i kept seeing these Claude Opus 5.5 motion design videos all over X, so obviously I had to try it with SketchMason
I gave Claude one prompt:
“make a 15-second motion graphics showreel for SketchMason. go all out.”
- no After Effects.
- no video model.
it basically wrote the entire film in code.
15 seconds = 900 frames at 60fps.
Claude wrote roughly 5,000 lines, rendered every frame in headless Chrome, synced the whole thing to a 128 BPM timeline, generated its own soundtrack from raw waveforms, added motion blur, film grain, colour fringing, and even checked whether the beats were landing on the correct frames.
the crazier part?
it reviewed its own render.
it generated contact sheets, looked at them and started fixing things a motion designer would notice:
- pen tip was outside the frame
- words were colliding
- end card was too low
- one 3D shot had too much empty space
final render used 8 Chrome instances in parallel feeding ffmpeg and took around 7.5 minutes.
but my favourite part is that the video actually tells the SketchMason story:
rough pencil sketch to finished interface.
AI generating videos is cool
AI writing an entire motion design system, rendering it, testing it and critiquing its own output feels like something else entirely.
anyway, here’s what it made
one sketch shouldn’t give you one screen
it should give you a system
- same visual language
- multiple screens
- one flow you can keep editing
that’s the direction we’re building SketchMason toward
7 days since I launched SketchMason
I had 9 people on the waitlist before launch. I warmed them up, sent everyone the launch message, and thought at least a decent number would come through
1 signed up on launch day
Right now SketchMason has 7 accounts. One of them is me.
0 paying customers
Instagram has 3 followers
Tiktok has 0 followers
and I’ve been posting on TikTok and Instagram and nothing has landing yet (yeah i think i need more time on this )
X has been slow too, but almost every useful conversation, piece of feedback and signup I’ve gotten has come from here
I built SketchMason alone in 27 days and launched it on my 21st birthday. Before launch I genuinely thought building the product was going to be the hard part.
lol.
Turns out I know how to build. Getting strangers to care is the part I’m learning from scratch
But there’s one number I actually like 6 people created accounts and tried something I built from zero
A week ago that number was 0.
I’m not going to call that traction yet, but it’s enough signal to keep going
Now I’m trying to figure out what the actual bottleneck is: the product, the positioning, or distribution?
If you were in my position, which one would you attack first?
https://t.co/2O6ToZxNxy
day 7 still building
one user asked me to add Google sign in.
that one DM ended up exposing 2 other things i shipped badly
the verification email was taking too long, and some Resend emails were landing in spam.
worse, SketchMason didn’t even tell users to check spam.
so i added Google sign in
then while testing the auth flow, i opened settings and realized i’d also shipped the old placeholder settings page
what was supposed to be one small fix turned into:
- Google sign in
- better verification messaging
- rebuilt account settings
- plans and billing
- credit top ups
- billing history
- profile controls
all because one person said the code was taking too long, this is probably the biggest thing i’m learning from having real users
they don’t always give you a roadmap
sometimes they just point at one annoying thing and you find 5 more things you should’ve fixed already
shipped the fixes today
generating a good UI from a sketch is only half the job.
you still need to be able to change the text, spacing, layout, components, positioning, all without starting over
that’s why editing is becoming a big part of SketchMason
3 users in 3 days.
one new user a day
is that growth? probably too early to call it that, but it is signal.
funny part is i’m still one of the heaviest users of my own product. anthropic just warned me about my credit usage because i keep generating UIs and comparing them against stuff on @awwwards
i’m trying to solve the boring details AI usually guesses the spacing, widths, grid gaps, nav rhythm. the things nobody mentions in the prompt but everyone notices when they’re wrong
- every user teaches me something
- every generation costs me something
- every day the product gets a little less sloppy.
marketing gets people through the door. product quality decides whether they stay
marc lou did over $100k in a couple days.
i’m at 3 users in 3 days, different stage of the graph
still moving right
Introduction to SketchMason
build the whole flow with SketchMason draw the UI keep every screen consistent
From dashboard to checkout, each new screen stays in the same design system.
2 free generations. no card needed
https://t.co/7ODa24aKOE
Introducing SketchMason
Draw the UI | Skip the prompt fight.
Turn a rough sketch into a finished interface, then keep the whole flow in the same design
2 free generations. no card needed
https://t.co/2WrF3EI6Zi
@sanketdattta appreciate this thanks
honestly the constraints have probably helped more than they hurt. forced me to keep the product simple and actually ship it.
turned 21 today
also shipping my first real SaaS
SketchMason turns a rough hand-drawn sketch into a finished interface.
- draw the layout.
- get the design.
- export the HTML.
built it alone, bootstrapped, no team, no funding.
you get 2 free generations, no card needed
https://t.co/2O6ToZxfI0
pretty good way to start 21 :)
this post was 2023
five hours after GPT-4 Vision shipped, @sawyerhood drew a UI, turned it into HTML and wrote:
“the future is bright.”
3 years later the future was actually bright
i’ve been building the version i wanted back then
- rough sketch into full UI flow
- desktop + mobile
- same design across every page
- HTML or Next.js export
he was right
SketchMason launches tomorrow
1 day left
tomorrow i finally open sketchmason.
draw the ui you have in your head. let mason turn it into something you can actually ship.
waitlist closes when we launch
https://t.co/2WrF3EI6Zi
3 days.
been building sketchmason around one idea:
if you can draw the ui, you shouldn’t have to explain it to an AI in 20 different ways
launching this week
3 days.
been building sketchmason around one idea:
if you can draw the ui, you shouldn’t have to explain it to an AI in 20 different ways
launching this week
i think prompting ui is backwards
if you can draw it, you should be able to build it.
- box for a card
- line for a heading
- circle for an avatar.
sketchmason turns the rough sketch into the finished interface.
opening access this week
who wants in first?
hand-drawn ui usually means the AI looks at your sketch, then does whatever it wanted anyway.
this one didn’t, every box i drew became a section on the page, in the same order
the reference image only decided the look
sketch means layout
image mean staste
rough rectangles in, finished interfaces out.
drawing the layout takes 10 seconds.
making it actually look designed is the part that eats the afternoon.
that’s what i built @usesketchmason for sketch the screen and mason designs it
one free generation a day. no card. no account.
free week is live
https://t.co/i6qr9sfZKJ
rules:
- sketch or draw a ui
- don’t make it pretty
- upload it and see what sketchmason does
pretty sketches aren’t the point
send me the ugly ones.