Schneider explains why his setup reads ad data from a warehouse and only uses the ad platforms' APIs to write changes. His point is that agent bans come from hammering the APIs, not from using agents at all.
…CMS so that it can actually go and publish that content that it produces or get the final products, you know, into the wild, actually, you know, into the public. So that's kind of the high level. Any questions about that? I can try to answer before we jump into this and actually do these builds.
…The Petri dish that Cody is talking about, the machine around all these models, agents, the machine around context, the machine around, you know, evals, that whole thinking is the job of the person who's building a business in 2027, if you actually care about it getting customers on day one.
…the data warehouse. Now, why do I do that? What is the point? Like, what is the reasoning for this? The reasoning is because you're trying not to abuse the Facebook ads API or the Google Ads API. Which they have terms of service, and agents will go and just like do a thousand requests at once. And this, that is against their terms of service. So, you probably have seen these online where, like, people are like, okay, Grokbot got my Facebook ads account banned, right? Like, don't use agents for this. That's not what's happening. Your Grokbot violated the terms of service of the API because it was sending thousands of requests per second
because it was trying
to do some data poll. So,
you should assume your agents are like drunk patrons at a restaurant.
