@elonmusk An alternative approach:
1. Find as many data points with a statistical correlation to scrapers as possible.
2. Combine them all using Bayes to get a single "odds of being a scraper" output.
3. Progressively lower the rate limit based on the odds.
Most users won't even notice.
The last few hundred years can be viewed as a transition from the rule of persons, to the rule of systems. Democracy and capitalism are canonical examples.
This trend continues to steepen as we move from policies to algorithms as the dominant encoding.
π