Ismael Canal

Decidido, a partir de hoy empiezo a publicar todas las ideas locas que me vengan a la cabeza! Es un ejercicio para practicar la escritura orientada a "benefits" VS "features" usando el estilo de nota de prensa que comentan en el libro de Working Backwards de Amazon.

@RamsesCabello Love it 🥰

🌸🌀🌟 just created ゆったり as a palette that has inspired me over the years. my website, most of the apps i use and all bits of technology are tinted by these colors. read more & download ↓ yuttari.ramses.work

Did some small collab with Tiago here to make sure our color engine works nice on Windows, MacOS and Linux

@zehf ❤️ love the style

@basiclines Yess!! We do a less cinematic version every launch week at Resend.

What if websites had credits?

@jonathantrevno So tempted to do it Star Wars style but fell odd in context

@basiclines star warsss

@hmorneau @DerekLegenzoff Thanks! We just have an open PR with the fix applied. Hopefully will land during this week!

@DerekLegenzoff Anyone else what their copilot cli randomly changing color. I went from black to red. It's happening quite often right now.

We just shipped some improvements to gpt-5.4 and gpt-5.3-codex in the Copilot CLI. The models are now better at providing updates as they work and have better formatting for their final responses. We have more updates planned too. If you have feedback on what else you'd like to see improved with these models, let us know!

@mattsilx You can actually check here: github.com/basiclines/ram… Supports rgb, lab and oklch interpolation

that’s basically what I do - rely on css vars so i can reassign them. e.g. text-base-11 / bg-base-4, then might tint via a utility that remaps the vars. So same element has same contrast regardless the hue being adopted. Curious about your formula for generating shades while keeping contrast consistent across hues.

What if websites had themes? 🎨

@JohnPhamous @adrianmg Love it. How did you read Arc theme values?

@adrianmg @basiclines x.com/johnphamous/st…

@mattsilx Then your contrast "contract" is just to make sure you have enough steps between backgrounds and foregrounds. Testing is a bit trickier because accessibility will depend a lot on the theme picked, so you need to test accessibility on mass, like hundreds of themes

Interesting! Any reason for not using oklch? I've been calc()-ing over oklch values to derive a base color from a primary + secondary pair. But we're really solving the same problem: how do you enforce contrast and hierarchy rules? My current approach: a 12-step shade scale per color, with defined usage for each step range. But feel a better api is approachable...

@mattsilx actually the interpolation uses oklch, but instead of manually creating the shades, i just create them programatically and append a css style tag with them

@mattsilx I've been exploring ways to dramatically simplify theme definitions to just 8 colors: foreground/background/+6 hues (RGBCMY). Then, interpolate each hue between FG/BG with different levels of depth. Use those colors in the components instead of the raw color values

@basiclines Been working on something very similar! Most systems nail assignment but completely miss that colors relate to each other. Blending that with component hierarchy is where it gets really hard. Curious to see your mental model!

Our new Color Engine is heavily inspired by the amazing work the Ghostty team did here: x.com/mitchellh/stat…

Copilot CLI now reads your terminal theme colors and builds upon it for a much richer experience. Works everywhere, zero setup. Literally infinite themes!

❯ Copilot make sure the TUI works in ALL themes. Don't make mistakes.

@stuebe2k14 If you use a code block they should look the same, but beware of tighter widths

@basiclines can you copy paste them well?

🧑‍🍳 Fresh out of the oven! New markdown tables are available now on Copilot CLI. Make sure to update!

@willmcgugan We will make it better! Thx for the examples this is really helpful

@basiclines You're gonna hate me, and I wouldn't blame you. The tables break when you crush them a bit. There is what i'm assuming is partial escape sequences. The column width calculation is also sub-optimal (see video).