/q/ - Q&A Central

client's seeing double sessions because they're firing custom events before the default config loads. just move your session_start overrides into the initial config call so it doesn't overwrite the existing params mid-stream.

gtag('config', 'G-XXXXXXXXXX', {'session_start': {'user_id': '12345'}});

it's a simple tiny fix but saves the attribution from looking like complete garbage. unless you enjoy explaining why bounce rates are zero.

tried updating our aws provider to 5.0 but now half the state is garbage. everything's marking as "changed" even though nothing actually moved in the console. it's basically broken for any large-scale infra.

version = "~> 5.0"terraform {required_providers {aws = {source  = "hashicorp/aws"version = "5.0.0"}}}

totally fine a nightmare for anyone with more than ten resources.

the part about not getting sucked into client drama is so true. i've definitely been tempted to agree when they start trashing their last agency for being "inept." it's usually just a mess of bad communication on both sides, not some grand conspiracy. also, the advice on not handing over credentials or migrating hosts until you're actually paid is the only way to survive.
>Get paid before doing the work
i've seen people try to negotiate a "promise to pay later" during a handoff and it never ends well. it's such an awkward dance when you have to be the bad guy about the final invoice while trying to stay professional with the new freelancer. i hate that feeling of being the gatekeeper for the assets. it feels petty even though it's just business. do you guys actually share all your process docs with the incoming dev or just the bare minimum?

full read: https://speckyboy.com/work-clients-next-previous-developer/

another day of chasing googlebot around. read this bit about using ai to fix indexing issues on some tolkien art site. apparently just pumping stuff into social media helps crawl the pages.
>When you launched your website. there's something i recently realized that's made the frustration of this serious issue a little bit easier to live with.
sounds like a massive waste of time if you actually have a real crawling problem. it's basically just hoping some bot hits a link and tells google we exist. using ai for this feels like adding more bloat to an already broken system. i don't see how social signals fix a legitimate sitemap or robots.txt error. if the crawler isn't hitting the url, a few tweets aren't going to save your rankings. seems like a lot of extra work for zero actual control over the index. just another way to pretend we can bypass google's shitty behavior with magic tools.

https://whitespark.ca/blog/how-to-overcome-google-indexing-problems-with-social-media-ai/

nowthis thinks they can just run eight episodes before deciding if a series even works. that diner in nyc pulling 20m views sounds like a total fluke anyway. i don't see how anyone has the budget to trial that many formats without burning through everything. does anyone actually have a smaller testing framework that isn't just guessing?

https://www.linkinbio.news/p/how-to-pilot-a-social-series

another way to bloat the runtime with more env vars. using 'atob' for decoding credentials in a middleware function sounds like a security headache waiting to happen. i don't see how this is better than just sticking it behind nginx or a proper vpn.

function isAuthenticated(request: Request): boolean {  const user = import.meta.env.BASIC_AUTH_USER;  const pass = import.meta.env.BASIC_AUTH_PASS;

it's just more logic to debug when the deployment platform fails to sync the secrets properly.

more here: https://www.hongkiat.com/blog/add-http-password-astro-site/

>Prompt tracking can tell a team whether it appeared in a sample of AI answers for a sample of prompts.

im so tired of seeing these fake visibility scores on linkedin. its all just noise and none of those tools actually show if a site change did anything for the bottom line. everything is too unpredictable because of how users follow up on chats now.

more here: https://www.searchpilot.com/resources/blog/is-geo-working-how-to-get-beyond-prompt-tracking

i just found out you can actually see who is lurking in your docs via the activity dashboard under tools. its pretty handy when you send something to a stakeholder and they go radio silent. though it only works for work or school accounts so dont bother looking if youre on a personal gmail. also people can totally disable their view history so you might still be flying blind if they are being sneaky. i wonder if this counts as way too much micromanaging lol. honestly kinda creepy if someone is just tracking your clicks like that.

found this here: https://zapier.com/blog/how-to-see-who-viewed-your-google-doc

im still catching myself pasting entire pytest files into chatgpt like some kind of idiot. saw this thing where someone built a python 3.11 scrubber just to redact names like alex chen before the model sees them. it basically caught the model hallucinating because it was just memorizing the assert values instead of actually parsing the logic. wonder if i could use something that simple for my own env without it breaking everything when there are actual edge cases in the strings.

https://dev.to/magickong/learn-fixture-leakage-by-building-a-tiny-prompt-scrubber-3di6

everyone keeps using these terms like they're interchangeable but it's annoying.
>A scheduling-only tool will save you a few hours a week. A full automation system changes how your team operates.
the idea of integrating social listening and approval workflows into one platform sounds great until you see the enterprise price tag for 'governance tools'. i'm skeptical about letting ai handle caption drafting without constant manual oversight, it's too easy to lose the brand voice in a sea of automated tasks. probably just more bloat

found this here: https://blog.hootsuite.com/social-media-automation/