Go code: CLIs, libraries, compilers, and the naming conventions that hold it together. Opinionated notes on a language that rewards restraint.
Four months ago I asked ChatGPT a dumb question while walking my dog. It told me building a programming language wasn’t that hard. Now I have a compiler, a type system, generics, structured concurrency, and a garbage collector.
Stateless agents repeat mistakes. lgtm and snap use a findings file, shared codebase context, and recovery loops so later tasks can use what earlier tasks already learned.
Why I built mokapot — a single Go binary that speaks SQS and SNS well enough for real AWS SDKs to connect without configuration changes. Persistence, filter policies, and the tricky parts of faking a cloud service locally.
Four real examples from recent Go CLI projects — bracketed paste handling, a scripting DSL, safe symlink restoration, and diff truncation — that show how small decisions separate a script from a tool people trust.
pin is a Go spinner with colors, prefixes, message updates, pipe detection, and no external dependencies. It exists because a spinner should not drag a UI framework into a small CLI.
I needed a spinner for a Go CLI and ended up inside a UI architecture. Some terminal libraries are great. They are also the wrong tool for a linear command that just needs progress output.
Go 1.23 added range-over-function iterators and the iter package. For CLI progress streams, that can replace channels, goroutines, and hand-rolled state machines.
Go uses names for more than readability. Casing controls visibility, short variables signal scope, and consistent verbs keep packages understandable.