Then came Diehr, which didn't say anything had changed, but it was opportunistically read as if something had changed. Software is patentable today because our legal system is inconsistent. Law is not law.
In Diehr, the inventor claimed "algorithm + machine" where machine was a computer. The Court looks at a computer as an infinitely configurable machine, with each new software algorithm creating a new invention.
That is the distinction that the courts make between Benson/Flook and Diehr, which has allowed "software" patents to exist.
The rest of the Diehr patent included a combination of elements claimed to be novel, not a general-purpose computer.
Also, from my understanding an infinitely configurable computer and general-purpose computer aren't the same thing. Infinitely configurable means a new machine every time new software is installed, which would allow the invention to fall within the patent statutes. A general-purpose computer is not a new machine, has already been invented, and therefore trying to patent an "algorithm + GPC" would mean a patent on the algorithm, which is outside patent protection.
According to the respondents, the continuous measuring of the temperature inside the mold cavity, the feeding of this information to a digital computer which constantly recalculates the cure time, and the signaling by the computer to open the press, are all new in the art.
From the first line of the opinion, "We granted certiorari to determine whether a process for curing synthetic rubber which includes in several of its steps the use of a mathematical formula and a programmed digital computer is patentable subject matter under 35 USC 101." Diamond v. Diehr, 450 U.S. 175 (1981).
Rehnquist, who wrote the majority opinion, said it right there we are talking about "formula + computer."
Claim 1 of the patent itself reads, "1. A method of operating a rubber-molding press for precision molded compounds with the aid of a digital computer, comprising..." Diehr at Footnote 5. The formula is the Arrenius equation (everyone already did this), and then Diehr added a computer for continuous monitoring, calculating, and output (the novelty).
Diehr themselves claimed in their arguments that their novelty to is "the continuous measuring of the temperature inside the mold cavity, the feeding of this information to a digital computer which constantly recalculates the cure time, and the signaling by the computer to open the press..." Diehr at 178-179.
EDIT: You edited your previous comment to include the same line as me. It seems as though we actually agree on what was novel here.
Most software patents are essentially an algorithm attached to a general-purpose computer. This relationship cannot be claimed to be novel. Diehr could at least claim that their device as a whole was novel and nonobvious (though the Supreme Court case didn't examine those questions).