Quackprep.orgt

The unraveling came via a data science blog post. A researcher downloaded all of QuackPrep’s practice questions and ran a statistical analysis. The findings were damning: over 40% of the questions were verbatim copies from publicly available old tests (some from defunct exams like the SAT II). Another 30% were AI-generated, but poorly vetted—one physics question asked for the “speed of light in a vacuum” and gave answers in “apples per second.” Most critically, the adaptive algorithm was not adaptive at all; it simply advanced users regardless of performance, creating a placebo effect of improvement.

The name “QuackPrep” reflects the organization’s core belief that learning should —i.e., break the silence of boredom, echo enthusiasm, and splash into a sea of curiosity. quackprep.orgt

The cracks appeared subtly. First, users noticed that the “explanations” for wrong answers were often circular—e.g., “B is incorrect because A is correct.” More concerning, the site’s predicted scores were suspiciously generous. A student who scored in the 40th percentile on a real College Board PSAT would suddenly see an 80th percentile prediction on QuackPrep. When questioned, the site’s anonymous forum moderators offered platitudes about “growth mindset” and “different scaling models.” The unraveling came via a data science blog post