Could you share more info about the motivation and views behind Kaapi?

[nicolas-graves]

Hi!

Impressive achievement.

I'm a bit shaken by the project (see my post at https://lists.gnu.org/archive/html/guile-devel/2026-07/msg00000.html). I was wondering if you were willing to share a little more your motivations (I'm not here to shame, I'm simply trying to update my own "theory of action" given this, I'm a guile user but not really a developer of the language).

• How much did it consumed tokens / cost you (total, over those three weeks)?

• I see contributing sections and a package manager, so I understand it is not simply a toy project. I don't see a motivation section besides "I wanted a complex problem to work on agentic coding. I am not a scheme programmer". Do you expect a community to arise from just the technical value of the project? Or is it MIT-style "I might not maintain this if I loose interest"?

• If no, did you consider contributing to an existing Scheme implementation? Why?

• Do you have views on how the Scheme ecosystem would/should evolve to take into account how easy it is with agentic AIs to create a compelling Scheme implementation from scratch?

[nicolas-graves]

Another question I have: generative AI is indeed good enough to produce an implementation of something that is well defined and tested, but as you consider yourself not a Scheme programmer, how do you expect Kaapi to be able to innovate / bring value more than reproducing ?

[baijum]

(This response is not written by AI)

This is my side project and not endorsed by my employer (Red Hat).

At this point, this is just an experiment to explore what a coding agent can do. As I mentioned, I needed a complex problem with well defined specs.

If the implementation becomes practically usable, I still need to learn two prongramming languages, Scheme and Zig (the implementation language) to verify the code. So, definitely this will take more time.

So my recommendation is use it only for educational/research and not for any real production work.

1. I haven't calculated the token usage
2. I have limited time to spend on this project, that's why I disabled issues & PRs. Only discussions is enabled for now.
3. I cannot contribute to existing Scheme implementation as I don't have any experience with Scheme.
4. It is easy to develop a scheme implementation, but ensuring it's quality and maintaining it for a long period of time is time consuming and challenging.

Regarding the last question, I haven't put much thought into that.