The most encouraging part here is that they published full calculations, code, and caveats rather than just a press release. DFT (especially PBE+U, and even HSE06) is notorious for getting band gaps and magnetic ordering energies wrong, and the YBaMnFeO5 candidate requiring perfect Mn/Fe checkerboard ordering is a huge synthesis ask — disorder could kill the compensated state entirely. Same with the 420K -> 490K calibrated Neel temp: calibration against a known magnet helps but doesn't remove systematic error. That said, as a screening workflow this is exactly how agents should be used: fast search over composition/structure space with a cheap objective function, then human-readable artifacts others can reproduce and falsify. The real test isn't the simulation, it's whether a lab tries to synthesize these and reports back, negative result included.
The redaction failure itself is almost the bigger story than the numbers - highlight, copy, paste revealing 52.65 MW peak and 13.3 megagallons is such a classic PDF redaction fail. Someone just drew black boxes over text instead of actually removing it.
On the substance: 13M gallons sounds huge until you see the context in the article - less than half of what the City of Lincoln uses in a single day, and about 20 Olympic pools for a year of operation. The Papillion site at 548 megagallons is the much larger water user. And the tax refund numbers ($55.8M for Agate alone) explain why Nebraska's reporting law and the governor's executive order exist - locals want to know what they're getting vs giving.
Claiming electricity/water usage as trade secret under 81-1527 feels like a stretch when the state explicitly requires annual self-reporting to DWEE. If everything material is redacted as trade secret, the reporting requirement is meaningless. Would be interesting to see if DWEE fixes their submission portal to reject non-flattened redactions.
