The chatbot use-case gets gets hyped up all the time, but they never seem to actually be particularly effective. I want to interact with one that lives up to that hype.

Haaa, yes—imagick is def. from a different era of web development. Still super powerful, but "clunky" is a good word for it.

The Browershot approach that @pascalbaljet.bsky.social wrote up works super well.

Haven't tried it yet, but @benholmen.com recently used imagecopyresampled() for howtall.is:

benholmen.com/blog/howtall...

Depending on how dynamic you need them to be, that could be a cool dependency-free approach.

Use chunking to try and avoid memory issues? Generate the sums in the background and cache them? Store an aggregate on the Account model?

For example, source-diving in Laravel Pulse shows a `pulse_aggregates` table that stores the sum, the type of info being summed, and the relevant period of time.

Using the audio of the poem was such a cool choice—so much more disquieting than quiet-loud-quiet BWAHHH sounds that editors normally use

Totally. Especially if you can actually use the tools first-hand like you did, the advantages become clear so quickly.

(If a stakeholder *doesn't* have that context, though, that's where it seems to get a bit squishier.)

That's really, really kind—thank you! We'll see how that particular discussion progresses. Having your conversation with Matt to use as a reference is already a tremendous asset.

I am, once again, in this exact conversation right now. Despite demonstrating all the tools and efficiency available in the Laravel ecosystem, I might lose the debate because of that intangible new-hotness factor that JS can have.

As my old business partner said, "It's like talking about religion."

Also, just switch the bool from false to true to restore your desktop:

defaults write CreateDesktop -bool true && killall Finder

(As someone who spends too many hours each day bouncing between different things, Raycast is the way.)

This was a great write-up! Using imagecopyresampled is a cool solution for your opengraph use-case—I'm actually not sure I knew that library existed before now.