Tyler A. Young ⚗️🧑🏻💻
@tylerayoung.com
#ElixirLang developer working at Jump. Former software dev at Felt and X-Plane. He/him.
I think I agree in spirit, but man, I wish it were easier to restore from a separate table/store when you inevitably need to.
January 11, 2026 at 8:26 PM
I think I agree in spirit, but man, I wish it were easier to restore from a separate table/store when you inevitably need to.
OMG, yes please!
January 11, 2026 at 7:33 PM
OMG, yes please!
Reposted by Tyler A. Young ⚗️🧑🏻💻
I like this setting for VS Code from Ben Frain. JUST show the vertical line for the current indentation, not all of them
```
"workbench.colorCustomizations": {
"editorIndentGuide.background1": "#00000001",
"editorIndentGuide.activeBackground1": "#ff9900"
}
```
benfrain.com/vs-code-high...
```
"workbench.colorCustomizations": {
"editorIndentGuide.background1": "#00000001",
"editorIndentGuide.activeBackground1": "#ff9900"
}
```
benfrain.com/vs-code-high...
January 10, 2026 at 5:19 PM
I like this setting for VS Code from Ben Frain. JUST show the vertical line for the current indentation, not all of them
```
"workbench.colorCustomizations": {
"editorIndentGuide.background1": "#00000001",
"editorIndentGuide.activeBackground1": "#ff9900"
}
```
benfrain.com/vs-code-high...
```
"workbench.colorCustomizations": {
"editorIndentGuide.background1": "#00000001",
"editorIndentGuide.activeBackground1": "#ff9900"
}
```
benfrain.com/vs-code-high...
Reposted by Tyler A. Young ⚗️🧑🏻💻
The release blog post: elixir-lang.org/blog/2026/01... #ElixirLang
January 10, 2026 at 7:41 AM
The release blog post: elixir-lang.org/blog/2026/01... #ElixirLang
Reposted by Tyler A. Young ⚗️🧑🏻💻
Reposted by Tyler A. Young ⚗️🧑🏻💻
TIL Oban's unique job period uses fixed time windows, not sliding windows.
A job with unique: [period: {7, :days}] can still allow duplicates 2 days apart if they fall into different buckets.
If you need sliding windows, use a custom query.
ftes.de/articles/202...
#elixirlang #oban
A job with unique: [period: {7, :days}] can still allow duplicates 2 days apart if they fall into different buckets.
If you need sliding windows, use a custom query.
ftes.de/articles/202...
#elixirlang #oban
Oban unique job periods are not what you think - Fredrik Teschke
Fixed windows vs sliding windows: a subtle but important difference.
ftes.de
January 9, 2026 at 12:33 PM
TIL Oban's unique job period uses fixed time windows, not sliding windows.
A job with unique: [period: {7, :days}] can still allow duplicates 2 days apart if they fall into different buckets.
If you need sliding windows, use a custom query.
ftes.de/articles/202...
#elixirlang #oban
A job with unique: [period: {7, :days}] can still allow duplicates 2 days apart if they fall into different buckets.
If you need sliding windows, use a custom query.
ftes.de/articles/202...
#elixirlang #oban
Yeah, I exaggerate when I say “every.” Mostly I’m just finding that a lot of tools have like a top-level collection of independent tasks that they run on the codebase, and they can be pretty trivially parallelized.
January 7, 2026 at 9:27 PM
Yeah, I exaggerate when I say “every.” Mostly I’m just finding that a lot of tools have like a top-level collection of independent tasks that they run on the codebase, and they can be pretty trivially parallelized.
Think I figured it out. Are you on OTP 28? 28 added optimizations for binary concatenation that I wasn't seeing on OTP 27. If I revert those change, I see a speedup on 28.
Wanna compare?
{:credo, github: "s3cur3/credo", branch: "ty/revert-binary-changes", only: [:dev, :test], override: true}
Wanna compare?
{:credo, github: "s3cur3/credo", branch: "ty/revert-binary-changes", only: [:dev, :test], override: true}
January 7, 2026 at 4:28 PM
Think I figured it out. Are you on OTP 28? 28 added optimizations for binary concatenation that I wasn't seeing on OTP 27. If I revert those change, I see a speedup on 28.
Wanna compare?
{:credo, github: "s3cur3/credo", branch: "ty/revert-binary-changes", only: [:dev, :test], override: true}
Wanna compare?
{:credo, github: "s3cur3/credo", branch: "ty/revert-binary-changes", only: [:dev, :test], override: true}
I'm on a mission to find every Enum.each/2 in the tooling I use and replace it with Task.async_stream/2 😁
github.com/nccgroup/sob...
#ElixirLang
github.com/nccgroup/sob...
