Before search engines got good, one perfectly reasonable strategy for learning science was to ask an actual scientist and hope somebody answered.
MadSci built an entire website around that idea.
MadSci says it debuted in September 1995 as part of Washington University’s Young Scientist Program, a student-run effort aimed at improving science literacy among K-12 students.
It evolved into a volunteer Ask-a-Scientist network covering subjects ranging from astronomy and biology to chemistry, computer science, engineering, earth science, and physics.
The site’s own history says that in 1999 more than 900 scientists volunteered and answered more than 25,000 questions. The current homepage advertises an archive of more than 36,000 answered questions and still offers the old core functions: browse the archive, search it, read FAQs, use the Random Knowledge Generator, or submit a question.
The examples are what make the project memorable.
Why do grapes spark in a microwave? Why does root beer foam with ice cream? Why does a black eye swell? Why might a frog urinate when somebody picks it up?
This is exactly what science outreach looks like when ordinary curiosity is allowed through the front door.
A human search engine for questions Google had not indexed yet
MadSci belongs to a specific era of the web when expert communities were experimenting with direct public access.
Instead of publishing only polished articles, the site preserved the questions themselves. That meant the archive gradually accumulated all the awkward wording, odd premises, school-project emergencies, and genuine curiosity that people actually bring to science.
That archive became more valuable with age.
A single answer might help one student. Tens of thousands of indexed answers become a searchable historical record of what people wanted to know, how experts explained it, and how online science communication worked before modern Q&A platforms and AI assistants.
The site also developed FAQs, web-resource libraries, experiments, research datasets, and educational tools around the question-answer core.
Most importantly, it kept the old structure visible.
The homepage still looks like a laboratory that escaped from 1997 and refused the redesign meeting.
CacheRat’s 1,967 Ancient Web Domains research list exists for pages like this: sites whose accumulated answers matter more than whatever visual fashion the current year says a homepage should follow.
MadSci’s copyright line now spans 1995–2026.
Thirty-one years later, the laboratory that never sleeps is apparently still combustible.