Role
UX research, UX/UI, UI animation
Tools
Figma, After Effects, Rive
Context
Feature concept on the 2025 Coinbase app
Try it(02)
Send Bitcoin to a new address and see what the app does. Paste, test, track, then send the rest.
Interactive prototype, built in Rive. Tap the address field to start.
hk48 gj3w lfat p3h8 p6kh 62yz 4gy7 fkp9 6zg8 kc5s dd3x 8wh8 mesn 58g3 2fpn ycgh gugg 98r5 9svy xvrn k0cd 7epr 9y0d lg3z gvlh 5tw9 05qq pv8u 6230 kuu5 h29h 3d8a t560 m3e4 hf0s grcx qn9d 3w33 6a0m gekw jf4w phxp gywf hqtx kznu g3zk nest jgwr 2dgz p5p6 7jc0 yj2e 6p8r axme u6df wl36 h9ej c02r gkw4 e8e2 5rv0 k9d8 rh8n wyey tf37 5rly 9dc6 xmtq 9l5q ymcc kl4e urjj me35 exsx gvg3 wd88 2ddz umk2 g6ha 8jl7 lph3 0hg0 m43a fpyv 60uz vslx 0am3 nday m5pq 084c 29hk jy3d 355c 4c4f augj wpdr g9t9 mfgv 4hf8 sq7r 4gs3 ea0g rc0h gk0t fqs3 ytx8 aewq vfet 6tll dn3g e3j8 0zm8 8y8r neu4 50rv n006 az00 8lm4 9h3h 0lqq yawp q05c jsls yw86 patc e8ax y0cc uzs7 dqku kats urpm 2tz4 hge5 d986 569h c4p6 364l cj3m uyej n5xj 8h34 nn7t qvdp lsxr
03 - The problem
How might we
add a safety net that keeps users in control, without slowing down people who know what they’re doing?
in ETH lost to holder mistakes, by one on-chain count
Source 1
stolen through lookalike “address poisoning” on two chains in two years (USENIX Security study)
Source 3
people I interviewed confirm transfers outside the app
My research
What I heard(04)
I interviewed 8 crypto users, met at Crypto Mondays and in my own circles, then grouped what I heard on an affinity map.
01 - Finding
confirm outside the app: Etherscan, Polygonscan, texts, Discord, screenshots.
02 - Finding
made, or are afraid of making, an address mistake. Even experienced traders.
03 - Finding
felt anxious while waiting. No status, so they kept refreshing.
The most careful users (3 of 7) already send a small test amount first, and 4 of 7 only send to people they know. The design builds on both habits.
The job underneath it all: when I send crypto to someone new, I want to know it reached the right person before I send it all.
Who I designed for(05)
Three composites drawn from the 8 interviews: someone new sending money home, someone who already tests by hand, and a business owner who needs trusted sends to stay fast.
01 - New to crypto
Chen Yuting
25 · sends money home to her brother Wei in Taiwan
“I send it, then I keep refreshing until my brother texts me back.”
Today
Sends to Wei every month and checks with him by text, because the app shows no status.
Worry
One wrong character, and no way to tell whether it arrived.
What Safe Send does
A small test she confirms with Wei, then a tracker that ends on “Received”.
Experience - Beginner
02 - Already careful
Maya Ortiz
33 · freelancer paid in crypto, sends some on to family and causes
“I always send a little first. Then I check Etherscan anyway.”
Today
Tests new addresses by hand, then confirms on Etherscan or asks for a screenshot.
Worry
Sending to someone new, like a charity or a new contractor, with nothing in the app to confirm it.
What Safe Send does
Her test-first habit, built in. Once confirmed, the address is saved with a name.
Experience - Intermediate
03 - Runs a business
Rob Nozic
35 · online shop owner, pays suppliers in Bitcoin every week
“Don’t slow me down for people I pay every week.”
Today
Pays the same suppliers again and again. Anything over $500 to someone new makes him nervous.
Worry
Extra steps on trusted sends, and one costly mistake with a new supplier.
What Safe Send does
Trusted addresses go straight through with a blue check. Only new ones get the checkpoint.
Experience - Advanced
Feature 1 · New means unconfirmed(06)
A new address isn’t a wrong address. It’s an unconfirmed one. So instead of a red error, the app pauses with a calm amber checkpoint: send a small test, and confirm with the person, through a channel you already trust, that it arrived. Addresses you’ve sent to before get a blue check and your last transfer, so trusted sends stay fast.
Why “a channel you already trust”: in impersonation scams, a fake support agent hands you the address and will happily confirm the test too. Checking with someone you know, on a number or chat you already had, is what breaks that script.
