The call for talks for #ElixirConf Chicago (September 10-11) is out!
Now to put together some proposals...
elixirconf.com
#ElixirLang
06.03.2026 17:59
👍 5
🔁 6
💬 1
📌 0
The call for talks for #ElixirConf Chicago (September 10-11) is out!
Now to put together some proposals...
elixirconf.com
#ElixirLang
OpenAI published a repo to orchestrate AI agents built primarily with Elixir (96.1%): github.com/openai/symph...
#elixirlang
News includes Expert LSP releasing its first RC, Elixir v1.20 compile time improvements up to 20% faster, #Livebook Desktop moving to Tauri with #Linux support, a new erlang-python library for ML/AI integration, and more! @elixir-lang.org #ElixirLang #AI #ML www.youtube.com/watch?v=Ed83...
Ah, I haven't had a chance to get back to it yet. I've not lost track of it, though, and I promise to post back when I do get there! 😅
👀
This is a huge milestone.
At Jump, we’ve had @katafrakt.bsky.social working full time getting Expert dialed in for our large codebase. The stability and reliability are now hands down the best #ElixirLang LSP experience I’ve had, and we’re not done yet.
Congrats to all the contributors. 🥂
phoenix_test_playwright 0.12.0 changelog
phoenix_test_playwright v0.12.0 released 🥳
Shorthand to eval JS in browser, assert_path fixes for LiveView navigations, and more.
Also ships with usage-rules.md for LLM coding agents. 🤖
#elixirlang
I learned recently (h/t @josevalim.bsky.social 😄) that you can totally omit the result line in an #ElixirLang doctest. This can be useful for pattern matching, like:
iex> some_thing = MyFactory.create_thing()
...> {:ok, "https://myapp.com/path?" <> _} =
...> MyModule.compute_path(some_thing)
Oh, this is amazing! Yeah, I'll give it a try and get back to you.
Gonna try it today. I had no idea it would be useful for that!
Oooh, interesting!
Is this a boondoggle? Are there other ways you've empowered designers to make small code changes?
I’d looooove to have a standard dev container that designers could spin up fully in the browser, with access to Tidewave Pro + Claude/Codex/whatever, where they could just talk to the locally running LLM and say “this button right here, make it blue.”
(Imagine: “here’s how you generate an SSH key from the terminal so you can pull the repo from GitHub—wait, first here’s what the terminal is.” 😬)
Does anybody have experience using GitHub-hosted dev containers ("codespaces") for #ElixirLang dev?
I *really* want to empower our designers to work on LiveView component HTML/CSS directly (they're already comfortable with those), but getting them a local dev setup is a huge blocker.
Once more, with feeling: My #ElixirConf 2025 talk is live!
www.youtube.com/watch?v=8mum...
Yeah, the bit about differentiating how people were using the LLM was interesting!
Ooh, I like that a lot!
YES. It's really important for teams to establish norms like this. There's no shirking responsibility for the code you ship.
Checks out. Some related research - "employees worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day, often without being asked to do so" "...leading to workload creep, cognitive fatigue, burnout, and weakened decision-making"
hbr.org/2026/02/ai-d...
I definitely, *definitely* have seen the problem where people (generally early-career devs, but often experienced people who damn well should know better) kind of blindly accepting whatever the LLM spits out, and if you ask them why they did something, they go "I dunno, Claude did that. 🤷♂️"
I think my experience (working at a LiveView shop for a year after ~3 years at another) is probably not a good benchmark for LLM use overall. I didn't really take my current job to learn new about coding, and the biggest challenges I'm working on are more about scaling expertise across the team.
Likewise I'll often turn the LLM loose on things that are just a chore... AST matchers are a good example. I *could* figure out the structure myself and encode that, but boy, it's a lot nicer to say "here's a test, make it pass" and again do the "LLM writes, tests pass, I refactor" loop.
I think those are generally cases where there's not actually a lot of learning or personal growth to be had. (You can only implement CRUD operations in a LiveView form so many times before you've peaked in terms of capabilities. 😆)
Hrm... it's complicated for me. I *feel like* when I'm implementing a new feature and kind of go step by step with the AI, reviewing everything as we go (and doing a lot of manual cleanup along the way), there *is* some speedup to be had without a drop in quality.
Yikes. 😬
> Participants who fully delegated coding tasks showed some productivity improvements, but at the cost of learning the library. Our findings suggest that AI-enhanced productivity is not a shortcut to competence and AI assistance should be carefully adopted into workflows to preserve skill formation.
"How AI Impacts Skill Formation" from Anthropic researchers:
> We find that AI use impairs conceptual understanding, code reading, and debugging abilities, without delivering significant efficiency gains on average.
(Continued)
arxiv.org/pdf/2601.20245
Quokka is always the first dependency I add to every new project. Such a great tool! And no more “oh my editor formatted it differently, sorry” or “damn I forgot to sort my aliases and now my CI failed”. Brilliant
