~~I have to admit~~, AI can be really useful

For the last few months, I have been trying to reconcile my single-file hledger journal and each time I always struggled with the unresponsiveness of hledger-ui. Previously, my journal was split by year. Sometime in early 2026, I decided to try Beancount because of [Mas Riyan](https://ruz.fi). He showed that Beancount's Python API can be really powerful. To slowly transition to Beancount, I had to combine all my journal files into one big journal. Combining the yearly journals was helpful not only to prepare them for transition, but also to spot potential account refactors. Because of that, I also reorganized my CoA. At the end, I did not move to Beancount, but I am still stuck with the single journal format.

The issue with a big journal spanning four years is that it's slow. Running `hledger bs` will take approximately 5 seconds, though I don't really think it's a big issue. The real deal comes when you try to run a plain `hledger-ui` for reconciling (at least for my workflow). Navigating between the menu will feel very sluggish, even without the `--watch` option.

AI, through Google AI Overview, helped me discover that I can use `-p` with `hledger-ui`. ~~Using `-p june`, for example, will make navigating inside `hledger-ui` become much smoother because (maybe) the hledger-ui only has to render register content for May 2026. The underlying query would still take time tough, but this is fine for now. I have to admit, while it's like a love and hate relationship, sometimes AI can be really useful.~~ Never mind, the `-p` does not really work. At the end, I think the only way to get hledger-ui become more responsive is by splitting again my journal file :)