Looking for an #ElixirLang dev? I'm officially on the market!
4+ years of Elixir in production. Natural affinity to testing, o11y and DX. My most used OS libraries are InstructorLite and PostHog SDK.
Remote from ๐จ๐ฆ, PDT, links in bio.
RTs appreciated ๐
#MyElixirStatus
Elixir v1.20 released! Now officially a gradually typed language: Elixir type checks every single line of code, finding bugs and dead code, without developer overhead (no typing signatures) and extremely low false positives rate. Plus a faster compiler! Links and reports below.
I once again have some extra free time on my hands, so I'm going to dig into why DBConnectionErrors are seemingly never grouped correctly and see how many yaks require shaving.
Our first stop is naive `Logger.error` calls. We need to figure out what to do with them! https://t.co/nkeNxTa10x
The problem is that you won't know if a patch is good quality until you already read it. So the cost asymmetry is very skewed in favour of the contributor vs the reviewer.
Basically, people are asked to spend some demonstrable effort upfront, so that maintainers knew they can spend their effort too.
๐ก New Blog Post
LLMs looove markdown, but the best chat interfaces go beyond and let models use rich UI elements. In this post I demonstrate one possible way to achieve this: by teaching models to use custom HTML elements
https://t.co/6vCC0uRWGS
#MyElixirStatus#ElixirLang
People talk about LLMs discovering and fixing bugs like it's normal and inevitable part of development process. But it's not!
Ask your QA, they know that bugs aren't finite set of issues waiting to be uncovered, but more like a quality gauge. If you see a lot of them, then even more have not been discovered, and you might want to step back and reconsider.
Sure, some bugs are normal, but they're a valuable signal first and foremost.
Code reviews are definitely dead and many of us, including myself, are grieving. I don't know how things will turn out, but I felt the need to write a farewell. โฌ๏ธ
@ZachSDaniel1 Agree, I think non-anthropomorphizing ways of working with LLMs are very underdeveloped. It doesn't help that LLM providers pretty much locked everyone into the chat completions API ๐ข
For a lot of people it doesn't seem very healthy to frame AI this way https://t.co/ZO6CNEfouk
Maybe I'm dating and deprecating myself in a single stroke, but I genuinely struggle seeing a productivity increase from agentic coding. I try, I retry, like every 2-3 weeks. I always end up tired and slow, either from hand-holding in real time or spending hours undoing the slop.