Glassdoor's gonna hit you with a ~300-result hard cap on every search, so you gotta get creative and slice your queries up by geography, salary ranges, and date filters if you want any real data out of it. Also, fair warnin, CSS selectors break constantly with React, so unless you wanna wake up to silent failures tanking your entire pipeline, you need to have backup selector strategies ready to go. For the dedup stuff, don't be lazy and just lean on URLs - compute a proper deterministic fingerprint using SHA-256 of your normalized company + title + location combo instead. And yeah, the AlterLab API works fine for smaller jobs, but it bleeds money real quick, once you're hitting 100k+ pages, direct curl_cffi pooling with residential proxies will save you a ton of cash compared to the API. Really solid jumping-off point though - would be fire to see a follow-up deep dive on tackling Cloudflare's JA4 fingerprinting and how to handle production-scale archiving without losing your mind