/q/ - Q&A Central

the text looks blurry as hell compared to chrome. i tried setting -webkit-font-smoothing: antialiased but it's making the weights look way too thin on the light stuff.
>it looks fine on desktop
it's definitely not fine when you're actually looking at it on an iphone. anyone found a fix for this?

I was reading about how being cited twice in ChatGPT doesn't actually matter if you aren't in the final recommendation list. They used Melp App as an example where they were cited but didn't even show up in the output text. It's pretty useless for conversions if people aren't clicking through to check sources anyway. I don't see how we're supposed to optimize for that without just basically spamming every possible niche.

article: https://www.growandconvert.com/ai/seo-for-llms/

The label in name rule isn't experimental anymore so it's going to flag everything where the aria-label doesn't match the visible text. I've seen plenty of buttons like <button aria-label="Start page"
>Home</button
> slipping through during scans before. It's basically going to start throwing errors every time we have mismatched names, which is going to be a massive pain for our backlog.

more here: https://www.deque.com/blog/find-label-in-name-issues-in-every-scan-with-axe-core-4-14-0/

Webflow's API behavior with CMS fields is so annoying. I always assume there's a bug in my fetch logic when certain keys just vanish from the payload. Turns out it's mostly just because the field is empty or someone used a display name instead of the slug. It's not even dropping the data, it's just skipping the key entirely if there's no value.
>The API usually isn't dropping it. Items with no stored value come back without that key

That explains so many of my undefined errors in Next.js. The idea of building a Route Handler on Webflow Cloud to diff the schema against live items seems like a lot of overhead though. It would definitely help find exactly which slugs are missing, but running a normalizer to force every field into the response sounds heavy for production. I'm not sure how much latency that adds when you're trying to hit a live endpoint. Probably worth it if you have a massive collection and can't manually check every item. It's basically just a way to map everything back to a consistent shape so the front end doesn't break.

more here: https://www.webflow.com/blog/fix-webflow-api-missing-fields

just saw some stuff about ghost citations and it's actually insane. ai models are literally using our data to answer questions without even giving us a shoutout lmao.
>power ai answers without brand mentions
it's basically just scrapin' everything and leaving us with nothing. seeing people talk about prioritizing fixes to get visibility back is cute but i don't see how it works if the model just doesn't care. feels like we're just training our replacement for free. i doubt any 'fix' is gonna stop the crawlers from just taking the info and running.

full read: https://searchengineland.com/ai-search-ghosting-brand-fight-back-499342

another day of fixing broken pipelines because someone bumped a minor version in a module. it's always the same provider-level dependency hell that breaks the plan. let's see if anyone can actually pin their providers to a specific hash for one week without getting an error from some random github action.

terraform {required_providers {aws = {source  = "hashicorp/aws"version = "~> 5.0"}}}

the token usage for max reasoning is insane.
>it took the "Max" reasoning almost 10x the number of tokens to get there.
setting verbosity high just makes everything muddy and i don't see the point when you're doing simple tasks. anyone actually use the mcp servers feature for anything useful?

link: https://zapier.com/blog/openai-playground

i was messing with our legacy modal component earlier and that annoying mustard-colored console warning popped up again. i used to think it was just a glitch in how bootstrap handles things but turns out the standard fixes like using a setTimeout or blurring the element are basically useless. it's actually breaking stuff for screen readers because focus is dropping into a void before the region even hides. honestly, if you aren't using the native <dialog
> element with. showModal(), you're just asking for trouble. i tried to just reorder my close function but realized i was still missing the part where you have to make the overlay inert. it's such a tiny detail but if you do it in the wrong order, you get that ghost focus issue. ngl, i've probably shipped at least three different versions of this broken logic in the last year. it's frustrating because the easy one-liners we all copy from stack overflow are exactly what cause the accessibility break. my current component library is too bloated to swap everything out for native dialogs right now so i'm stuck trying to fix the sequence manually.

function closeModal() {  overlay.setAttribute('aria-hidden', 'true');  overlay.classList.add('fade-out');  // ... still failing here}

https://css-tricks.com/blocked-aria-hidden-fix/

just use css transitions instead of trying to math out every single delay in js. i saw someone doing this today:
>it takes about 400ms for the slide to finish
400ms 600ms? it's literally impossible to tell without a stopwatch lmao. it makes the whole site feel sluggish and janky if you don't get the easing exactly right.

i'm tired of seeing people just dump forty terms into a doc and calling it a day. it doesn't matter how tidy the list looks if you're just writing descriptions from memory three weeks later. there's gotta be a better way to actually use the real search data without it becoming a useless graveyard of dead tabs. do you guys actually use these lists for anything besides feeling productive?

found this here: https://www.tailwindapp.com/blog/pinterest-keyword-research