Is the ARC Theory science, or only a hypothesis?
A hypothesis with published tests and published abandonment conditions is science in progress, and this one is not yet accepted science. The register holds 44 kill conditions with live statuses, one fired; the deciding studies are written and awaiting human submission; there is no peer review and no outside replication.
Take the standard first, because it is about science and not about this programme. A hypothesis that states what would refute it, publishes the conditions in advance and records what happens when one fires is science in progress. Acceptance is a later stage. Peer review is a venue, and affiliation is an address. Neither is the definition of the thing.
Conceptual synthesis under a new assumption is how discoveries have been made, and every law is written in borrowed mathematics. Recognition arrives with independent empirical testing, not with the synthesis. This programme's synthesis is stated and dated, 8 December 2024, its deciding studies are written, dated and prepared as draft registrations awaiting human submission, and its own first measurements are published with their intervals and their corrections. No outside laboratory has run a test of it.
Now this programme's position on that standard, without raising a grade. The falsification register publishes 44 kill conditions with live statuses, and its own semantics say that an open condition is not evidence of support. One has fired. Nothing here is presented as established, and the home page opens its claims section by saying so.
The retraction is the method working. An early estimate of 2.24, the most striking number the programme ever produced, was retracted in public when cross-architecture replication did not hold at that magnitude. What replaced it came from the programme's own blinded six-model protocol: roughly 0.49, interval [-1.3, 2.9], printed with the interval every time it appears. A second figure, the earlier Fisher-combined one, was withdrawn because a Fisher combination assumes independence among component tests that was never established.
The concessions stay, in the site's own words: the work is self-published, is not peer-reviewed, and the author holds no institutional affiliation, no doctorate and no journal's imprimatur. What follows from that is a stage, not a verdict. The protocols, data, code and predictions are public precisely so that a group other than the author can decide it without permission.