Wolfram Alpha

A computational knowledge engine, launched in 2009 and built on Mathematica, the system Stephen Wolfram had been developing since the eighties. It was never a search engine. Ask it a question and it did not hand back a list of pages that might contain the answer. It computed the answer, from curated data and encoded mathematical rules, and showed the working.

Scientists, engineers and students used it for exactly that. Solve the integral. Convert the units. Give me the population figures, the orbital period, the chemical properties, plotted.

It sat in a gap nobody had a name for. More than a search engine, because it produced answers rather than documents. Less than what we now call AI, because nothing in it learned anything. Every fact was curated by hand and every rule written by a person, and that constraint was the whole point: it could not be wrong in the way a language model can be wrong, because it never guessed. It also could not answer anything outside what had been encoded for it, and could not parse a question phrased in a way its authors had not anticipated.

Which is why it survived the arrival of AI by becoming a component of it. Language models are poor at arithmetic and worse at facts they were never trained on, so several of them now call Wolfram Alpha as a tool for the parts they cannot do reliably. The pattern is older than the current era, and it holds: the model handles the language, a deterministic system supplies the number.