Live · reviewed in batches
Every capability, deployment and labour-market finding that bears on the thesis, filed with its source, its date and a verdict. Including the evidence against.
How the register works
The thesis makes claims that can be checked against the world. This register is where they are checked. Each entry records what was claimed, by whom, when it was observed, which part of the argument it bears on, and a short verdict on what it does and does not show.
Entries are found by the CopeCheck evidence engine, which watches lab announcements, research releases and labour statistics and gives each claim a cope score and a verdict. Nothing enters the register until it has been reviewed. It is updated in reviewed batches, not as a continuous stream, and the page below loads the current register every time it is opened.
Three rules govern it. Vendor and benchmark figures are reported as the vendor reported them, and labelled as such. Disconfirming evidence is filed alongside confirming evidence, with the same prominence. And no entry is ever rewritten after the fact to fit the argument.
Reading the categories
Benchmark entries track measured capability on public tests. Vendor and deployment entries record what labs and enterprises say they have built and rolled out. Labour market entries are the ones that matter most, because they measure the dials the thesis predicts will move first. Confirming mechanism entries show the propagation working in a specific setting. Disconfirming entries cut against a premise. Consultancy entries record industry framing, which is evidence of what firms want to believe as much as of what is happening.
What the register cannot do
A register of announcements over-represents what is new and under-represents what quietly fails. Benchmarks measure tasks, not jobs. Vendor case studies are marketing. The labour-market series are early, noisy and sometimes in conflict with one another, as the New York Fed and Stanford findings on junior hiring currently are. The register is evidence for and against the thesis, not proof of it. The standard it is measured against is Appendix I.
The register is also published as structured data at thesis.copecheck.com/data.json for anyone who wants to test it independently.