How to Save Claygent Credits by Pre-Filtering Data with Native Formulas and Regex

Hey everyone, quick rant/find from a setup I was building today. Most people just dump a huge raw list and let Claygent do all the heavy lifting to find ICP fitment. But Claygent gets expensive and slow if you're running it blindly across 4k+ rows of unverified data. I started aggressive-filtering at the source this morning. Instead of sending every domain to Claygent, I use a fast native formula column to check meta-tags and Regex for immediate red flags first. If the basic keyword check fails, the row gets blocked and never hits the Claygent credits. Saved me around 4,000 credits on a dirty list today. Anyone else hard-filtering before letting Claygent loose, or are you guys just letting the agent filter the noise? Let me know what's working best for scale.