See also the list of Things I did with Claude on a Week-End.
I feel there's a better way to break lines in wz's tooltips (in particular) and in citations (in general); see for instance the horrendous indentation of Love's opening citation, and how the most perfect prose can be butchered by wrong line breakings:
I'm resisting using the phone which I see as a black box that is difficult of access except as a passive user. With AI, however, it's probably possible to make the most out of it, starting with cleaning all those viruses and pop-up that constantly shower me with stupid notifications. I already implemented a basic π feature to upload directly phone-camera photos to the server; this could be made better (faster, offline, etc.)
This looked like a nightmare since I promised I'd do it one day: importing my old Wordpress blog to laussywiki. For AI it should be a piece of cake. Install the corresponding tarball of the time, make the conversion, and done. I should then recover my beautiful Gaudi and Sagrada Familia posts.
The site can be quite sluggish. It has a lot of functionalities. AI could help optimizint it, which is both a complex and dangerous process!
One such bot should take care of things that get broken upon upgrades, such as broken youtube videos, or broken equations, or even broken references, e.g., as seen in my Blog post on CSI with frequency-filtering (a lot got broken here).
Another bot should clean behind me, such as removing the emtpy <wz tip=""></wz> put by automated figure uploads. Not a big deal, but if this can be cleaned, why not? I like the idea to have a swarm of little bots cleaning the mess after me. Such a botβlet's call it M-Oβcould be upgraded as time passes and realizations strike with a list of things to keep tidy. Another example is the double periods after citations or misplaced punctuation in references, e.g.[1], (the comma should be put before the reference). Here's an example of doubled periods, with the expanded Spontaneous coherence buildup in a polariton laser. F. P. Laussy, G. Malpuech and A. Kavokin in Phys. Stat. Sol. C 1:1339 (2004).
. I would have to keep M-O off this page too so that it doesn't clean it!
Here's a list of things for it to do:
There was some nice KIM service-menu way to get essential shortcuts at the tip of your ring finger but this became nightmarish to install and maintain, so I essentially gave it up. Claude could probably bring this back in an efficient/portable way, to do things like:
etc.
Add a functionality so that only bit of a yeet is disclosed, possibly with an embargo. Something like this:
I have shown that ββββββββββββββββββββββββββββββββββ doesn't apply in the problem of ββββββββββββββββββββββββββββββββββββββββ
The remaining 2kiB characters of this yeet have been embargoed until further notice.
I wouldn't like to use it, though.
Extend {{pan}} so that it loads a suite of images as one zooms: a first picture that can be zoomed down to some level, then higher-resolution tiles that take over as one zooms further, each level twice as fine as the previous one and only the visible tiles fetched. Like this, people who never zoom don't download a 260 MiB file, while those who do can go down to the most minute details. Since my figures are drawn from data, the tiles could even be rendered directly at the resolution they need rather than cut from one giant picture. To be settled: where the tiles live (hundreds of File: pages, or a folder on the server). Not needed now, but it will be when we enlarge the scope of panoramas such as Fig. (0OG5J).
Coupled π to π through π, i.e., add a β‘ button to π/β€ posts that limit them to Twitter's size limit and post a copy online.
The listing of all the papers of bib in Scientific articles works again and renders three times faster (see what was done), but it remains a single page of 7 MB that only the local wiki can render within a web request. Possible further improvements:
Tags such as <tt></tt> or <nowiki></nowiki> should include what they're selecting, as opposed to overwriting them.