https://expanse.to/blog Insights on AI infrastructure, HPC, GPU utilisation, and the engineering behind compute certainty. en-gb https://expanse.to/blog/measured-on-a-national-supercomputer https://expanse.to/blog/measured-on-a-national-supercomputer Tue, 27 Oct 2026 00:00:00 GMT Ismaeel Bashir How an early version of Expanse's prediction model was built and measured against ground truth on production clusters at a national supercomputing centre, and what a month on a national facility showed about how much compute is paid for and never used. https://expanse.to/blog/expanse-raises-5-3m-seed https://expanse.to/blog/expanse-raises-5-3m-seed Wed, 16 Sep 2026 00:00:00 GMT Ismaeel Bashir Why we started Expanse, and what the $5.3M seed led by Crane Venture Partners goes towards. https://expanse.to/blog/frontier-models-guess-at-compute https://expanse.to/blog/frontier-models-guess-at-compute Tue, 02 Jun 2026 00:00:00 GMT Ismaeel Bashir We gave eight frontier models the code and submission script of 665 real HPC jobs and asked what each job would need. None of them got close, the confidence ranges they gave were made up, and Expanse was 8x more accurate than the best of them working alone. https://expanse.to/blog/its-2026-and-were-still-writing-slurm-scripts-by-hand https://expanse.to/blog/its-2026-and-were-still-writing-slurm-scripts-by-hand Sat, 28 Feb 2026 00:00:00 GMT Ismaeel Bashir Why the friction between scientific intent and cluster execution remains one of the most solvable hard problems in research computing.