The lookalike check (right) is a 2026 idea and hasn’t been tested.
evrq z04n zz92 t9px 47t5 4vtt hkp4 r65u p07m 4afw kjsd a446 e9l0 8yde pv75 z3u3 f4ln fx66 jfts tq4r 9xny px6s 0dlc vla5 5anq srq5 vjtf l6ua qk4s xeuf m9g6 hfhy rx2u z640 t0ha ys4c amkh 92sh 8kpk 88d9 m507 9yg6 hr9z hzx4 qp2w prxd j84v 8d7c jwy0 fk85 hnj6 ny3n um0m 59q4 rhyw f34x vuet g88f wm5s d09v hulx vwts 8jz0 gxp3 0lsn 7y8x 385j 0cq0 w2p7 zdrx ku9l zfyl v00x 8s40 6ssq dwr9 trxy 3rr7 5lkk wgd3 f0lu xt3g 5m43 u8y5 a70p 9ttk y4hw zg0v xs58 j9wq n8sp r4ss 5et9 x7xs ekdu j5w8 h06u mql3 96v9 0c5j za46 47c2 6vjj 7xls quy8 n8zg ycqk qshj 3462 w547 sphj 56tx p0hd l7qx h4gs ymhq t0sf tkw8 gnsx 3tdn r6q0 tr2n 4849 znmr nlxk kxrz 7n94 qu4a 3mvf 0uas l59l cx79 5vfe 0722 zjd7 8fan mp4d q32w y6l6 2756 79xy zrxq 0hu7 3sp4 zv0u 66s2 28x0 l5h9 7xqg rada ha3p qf90 upnp dydq fd4g 00ft vt30 zcxe x0fa pjwa 32dj 999x frf8 c74h pjm8 t5dg geuz td6z hedm 3d5p djkt 564f xqt7 znhd f49j l5e9 celc zxjl samh 4gjd j9ha 5w4l ygz5 qtv9 ejm7 k3nc 65c8 4xqz k05m l2jt 2r4n mx6m n45q umht fge0 z0tk 57h5 n9gj ah80 n8td x30r 942s zc23 yhtd hlyj u6xz e7f3 qlmv vf4r xvkj ksnj sr90 66vd pts5 6t2g us2a sdne t94s kcrz h3xy 8t6m 2km2 hp9g 73um
Dec 2025 · Address poisoning · Source 2
Interactive, built in Rive: one lookalike hides in the recent activity. Hover an address to check the middle.
Feature 2 · See it arrive(07)
I tested three tracker designs with 10 users. They chose C: it felt like their banking app, and testing showed they needed a clear “Received” step. A blue check ends it, so nobody has to text the recipient to ask.
Option A
Option B
Option C - chosen
Interactive, built in Rive: tap Before and After to compare the old status screen with tracker C.
Motion with a job(08)
Every animation answers a feeling users described. Before sending, an amber glow flags an address that only looks familiar, and the amber warning answers “am I doing this right?”: look again, it’s not an error. After sending, progress and the blue check answer “did it work?”, so nobody has to leave the app to find out.
Spot · amber address · 2026 concept
Wait, this isn’t the address you know.
Before · amber warning
Look again, it’s not an error.
While it travels · tracker C
Progress you can watch, step by step.
After · blue check
Done. No need to text the recipient to ask.
How it was made
One tracker, six tools: the same blue bar from a rough sketch to the page you’re reading.
01 · Procreate
Three rough takes on one question, where is my money right now? A segmented pill, a filling bar, a dotted line.
02 · Figma
Design C as a static frame: the bar, the C, and a step list that says where your money is.
03 · After Effects + Cyclops
Motion with a job: easing, timing and the C riding the bar. Cyclops lays each layer’s wireframe over the render: nulls, trim paths, no shortcuts.
04 · Lottie
Exported clean: the expression baked into keyframes, mattes kept, one small file that plays anywhere.
05 · Rive
Converted to Rive and wired to real taps: the same animation, now interactive. Tap Before and After.
06 · Framer
Shipped on the case study you’re reading.
2025 → 2026: what I’d refine(09)
Looking back with today’s scam data, I kept the core rule, new means unconfirmed, and tightened it: the test send became the main action, the full transfer is locked to the confirmed address, and lookalikes get flagged. These refinements haven’t been tested with users yet.
2025 · new-address sheet · tested with 10 users
2026 · test-first + lookalike · untested
Outcome & learnings(10)
What testing with 10 users showed (2025)
Users picked tracker design C and asked for a clear “Received” step, which shaped the final animation.
What I learned
Motion is copy: a shake reads as “wrong”, amber reads as “look again”. And the costly errors here are slips, not mistakes: people know the right address and still paste the wrong one. So the fix is to lock the confirmed address, not to warn louder.
What I’d measure next
How many first sends to a new address start with a confirmed test, and fewer trips to Etherscan. I’d also test the extra fee a test adds, and sends to exchanges, where there’s no person to confirm with.
Sources(11)
Next project(12)
Next case study
Amara
Read the case study →