Last update is almost one year old! What happened?
Well, end of last year I was busy looking for a job. I joined ClickHouse and got busy onboarding. Awesome company by the way, and we are recruiting across the board!
Then personal events kept me away from my OSS projects.
Finally back to my usual lifestyle and guess what? Javelit is pretty much useless with the current state of AI!
A Javelit app ticks all the boxes of an app that can be vibecoded: exploratory, POC-like or internal, simple structure, and a straightforward feedback loop ("do I like what I see?").
As a last stand, I did try to make Javelit evolve: I first explored a notebook mode: it worked great, and it was the best notebook experience for Java. Then I looked at the sheer number of Java notebook projects that never took off (including my own!) and decided it'd be a shame to give this treatment to Javelit.
I then tried to make Javelit "Lovable-like". I added a chat component in a left panel, plugged to the code. It was possible to generate and evolve a Javelit app by simply chatting with the app itself. Pretty cool, but the experience was worse than Lovable in many ways and pushing the Javelit architecture to its limits. A big rewrite would have been necessary to get to something good enough.
Javelit is dead. Wishing good luck to Streamlit to survive. I think their last strong point is being python-only and single-file, which could resonate for data analyst and data scientists, but is of lesser value for Java-ists.
This blog is (maybe) not dead though! I have a few thoughts about data representation I'd like to share, similar to what's done in junkcharts.com, one of my favorite blog. I may also write about ClickHouse, who knows.
Take care.
Cyril
Well, end of last year I was busy looking for a job. I joined ClickHouse and got busy onboarding. Awesome company by the way, and we are recruiting across the board!
Then personal events kept me away from my OSS projects.
Finally back to my usual lifestyle and guess what? Javelit is pretty much useless with the current state of AI!
A Javelit app ticks all the boxes of an app that can be vibecoded: exploratory, POC-like or internal, simple structure, and a straightforward feedback loop ("do I like what I see?").
As a last stand, I did try to make Javelit evolve: I first explored a notebook mode: it worked great, and it was the best notebook experience for Java. Then I looked at the sheer number of Java notebook projects that never took off (including my own!) and decided it'd be a shame to give this treatment to Javelit.
I then tried to make Javelit "Lovable-like". I added a chat component in a left panel, plugged to the code. It was possible to generate and evolve a Javelit app by simply chatting with the app itself. Pretty cool, but the experience was worse than Lovable in many ways and pushing the Javelit architecture to its limits. A big rewrite would have been necessary to get to something good enough.
Javelit is dead. Wishing good luck to Streamlit to survive. I think their last strong point is being python-only and single-file, which could resonate for data analyst and data scientists, but is of lesser value for Java-ists.
This blog is (maybe) not dead though! I have a few thoughts about data representation I'd like to share, similar to what's done in junkcharts.com, one of my favorite blog. I may also write about ClickHouse, who knows.
Take care.
Cyril