raf (@rafcespedes) header
raf (@rafcespedes) profile picture

proof of existence designing @robinhoodapp building https://teapaw.org/t.co/WyKJbl5NA6 & more figuring it all out in public

nyc
Joined June 2012
Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)Photo by raf (@rafcespedes)
been working on how the mini logo adapts across every mini tool. turned out to be a pretty flexible system at its core, and unmistakably mini at every beat all of it is built piece by piece, in public. tools and brand side by side. still naked. somehow it keeps coming together
37
52
4
1,592
40,834
stuck at sfo during this cyberattack. if you ever wanted to tell me i'm your crush, now's the time
1
4
348
agents are about to outnumber us online, and right now they all look like nothing working on an open source identity system for them. any agent, from any lab, can claim a name and a face. the avatars move too, showing when an agent is working and when it needs you
1
4
294
sometimes you literally have to tell your agents to think harder
2
6
397
i need to practice restraint
1
3
319
got a chance to see the presentation and is one of the best things i’ve ever seen 😮‍💨
thank you @scspeier for inviting me to speak to the @perplexity_ai design team last week! the team has been a huge source of inspiration for me over the year, so it was really lovely to get the opportunity. a few slides from the talk :)
3
421
was going to wait until this felt perfect, but then i remembered the whole point of this account is building in public. and perfect doesn’t really fit that so before i take a little break, here’s mini Shadow. a different way to make drop shadows. move the sun and the shadow follows it’s the first tool in mini, a product and brand experiment rethinking the controls we’re used to in design apps. every shadow exports as css, tokens and more. take it for a spin and tell me what you think
6
2
44
1,511
i spend a lot of time giving my agents context. i already talk to them in metaphors. what i want to try next goes further than that. a kind of direction that's outside how i normally think. not sure what it looks like yet, which is kind of the point
2
1
5
1,000
i want my feed to be 30% shitposts, 60% decoding the world and the future of design, and 10% cats
4
14
456
should be 100% cats, but that's my feed on every other platform
95
the level of delusion i operate at
Honestly if they sent a honeypot after me she’d probably fall in love with me and betray China
4
381
A lot of people suck at optimism because it requires creativity. Negativity doesn't demand you challenge your assumptions, which is why it's the default. Finding the good asks more of you. You have to be creative enough to imagine and consider possibilities that aren't directly in front of you.
19
214
22
829
25,022
the first thing i delete when i make it is linkedin
1
6
452
you should get to decide how ai agents read your site so i made a skill. point your coding agent at your site and it asks how agents should read it • anything goes
• files first. agents get a clean version made for them
• no ai at all. a polite please don't crawl
15
2
1
37
2,157
the replies on this are insane. looks like the bots found my post about bots
62
explaining this one better for my designer and artist friends ai tools are already reading your website. this lets you leave them a note saying please don't. it's only a note, they don't have to listen, but the well-behaved ones do if you're fine with it, it gives them a clean version of your site so they get your work right instead of guessing. and on the way out, they can leave you a note
you should get to decide how ai agents read your site so i made a skill. point your coding agent at your site and it asks how agents should read it • anything goes
• files first. agents get a clean version made for them
• no ai at all. a polite please don't crawl
12
707
nothing on the web can force an agent to listen. it's a stated preference, not a lock. but this puts your wish where agents actually look, including robots.txt, and the well-behaved ones follow it
1
333
raf retweeted
spent far too long working on a stylized water simulation this weekend
10
5
1
157
4,015