I couldn't find a job, so I built an open source project
Published: 2025-06-23
Instead of chasing a job, build an open source project, and the job will come. And if not, at least you’ll have an open source project.
My interpretation of the famous German saying
job search
In April 2025 I started looking for a job. For the past 5 years, I’d been a founder and CTO. I didn’t know what role would best fit me and what to expect. So I was taking all the calls and interviews.
Luckily, I quickly landed a couple of interviews. Unluckily, they were all Leetcode-based and I didn’t do well on them. Embarrassed, I decided to brush up on my Leetcode and spent a week practicing it. But as much as I enjoyed solving puzzle-like questions, it felt like a huge waste of time. So I decided to simply reject interviews with any company that uses Leetcode for screening.
I then had a couple other interviews that were project-based, which I did better in. Unfortunately, none of those opportunities materialized; some because of me, others because of a mismatch in culture, geography, or other reasons.
To stop wasting my and other people’s time, I decided to figure out what I’m good at, what I want, and what company would benefit from having me.
As a former technical founder I can build and deploy a product, front and back. But because I covered the entire stack, I was an expert in no particular area. I am what people call a generalist and I wasn’t sure if it’s a good thing.
I also noticed that most companies are still looking for programmers - someone to write code. It seemed like many missed the seismic shift brought by AI assistant coding.
Looking at the market through the technical founder lens, I feel like the profile of a 100x engineer of today is completely different from before. My thinking was: “Why pay someone $200k/year to write code when you can get it for 20$ with Cursor”?
And being on the other side, as a prospective employee, I think getting a job where you’re paid to write code is akin to a career suicide. It’s like becoming a telegraph switchboard operator right when mechanical call switching technology was introduced.
If I were hiring, I’d be screening for agency, adaptability, and creativity. I was fairly confident in my assessment, but what I saw in the labor market contradicted it.
After another month of fruitless meetings, annoyed, I decided to put the interviews on the backburner and work on something I was itching to build for a while: an open source database editor with AI assistant.
open source project
All my previous startups dealt with data and I used DataGrip extensively. I love its features like editable tables and copy-pasting data ranges. I’m also a big fan of Cursor for obvious reasons. So I wanted to build an open source version of DataGrip with a Cursor like assistant.
Around mid-May I started working on Dribble. The idea was simple: rebuild DataGrip with React and add an AI assistant to it. But there were a couple features in DataGrip I wanted to improve, the biggest one being the management for saved queries.
In Dribble, I made queries into first-class citizens. They’re treated as files. Each query has its own data table, versions, and logs associated with it. They’re also designed to be sharable, searchable, and permissioned.
I also abstracted db connection and query execution logic into a separate docker based worker. Each db source has its own worker, which enables sharing connections, access management, and scalability.
The AI assistant I had to build from scratch. Some interesting solutions I came up with were adding db schemas and context queries into the system prompts. I also added a query execution tool so the assistant can run and verify SQL before returning it to the user.
There were a few other engineering challenges I had to address, such as:
- SQL linting in editor
- An intuitive folder tree
- Data pagination (copied clever DataGrip method)
- UX for query versions and run logs
There are still quite a few improvements that are needed. For example, the app works with GPT, but I want to add support for open source LLM models. I also want to add SSE to support streaming. And, ironically, editable tables are still missing.
Nonetheless, I’m very excited about the way it’s shaping up and quite proud of it. So today I’m opening it to the world.
plans for the project
The more I was working on it, the more excited I would get about it. I saw the potential of it being useful not just for developers, but also for startups and SMBs. Specifically - a company that has multiple data sources (database and data warehouse) and needs to safely expose data to their team.
To support such use cases, I decided to release Dribble with an open-core model. There will be 2 versions: free open source for developers and enterprise with on-prem only.
Next I’ll be working on the following enterprise features:
- On-prem deployment
- On-prem AI assistant
- Support for data warehouses and data lakes (Snowflake, Redshift, S3, etc.)
- SSO and user management
- Access management
- Logs
The goal would be to make this solution attractive to companies that have strict regulatory requirements (health care, fintech).
I also want to keep pricing simple and start with a 500$/month for an on-prem enterprise plan. Unlimited usage and users. Might be too cheap or too expensive, I don’t know yet.
If you are a startup and can benefit from such a product, please reach out. I’m looking for design partners. And if you’re a developer, please give it a try and share your thoughts. I’d appreciate any feedback.
final thoughts
When I started this project, my objectives were simple:
- Build something cool
- Regain confidence in my skills
- Learn more about AI
Prior to this project, I had a limited exposure to AI assisted coding. So at the beginning I still had this mistrust. I was worried I’d lose control over the codebase, so I’d use the assistant cautiously. I started by delegating smaller tasks to it, like styling, code completion, and simple functions. But it kept writing good code, often better than mine. Gradually, I delegated more and more complex tasks to it.
By the end of the third week the dynamic had changed drastically. I was no longer the coder, but a project manager, architect, tester, and coordinator.
This experience has confirmed my suspicion - software development is no longer about writing code. It’s about delivering useful solutions. Writing code is now an AI domain and it’s ok.
It doesn’t matter if a cat is black or white, as long as it catches the mice
Deng Xiaoping
More importantly, this experience has changed my outlook on the future.
Before this project, I thought my current skills will be irrelevant in 5 years. I thought it would be harder to learn new technologies as I got older and AI will make whatever I already know irrelevant. So I might have to switch to a managerial role.
But after working on this project, I feel more empowered than ever. The barriers that would normally hold me back, such as lack of skills, lack of capital, or lack of time - are now gone. I’m only limited by imagination and ambition (I know, a bit cheesy, but that’s genuinely how I feel).
Working on this project also reminded me of the good startup days - that excitement of building something new.
Overall, I’m very happy I started this project and looking forward to what comes next!