I think you can do this with a Fisher-Yates shuffle (https://en.wikipedia.org/wiki/Fisher%E2%80%93Yates_shuffle), right?

I wonder how much CPU this uses. An immediate mode GUI (ImGUI, which CodePerfect 95 uses) constructs the entire GUI on every frame by default.

It does run through every node in the would-be scene graph by default, but IIRC the Dear ImGUI library caches the actual expensive toolkit, graphics, and/or window system calls for each subtree. The API is immediate-mode, but its implementation does not have to be stateless. (Cf. React.)

It's great until you want to cancel; they require you to call their customer support number to do so.

I got so pissed off by this that I rang them up right after subscribing and cancelled. At least they lost one customer for it.

I wish the article detailed the performance issues with the old implementation, and why those issues necessitated a rewrite (other than "strong primitives" and "difficult to retrofit").

The same blog has other benchmarks[1], one of which is SQLite[2].

[1] https://tech.marksblogg.com/benchmarks.html

[2] https://tech.marksblogg.com/billion-nyc-taxi-rides-sqlite-pa...

Maybe it’s more of a security feature? What if phishing emails had a bogus email for the Reply-To name, E.g., “service@paypal.com <hacker@evil.com>”

See https://news.ycombinator.com/item?id=18452992

That was my first thought, but if that's the case, then why is it implemented in the mobile apps?

Different people implementing them?

Thanks for providing this service and having such great communication with users!