Fuzzy logic is less fashionable than it was in the 1980s and 1990s. That is not the same as unused. These three notes ask whether it still earns a place, what the current libraries look like, and what public evidence there is of real deployments.
Part 1 is the argument: graded predicates, readable rules, and a toolbox that does not expect an LLM to do inference, vagueness, and policy all at once. Part 2 is the survey of MATLAB, FuzzyLite, Python, Julia, Rust, Elixir, and Zig. Part 3 is the 2026 landscape: who documents production use, which packages are research-grade, and where a systems-language or Datalog-shaped implementation would still be new work.