https://sparsenotes.com/series/weekly-papers-2026-08-08/ Recent content in weekly-papers-2026-08-08 on Sparse Notes https://sparsenotes.com/images/og-default.png https://sparsenotes.com/images/og-default.png Hugo -- gohugo.io Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/papers/architectural-implications-agentic-ai-workflows/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/papers/architectural-implications-agentic-ai-workflows/ The first production-fleet architectural characterization of agentic AI, showing that host orchestration puts the CPU back on the critical path and that uniformly provisioned servers are the wrong machine for the job. https://sparsenotes.com/posts/2026/08/papers/communicating-sequential-processes-hoare-1978/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/papers/communicating-sequential-processes-hoare-1978/ The seminal paper of the week: Hoare's 1978 CSP, which proposed that input, output, and concurrency are primitives — and gave us the intellectual foundation for occam, Erlang's mailboxes, and Go's channels. https://sparsenotes.com/posts/2026/08/papers/topology-role-ml-training-performance/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/papers/topology-role-ml-training-performance/ An analytical head-to-head of fat-tree Clos vs. torus for ML collectives, showing Clos wins almost everywhere except AllGather — and wins decisively once failures and placement enter the picture. https://sparsenotes.com/posts/2026/08/papers/ Sat, 08 Aug 2026 00:00:00 +0000 https://sparsenotes.com/posts/2026/08/papers/ This week's picks: interconnect topology for ML training, the first production characterization of agentic AI workloads, and Hoare's CSP as the seminal paper of the week.