Ask HN: Programmers who don't use autocomplete/LSP, how do you do it?

Hacker News Discussion ↗

How People Work Without Autocomplete/LSP

  • Many use plain Vim/Neovim/Emacs (often near‑stock) plus CLI tools: grep/rg, find/fd, fzf, ctags/etags, sometimes cscope.
  • Typical setup: editor in one pane, shells in others for compilers, linters, tests, REPLs, watching logs, docs or browser on another screen.
  • Navigation techniques: search for definitions by name, use tags, or maintain mental maps of file layout and naming conventions.
  • Some rely only on syntax highlighting and auto‑indent; others even disable highlighting and all pop‑ups.

Perceived Benefits of Minimal Tooling

  • Forces deeper understanding of languages, libraries and codebases; people say they remember more and can reason “offline” without tools.
  • Encourages better naming, file organization and smaller, clearer modules so navigation via search is enough.
  • Reduces visual noise: many find auto‑popups, autoclosing brackets, inline warnings, etc., break concentration and “flow.”
  • Some deliberately avoid heavy tools while learning a new language to build stronger fundamentals.

Arguments For LSP/Autocomplete

  • Strong support for LSP in large, complex or long‑lived codebases; features like go‑to‑definition, find‑usages, refactors and early error feedback are seen as huge time‑savers.
  • Particularly valued in verbose ecosystems (Java, C#, C++, AWS SDKs, ORMs) and for unfamiliar APIs where remembering everything is unrealistic.
  • Some see LSP as primarily a discovery and understanding tool, not a typing accelerator.
  • Others argue skipping such tools on big projects is unnecessary self‑handicapping, likened to refusing CAD in engineering.

Copilot and AI Tools

  • Opinions are sharply split:
    • Fans claim large productivity gains for boilerplate, scripts, tests, glue code and “inline documentation,” plus faster iteration on designs.
    • Critics find suggestions distracting, often wrong, stylistically inconsistent, and say they reduce upfront thinking and long‑term skill.
    • Several note extra review burden and fear over‑reliance will erode the ability to spot AI mistakes.

Personal Preference, Scale, and Neurodivergence

  • Many emphasize it’s not moral or skill‑based: different brains and contexts (e.g., neurodivergence, remote servers, niche languages, huge monorepos) favor different setups.
  • There is tension between “artisan” pride in minimal tooling and “power‑tool” pragmatism; multiple commenters warn against turning either stance into dogma or gatekeeping.