Hi all,

A few days ago, the OSS program for Claude went live  
(https://claude.com/contact-sales/claude-for-oss ) and my application was 
approved yesterday, so I started experimenting.

The question I wanted to answer: could I pair with a GenAI to build a Jakarta 
Data 1.0-compatible implementation, integrate it into TomEE, and actually pass 
the Jakarta Data TCK?

I was skeptical about the last part at first. There are reference points to 
draw from: Deltaspike's data repositories, Hibernate's approach, the spec 
Javadoc, and so on but getting all the pieces to fit was another matter.

In practice, Claude had solid initial ideas but needed direction at key points: 

- pointing it to specific parts of the TomEE codebase, 
- telling it to follow the pattern used in a particular CDI extension, 
- nudging it to consider edge cases,
- nudging it to built more application composer tests

It was very much a collaborative process rather than a hands-off one and took a 
few hours :)

Before anything gets integrated, I still need to do my homework against the ASF 
GenAI guidance ( https://www.apache.org/legal/generative-tooling.html ) and 
make sure all conditions are met (and of course some quality post-processing …).

That said, I'm genuinely surprised this worked as well as it did, and I thought 
it was worth sharing with the list.

The code is here: 
https://github.com/rzo1/tomee/tree/jakarta-data/container/openejb-jakarta-data 

Gruß
Richard

Reply via email to