Mehjabeen Arsiwala

Mehjabeen Arsiwala is an assistant editor at Fathom, covering AI research and the frontier labs. Her beat is the work itself: model releases and the claims made for them, benchmark results and what they do and do not measure, published papers, evaluations, and the safety and alignment research that rarely makes headlines until something goes wrong. She reads the model card and the paper before the write-up, on the view that most of what matters in a launch is in the technical appendix rather than the announcement. She writes about the competition between OpenAI, Anthropic, Google DeepMind, Meta and Mistral as a research contest rather than a horse race, and treats a benchmark number as a starting question rather than a conclusion. She is also the responsible editor for Fath Consultancy Ltd, which publishes Fathom, and handles corrections and right-of-reply requests directly.