CodeMonument

Just ported @ryanfeigenbaum.com's amazing color logic into a standalone, MIT-licensed library. Perceptually uniform palettes for your next generative art or design project. 🎨 meodai.github.io/pro-color-ha...

This sounds like htmx but directly in the browser 😃

Nice!

Gemini 3 Pro did an oooopsi-daisy in cursor for me.
I told it to improve the ui of my landing page and it did. Then I wanted it to do some changes and it realized: "hey, some code exists already, i don't have to print it again!".
And cursor thought: Yes, good idea, lets replace all existing code!

In the end: telling jokes from real takes as text on the internet proves maximally difficult again 😅😂

But on the other side, if we want to have a welcoming space in tech we need to stand up for it. Probably not canceling DHH, but clearly speaking up against the views he throws into the world if they don't align with one's values.
Simply freedom of speech in both directions 😄

Oh okay :O but what is jokey about it when there is a real argument to be made here?

Okay, one could ask the question: what does his omarchy set of dotfiles and configs have to do with its political view?
So: framework is "simply" supporting tech they like/see need of.

Same thing as here in Germany. 😞

As I remember London As soon as I was old enough to travel on my own, London was where I wanted to go. Compared to Copenhagen at the time, there was something so majestic about Big Ben, Trafalgar Square, and even the Tube...

Here's the original dhh article with his claims.
world.hey.com/dhh/as-i-rem...

My opinion:
This is so sad and scary to see. Mixing a tiny tiny core of the real problem of not-successful police prosecution of criminals with the racist views and talking points of the extremist right. :/

But: I couldn't find sources for the claims yet. Will dig deeper, since trusting some article blindly in this internet age is bad 😃😄

I searched what framework has donated to and it seems it was DHH?!? 😳 what the f*ck is our timeline! I thought: yeah, this guy might have a lot of controversial tech-takes but also a lot of experience into he field. But now this...
Here's an article about that.
www.osnews.com/story/143520...

One alternative idea I had, but rejected:

Create a new SQLite file for each test, then run drizzle-kit push for the schema and then use drizzle-seed to fill the db with demo data based on my schema.

But I wanted to have the speed of in-memory and temporary files seemed a magnitude slower to me.

- let the server use drizzle migrations on startup, which where generated with drizzle-kit generate before
- restore the state of the server completely via public API (aka: user registration, login, doing stuff on the page)

I'm coding a dashboard tool for myself getting some time tracking entries from a provider, caching them and showing some aggregations to the user.
It uses: drizzle as orm and SQLite.

I decided to:
- don't touch the database in my tests
- pass ":in-memory:" as database url

Yeeees, this is ultra-cool 😍

Extremely nice having sources and source maps on traces! 😍

Massively complex environment (aka the hardware stuff)

Basically what I imagine:
If you have a vite app, running unit tests in vitest, component tests in vitest browser mode and e2e in playwright, this is kinda possible.

Replace any part of this stack and some things get impossible due to slow test tooling, overly complex business setups or simply

Because playwright has its own runner 🙏

But cache invalidation still has to be considered strongly.

At last: a good portion of the test still use jest, so there is legacy stuff to migrate to modern, faster and more compatible test runners, e.g. vitest.

I'm grateful that I dont need to migrate the playwright tests

This sandbox could be docker, we're using bob build tool right now but want to fully switch to docker.

The infrastructure must be fast enough to build code (we build a custom electron app with native rust and cpp extension code in it)
- not a real issue for testing of cached correctly,

Your tests should be written so that they can be relieably executed (not easy if you have dependencies on real/emulated hardware state)

The infra must be able to sandbox those test environments beyond what playwright offers with the browser (for the hardware emulation I mentioned)

The infra must be fast enough to handle the need to basically bootstrap the full environment for each test.

The infra (aka ci runner and my dev laptop) must be able to get all required data easily, like internal git repositories, internal artifactory repos, external repos, if needed and secrets.

Introduced a loooooot of shared state into the system again.
Mitigating this with mocking was also easier said than done and also not the most optimal solution.
So the solution we settled on in the end was to break down the test cases even further and have more of them in general.

I also worked at another company before where we couldn't simply authorize like in your example. We had to ORDER new test users, like steak on a menu and infra had to prepair them for us. Since we couldn't have hundreds of them, we had to re-use them between tests which

I'm currently working on a project where isolating the test runs into sandboxes tooks a looong time, since hardware components need to be emulated and primed with special state. Also, our tests are kinda air-gapped (with only some artifactory) which limits available packages and testing env.

No, I love this thing! This is the dream!
Unfortunately, this sets an extremely high bar on the infrastructure, tooling and coding environment to be able to pull this off.

Nice! Finally! 😃 you're probably the only reason why I still have a Twitter account (and random trash posts ofc)

Linus Torvalds being unfathomably based