About the Atanasoff-Berry computer...
I knew of it, which is the reason my original post on the topic was carefully worded. "What was the first digital computer" is a contentious question because some of the contenders didn't have all the traits we now consider diagnostic. The question we can answer is "What is the earliest contender that is (a) a true digital computer, and (b) from which digital computers today derive?"
Three traits at issue are (1) Turing-completeness, (2) whether the machine was "stored-program" - that is, stored its code in digital memory rather than needing to be rewired for new computation, (3) whether it has architectural descendants today.
The ABC computer was not Turing-complete and not stored-program. It influenced the design of ENIAC.
The Z3/Z4 was not stored-program and had no descendants. (Its language, Plankalkul, was
magnificently weird.) It was Turing-complete.
The first version of ENIAC was Turing-complete but not stored-program. (Used decimal arithmetic internally, not binary) It had many later descendants.
BABY (aka the Manchester Mark I) was Turing-complete and stored-program. It influenced later designs.
The Bombe (the first Bletchley Park cryptanalysis engine) influenced BABY. It was neither stored-program nor Turing-complete. Colossus (1943) was Turing-complete but not stored-program; it was a deep-dark secret until the 1970s and did not influence BABY, ENIAC, or later designs.
Though a patent court gave ABC priority, most historians consider it only a precursor of the true digital computer because it failed Turing-completeness. So did the Bombe and Colossus. Thus, they are generally considered to fail criterion (a). (Wikipedia waffles and back-steps about the Colossus rather entertainingly.)
Choosing among the Turing-complete Z3/Z4, ENIAC, and BABY is more difficult. The Z3/Z4 is generally dismissed because it had no architectural descendants. (Some German patriots half-seriously dispute this.)
BABY has a case based on the fact that early versions of the ENIAC were not stored-program and it can therefore be argued it fails criterion (a). Historians have generally chosen not to disqualify it because it was upgraded to stored-program operation. in 1948 before BABY became operational. See
https://www.computerhistory.org/atc...c-an-example-of-why-computer-history-is-hard/
Blunted Sickle relevance: if ENIAC hadn't happened, would BABY have? pdf27 appears skeptical on the grounds that the designer of Colossus, absent the later phase of WWII, would have continued upgrading telephone networks. The reason I project that BABY would have been built anyway, even without Colossus, is Alan Turing. He had already designed the Bombe at POD, he had formalized the idea of Turing-completeness back in 1936, and IOTL he assisted in the Mark I's design.