Cory Dransfeldt opens with the Merriam-Webster definition of a parasite and maps it onto a five-step business model. The steps are scraping content without paying and calling it fair use, training on it, selling metered model access, repeating until the source is exhausted, and then hoping to contain the effects of models training on model output. He says the old trade of search traffic for indexing is gone, scrapers strain infrastructure, and clickthroughs fall. He gives Reddit's data licensing as an exception, but argues it depends on a moderation community the company seems indifferent to. The post also covers data centers, citing gas generators and higher electricity and water rates. The Hacker News thread pushes back on the framing. One commenter says heavy use of the models undercuts the premise. Another replies that usage feeds a self-improving loop that strengthens the parasite. For founders, the essay reflects a growing complaint that training data is taken without payment. It also points to local opposition to data center siting. Builders who depend on third-party crawled data or large compute footprints should expect closer scrutiny of both sourcing and infrastructure.