luka

@ajambrosino how big is the codex app team?

i am kidding. i remember early on being 1% annoyed that we couldn't make everyone at OpenAI use the Codex app and the app is 10000% better because it was an organic choice

next time i get grilled to grow faster i'll just pull this up

Introducing the Lovable mobile app. Your ideas won’t wait. Now you can build them anywhere.

@kly_cl very beautiful

first ship at OAI :) it took a village, we hope you enjoy

@msllrs @iamdavidhill @opencode is this designmogging?

deep breath in (i think you already know my choice). ive gone with option 1 for the past few systems ive worked on, and also took this route when defining the new color system for lovable (coming soon). with option 1, as you point out, token count stays manageable. with option 2, every intent times every state times every variant would blow up the token set, and potentially box you in. option 1 keeps it composable... a small set of state tokens that layer on top of any color. option 1 also scales with how dynamic your color system is. our main intent set is fixed, but the actual values arent hardcoded... theyre derived from seed colors through an oklch derivation engine that generates the full token set (solid fills, muted fills, foregrounds, borders, shadows, gradients, and "glows" (our term for these state overlays)) across ~4 themes automatically. with option 1, the glow overlays are color agnostic. theyre just something like oklch(255 255 255 / 0.12) compositing on top of whatever fill is there. so if we tweak a seed color, add an intent, or eventually open it up to user defined brand colors etc, the existing hover/pressed/whatever tokens just work. option 2 would mean hand picking or precomputing unique state values for every base color across every theme... the whole system is built on oklch, so its perceptually uniform. equal numeric shifts produce equal visual shifts, so our color ramps have consistent visual weight across hues. that means a white 12% overlay on a blue button looks the same intensity as on a green or red one (or close enough). the contrast ratios stay consistent across the ramps too, which is why the stacking feels coherent regardless of which color it's placed on. hover tokens are split into categories (neutral, intent, and inverse) which gives fine grained control without the token bloat. we also have "on surface" glow tokens for things like outline/ghost buttons where the hover/pressed is the fill itself rather than an overlay on an existing fill. in the case of lovable we use pseudo elements (::before for gradients, ::after for glow overlays) rather than color-mix(). in previous systems ive leant on color-mix because components were flat. but for lovable we wanted gradient fills on solid buttons etc, and color-mix struggles/cant express gradients simply. pseudo elements let us stack gradient + glow + box-shadow borders independently. the tradeoff is that designers need to think in stacked fills rather than flat swatches in figma. option 2 is more wysiwyg for them. but once the mental model clicks, its actually faster(ish) imo... they dont need to hunt for or maintain a unique token for every state of every color. they just stack a base fill and a gradient fill with a state fill and the composition handles the rest. box-shadow borders simulate inset borders and highlights, so the stacking of fill + gradient + glow + border all composites cleanly without layout shift or border-box math issues. the layers sit on top of each other naturally. it all just works. so option 1 isn't just simpler in dev, it's architecturally "better" because the state layer and the color layer are independent concerns. add a new intent color? the existing hover/pressed tokens just work. no new state tokens needed.

we're exploring options for our color system given the differences between figma and dev implementation there are pros and cons to different approaches option 1 feels simpler and more dev aligned option 2 feels safer but more design aligned 🤔

@miralizain @ayushsoni_io @alifdotbuild @helloHEX huge

I'm excited to announce I'm joining hex.inc as a Partner. Every founder in the @alifdotbuild ecosystem now gets access to world-class creative direction – by the same studio that's worked with Replit, Whop, Framer, Exa and more. DM me if you're looking for design/brand/videos.

@PeterMmuo @Lovable just work and have fun is how it worked for me

@lukaivanovic @Lovable Congratulations bro How does one achieve a feat like that tho

i've joined @Lovable :)

@davidhoang @Lovable thank you!! 🙏

@lukaivanovic @Lovable Congrats, Luka!

@aydaoz @Lovable i have some tech background, i started designing but transitioning towards eng for the last two years

@lukaivanovic @Lovable congrats! Im curious what are the skills u need to have fo design engineering? were you a dev or designer in origin?

@stevelauda_ @Lovable thank you! 🫡

@lukaivanovic @Lovable congrats man! Happy for you and well deserved!

@emirayaaz my first dislike click on x jk, i also started with lua scripts for gta

@lukaivanovic i was playing mta lol

1 year since arc's first paid client > all solo > $90k in revenue > 15 companies, 6 yc > 4700 followers on x > prices $1,600 → $5,000 still feels unreal. the road here: > 2021, started coding to write scripts for gta:sa, 13yo > made my first internet money selling those scripts > 2023, joined my friend's startup, learned web & design, 15yo > existential crisis > 2024, ran a marketing agency with a friend for local turkish businesses > cofounder left for studies > jan 2025, started arc for local us/uk businesses, 16yo > march, pivoted off local businesses > started posting on x > march 17, 100 followers > april 4, first arc client — sim (yc) > april 24, second client at 7x the first > may, moved to design partnership model > june, sim offered full-time // paused arc > september, left sim > second yc company came in > back to $10k+ mrr in 10 days > october, broke up with my gf of 2 years > november, best month ever > closed 2025 at $65k > feb 2026, worst month, $0 > april, we are so fucking back

@GrizzShutsDown very cool

I saw luka's effect and i had to add some details and love to it.

@faustisches this is beautiful

Color around and find out — a spectrum-driven system exploration for an unboring* AI product

@Brockandersn Did you find someone for the eng?

want to scroll reels, but forced to review code? i made preels™ - app that let's you scroll through your organization's pull requests as reels built with @Remotion @ElevenLabs @elysiaJS

@SamuelBeek @schematikio A little jumpy, but looks really good I think people will increasingly get used to just waiting for certain tasks e.g. i dont mind codex taking up to 15m right now

One of the hard bits with @schematikio is that initial generation takes 90 seconds I've tried to decouple it in small tasks to make it look "faster" thoughts?

I was constantly coming back to this amazing website from @shuding but needed some more control over stop positions i started building a similar tool with stop controls and oklch default colors, link below

@reve it would be amazing if some of these are available on the api 🙏

Make specific edits