, to give me the whole codebase and all engineering lifecycle data, and honestly the problem wasn’t that burning enough that people would want a separate tool. Then GitHub too released documentation and issues as data sources for Copilot Chat, and that basically stopped my startup in the tracks.

I loved it though, and it was quite fun and exciting being a founder. I learnt a lot about how to conduct user interviews, how to interface with people, the process of customer acquisition, and most importantly how damn hard being a founder is, mentally and physically. I also learnt that I have a major flaw: I do not understand the basics of business properly. To fix this, I have started again from basic books: Zero to One (Peter Thiel), and Influence (Robert Cialdini). Would love some feedback on what else I could read to get to 80%ile of business understanding.

❓ what’s next?

What’s next then? Somewhere, I felt that I have been too deliberate about trying to validate ideas and get things going that way. It did not feel natural to me. In fact, the last time a project of mine succeeded (GitaGPT), was just a fun thing I built and released within a day.

After all, I do get a huge kick out of just building things, and since Cursor has accelerated my development cycle so much, it might just be faster to make things and show the world, get a tight iteration loop right there. So now the plan is simple. I will make things, and release them.

🛠 getting back to basics: the builder way

No playbooks, no “will it get funded?” thinking, no extensive user research, just pure building the things I want to see in the world. High failure is expected, downside is non-existent.

So as a new year resolution, here’s a thing I’m going to do. I’m calling it, weekly build.

📅 weekly build: making something every week

Every Saturday, I will release something I’ve been building over the past week. It might be something new, or a major upgrade of something already being built. The idea is to monetise right away as well, to test out whether it is actually a viable business idea.

the first build is quite simple, and has been personally quite useful to me

💻 terminal agent

Every time I open my GNOME Terminal, I am disappointed. I want to write a natural language instruction. I want the thing to write bash itself and execute the steps. Maybe even pull from the web. Sometimes, I just want to talk ChatGPT, and not have to copy commands in. I just want to type out “Move the docker image location from /root to /home”, and instead of yapping and back-and-forth conversation, give me the command and ask me if I want to execute it.

So I made an agent that lives inside the terminal, invoked by tgpt.

Here is the full post about it, including idea, implementation and other details.