i find it interesting that you can detect slop-generated static sites with a high (although not 100%) certainty with statistics alone, without even looking at the content

the tell is the small size of current version (the deployed one) coupled with a very large size of non-current versions (the ones deployed in the past, excluding any files reused in the currently deployed version)

slop machines love to rewrite your entire site 500 times in a row. there are a few non-slop use cases which result in this but they are few and far between in practice

(note that the quota accounting is weird; the quota is per-site, but the reporting is per-domain, and non-current versions are (obviously?) not counted against your quota. so your domain can incur 180% of a quota despite that quota being a hard cap on the size of a single site)

so i'm having a great time with a new TI chip (different from the other TI chip that i've complained about this morning, basically completely unrelated except the PCBAs they're installed in have the same manufacturer (us))

writes seem to work, reads lose about half to three quarters of the bits of every byte. no i can't explain it better

cursed concept that totally makes sense for something i'm building

SMBus shaders

i have four identical ports that need some mostly but not entirely uniform actions applied to them. i want the behavior to be driven almost entirely by tables (to conserve code size). and the inputs are network packets that have a uniform port mask and an array of identically shaped configurations. which i want to process per-port