Emacs

Emacs is where most of our work actually happens. Not because it is the fastest editor — it is not — but because it is the only one that has been continuously maintained for four decades and can be bent to the shape of the job.

Why we still choose it

Because a working environment is a long-term investment, and the ones that ask you to migrate every five years quietly tax everything you build. Configuration is a program, kept in Git like anything else, so an environment moves between machines and years without ceremony.

The practical value is that everything lands in one place. Editing Nix expressions, reading a Rails codebase, writing Markdown for a Jekyll site, running a shell, and — increasingly — talking to a local model through Ollama, without the context switch of a different application for each.

Org-mode deserves a specific mention: notes, runbooks and literate configuration in plain text that will still open in thirty years, which is more than can be said for most tools that have tried to replace it.

Where it meets the local models

Emacs turns out to be an unreasonably good interface to a language model running on your own hardware, for one reason: it is programmable all the way down. Wiring a model into a buffer, a region or a code review is a few lines of Lisp rather than a plugin somebody else has to ship — so we offer that as an engagement in its own right, alongside Ollama.

The honest limit

The learning curve is famously steep and we do not impose it on anyone. A proportion of any team will bounce off it, and that is a preference rather than a failure. Everything we hand over degrades gracefully: runbooks export to plain Markdown, configuration is readable in any editor, and the model endpoint is OpenAI-compatible so a colleague can point their own tooling at the same GPU and lose nothing but our key bindings.