More and more code is written by language models and read by people. The best language for that partnership is the one where intent is easiest to say, easiest to check, and pleasant to live with. That's Ruby.
Language models learn from human language first, and Ruby was designed to sit close to it. The code says the intent, so a model has less to translate, and the person reviewing it can see at a glance whether it's right.
active = users.select(&:active?).sort_by(&:signed_up_at)
send_reminder(user) unless user.subscribed?
3.times { retry_upload }When a reviewer can read code aloud and hear the requirement, mistakes have nowhere to hide.
No type annotations on every line, no boilerplate classes, no build step to run a script. Every token of a model's context window and every minute of a reviewer's attention goes to logic, not scaffolding. A small program stays small.
Ruby's conventions put information right in the name: empty? returns true or false, sort!
changes things in place, save! raises when it fails. Models pick these patterns up reliably, and people
can tell what a call does without opening its definition.
Ruby culture, and Rails especially, settles most decisions in advance: where files go, how things are named, how tests are laid out. Fewer open choices means fewer ways for a model to guess wrong, and code from many authors, human or not, ends up looking like it came from one team.
Agents work best in a tight loop: write, run, check, repeat. Ruby runs instantly, irb answers
questions in seconds, and the standard library covers JSON, HTTP, files, processes and templates out of the box. Testing is
part of the culture, with Minitest built in, so "done" can mean "the tests pass" from the first line.
Matz designed Ruby to make programmers happy. That matters even more now that much of the code is generated: people spend their time reading and reviewing it, so code that is pleasant to read is code that actually gets read carefully.
GitHub, Shopify and many others run on Ruby. Two decades of open-source Ruby means models have seen a lot of idiomatic, well-tested examples, and the gems they reach for are mature.
Dynamic typing. Errors that a compiler would catch show up at runtime instead. Tests close most of that gap, and RBS or Sorbet add types where you want them.
Raw speed. Ruby isn't the fastest language, though YJIT has narrowed the gap and most real workloads wait on networks and databases, not the CPU.
Machine learning. Python owns model training. Ruby is for everything around it: the services, tools, glue and products that use models.
Provided as is, for information only, with no warranty and no liability. Not legal or professional advice. Security and radio material is for authorized, lawful use only. Full disclaimer.
Visits are logged on our own server: IP address, browser and device details, and pages viewed. We use them to understand traffic and spot attacks, and never share or sell them.