Magnus Coldwell
Benchmarking the Benchmarks: Why Current AI Evaluations Measure the Wrong Things
A systematic review of 147 AI benchmark datasets reveals that most popular evaluation methods are testing for capabilities that do not predict real-world performance. The gap between what we measure and what matters is growing.
Beyond Poynting: A Forensic Look at Energy Transfer and Maxwell's Limits
A forensic dissection of the Poynting vector reveals that our standard picture of electromagnetic energy transport has been hiding assumptions — assumptions that break down in extreme conditions.
Room-Temperature Superconductivity Claim Survives Replication — This Time With an Explanation
A team at Caltech has independently confirmed lutetium hydride superconductivity at ambient pressure AND provided a lattice-level mechanism. The field is cautiously watching.
Permafrost Carbon Feedback: The Number Climate Models Keep Getting Wrong
A new synthesis of Arctic soil measurements suggests carbon release from thawing permafrost will outpace optimistic projections by a factor of three. Here's why the models have been underestimating it.