#ElixirLang
perf: Run check tasks in parallel by s3cur3 · Pull Request #184 · nccgroup/sobelow
Hi there! Thanks so much for Sobelow. We're using it in CI at @Jump-App and we take its findings seriously. ☺️
This makes the main mix sobelow task run each of its component parts in parallel. ...
github.com
January 7, 2026 at 3:57 PM
I'm on a mission to find every Enum.each/2 in the tooling I use and replace it with Task.async_stream/2 😁
github.com/nccgroup/sob...
#ElixirLang
github.com/nccgroup/sob...
#ElixirLang
Whoa! Congrats Sean!
January 6, 2026 at 1:20 AM
Whoa! Congrats Sean!
If it had been at peak power draw, it would have cost like $15/mo in power, but of course it was mostly idle.
January 5, 2026 at 11:56 AM
If it had been at peak power draw, it would have cost like $15/mo in power, but of course it was mostly idle.
When I was self-hosting SleepEasy, I ran Dokku (a Heroku-like self-hosted PaaS thing) on a retired server I bought on eBay. It had 24C/48T Xeons (older models, but still) and 256GB RAM and I paid under $300 shipped.
January 5, 2026 at 11:56 AM
When I was self-hosting SleepEasy, I ran Dokku (a Heroku-like self-hosted PaaS thing) on a retired server I bought on eBay. It had 24C/48T Xeons (older models, but still) and 256GB RAM and I paid under $300 shipped.
You’re talking paying for colo space, or hosting it at home? I’ve done both at different times. Home hosting is great as long as you have reliable power and internet.
January 4, 2026 at 12:12 PM
You’re talking paying for colo space, or hosting it at home? I’ve done both at different times. Home hosting is great as long as you have reliable power and internet.
Reposted by Tyler A. Young ⚗️🧑🏻💻
A simple thing to kickstart blogging in 2026: katafrakt.me/2026/01/03/e...
Using Elixir head version with Mise
Just my website. Some Elixir, some Ruby, some more exotic languages and general thoughts on tech.
katafrakt.me
January 3, 2026 at 6:38 PM
A simple thing to kickstart blogging in 2026: katafrakt.me/2026/01/03/e...
Hahaha, you don’t have to do that 😆
January 3, 2026 at 3:32 AM
Hahaha, you don’t have to do that 😆
Oh I love stuff like that. 😄 I gave a talk on CPU optimization tricks on the BEAM. Stuff that’s useful to know like once a year at best, but lots of fun when it does become relevant.
youtu.be/cdL8hYueAZA
youtu.be/cdL8hYueAZA
Tyler Young, Fantastic Hacks and Where to Find Them
YouTube video by Paraxial
youtu.be
January 3, 2026 at 3:03 AM
Oh I love stuff like that. 😄 I gave a talk on CPU optimization tricks on the BEAM. Stuff that’s useful to know like once a year at best, but lots of fun when it does become relevant.
youtu.be/cdL8hYueAZA
youtu.be/cdL8hYueAZA
In a context without a user, I’d expect to roll the feature flag out on a percentage-of-calls basis. Day 1, do the new thing 5% of the time, day 2 10%, etc. If it’s a truly global boolean, I don't know why I’d bother with the feature flag… make it a code change and you can always revert. 🤷♂️
January 2, 2026 at 4:16 PM
In a context without a user, I’d expect to roll the feature flag out on a percentage-of-calls basis. Day 1, do the new thing 5% of the time, day 2 10%, etc. If it’s a truly global boolean, I don't know why I’d bother with the feature flag… make it a code change and you can always revert. 🤷♂️
2. There’s no mention of what is maybe the most important aspect of a feature flagging system: being able to roll out flags to some users but not others! It makes me wonder if the author has ever actually used feature flags in prod. 🤔
January 2, 2026 at 1:09 PM
2. There’s no mention of what is maybe the most important aspect of a feature flagging system: being able to roll out flags to some users but not others! It makes me wonder if the author has ever actually used feature flags in prod. 🤔
This is some reeeeeal ivory tower academia stuff. 😆 It’s an interesting thought (I totally believe we could and should be leveraging dynamic recompilation more), but:
1. I have trouble believing you could even measure the time spent branching when your real app has a dozen DB calls to load a page
1. I have trouble believing you could even measure the time spent branching when your real app has a dozen DB calls to load a page
January 2, 2026 at 1:09 PM
This is some reeeeeal ivory tower academia stuff. 😆 It’s an interesting thought (I totally believe we could and should be leveraging dynamic recompilation more), but:
1. I have trouble believing you could even measure the time spent branching when your real app has a dozen DB calls to load a page
1. I have trouble believing you could even measure the time spent branching when your real app has a dozen DB calls to load a page
Reposted by Tyler A. Young ⚗️🧑🏻💻
For those of you wanting to start a new project in the new year but afraid you're gonna suck, heed the words of the ancient texts:
December 31, 2025 at 11:14 PM
For those of you wanting to start a new project in the new year but afraid you're gonna suck, heed the words of the ancient texts:
Unbelievable. 🤯
December 29, 2025 at 12:26 PM
Unbelievable. 🤯