To expand on the maintenance status in the readme, I spent roughly 10 hours over the last week reading and evaluating the kibit code base, looking for ways to improve the overall speed and robustness. However, the fundamental requirement of core.logic is too great. It's an incredible library that does exactly what it needs to, but it's entirely too slow to be used at any scale. Even on simple cases such as '(do (when test then)), it takes 10-20x longer to evaluate with the current set of rules than Splint does with nearly twice the number of rules.

Any changes to the code-walking logic (using hand-written code walking instead of tree-seq, only constructing a simplify-map after checking the guard, removing laziness, etc) barely change the performance because of how much time is spent in the core.logic code running the rules.

I experimented with embedding #C04SCGV2ATX directly in kibit, treating kibit as a wrapper around splint, but that required enough effort that I think it's better for interested users to just switch to splint.