I recently started playing with Igniter and it’s a great code generator already as it is. With Chris’ keynote, I can also see our trend as a community towards the next generation of codegen tools. My hunch is that Igniter can stretch even further in that direction. Codegen can become an edge for the Elixir community.
I won’t push too much nuance into making a vibecoding case one way or another but this is my attempt to start a discussion:
- What’s your thought on Igniter, Phoenix.new and our present codegen possibilities?
- What directions do you imagine we take things in the future?
- What will you recommend as next contributions from the community?
I was simply going to contribute back my learnings to the igniter docs (my answer to 3) and watch the broader things unfold, but I decided to make this post to hopefully elicit better ideas
EmberJS had what they call codemods for ages.
It still managed to be left behind, so I doubt it’s gonna be the killer feature for the elixir ecosystem.
phx.new is like any other generator: just produces code and is entirely clueless after. Igniter I heard good things about – including that it might actually be able to modify code – but haven’t had a chance to try it.
To me, if we get to the point of having tools being able to modify our code in meaningful ways (almost akin to what LLMs could do), then it becomes a competitive advantage and likely many people would not ever open-source that.
Well if it can’t modify code and install itself non-disruptively then to me it’s mostly useless. One-off code generation is fine to start with, and Phoenix’s project and auth generators are quite good, but beyond that there’s much more fruit to pick.
Meaningful modifications to existing projects. But that’s a huge topic in itself and it can become quite academic.
It’s getting difficult these days. phoenix.new is not the same as phx.new. Maybe you mixed them up.
Very likely. I use them like 3 times a year.
Phoenix.new is “new”: https://phoenix.new/
Chris announced it recently:
To add to the possibilities, there’s also Tidewave (MCP server connected to our own Phoenix app) from José & team: https://tidewave.ai/
1 Like
I believe it could help especially coupled with main features of elixir like fault tolerance , concurrency and high scalability. Tools that can generate easily most parts of required building blocks for an application would make Elixir the go to option with all this emphasis on quick POC and go to market.
Besides the tools mentioned above , Please note just recently @zachdaniel introduced also Ash AI which is focused on enhancing your application with AI capabilities which is something already started getting being requested and you can see big enterprises moving into this direction.
I think there is still very much a place for igniter in a world of LLMs 
I’ve tried to get an LLM to bootstrap a project from scratch with even a few of the Ash dependencies and it got nowhere close (unsurprisingly). this installer is powered by igniter https://ash-hq.org
Packaging up changes as deterministic codemods is still a very big deal and is a rich vein we’ve barely even tapped at this point.
2 Likes
