Adam Neumann is out as WeWork CEO, Tuesday, under investor pressure, the IPO shelved indefinitely, the prediction chain closing its arc in under six weeks from S-1 to defenestration. The final-week reporting delivered details beyond parody (the tequila-and-Run-DMC layoff meeting from 2016; “cereal entrepreneurship”-grade mission language; a private jet story involving a cereal box that I decline to summarize in a family archive), but the structural postmortem is the keeper: every guardrail, board, bankers, auditors, late-stage investors, had economic incentives pointing the same direction as the founder’s narrative, so the system had no adversarial reviewer until the public S-1 process installed one. Code review works because the reviewer doesn’t share the author’s sunk costs (corporate-governance edition). Institutions that pay all their reviewers in the author’s equity should expect author-quality review.
The fortnight’s second story leaked early: a paper briefly posted (then pulled) on a NASA site claims Google achieved quantum supremacy, their 53-qubit Sycamore chip completing a (carefully chosen, practically useless) sampling task in ~200 seconds that they estimate would take a classical supercomputer millennia. IBM is already drafting a rebuttal about better classical algorithms. Measured take pending the official publication (fifteen days, probably): “supremacy” on a benchmark designed for quantum advantage is a Wright-Flyer moment, it doesn’t deliver anything, it exists, which is different and still enormous. The Area 51 raid meme also peaked this fortnight (“they can’t stop all of us”; ~150 people showed; two arrested for, essentially, vibes), logging it purely because future-me deserves to remember that 2019 contained multitudes.
TIL: cross-entropy benchmarking, how you verify a quantum computer’s output when no classical machine can check it directly: statistical fingerprints on samples. Trust-but-can’t-verify is about to be a whole subfield. (Blind injections will find a way; they always do.)