Building a terminal for meme coin traders

A cross-chain trading terminal where acting on a signal never means leaving the screen. Built from scratch, from sign-up to multi-wallet execution, for traders who have seconds to decide.

Overview

Meme coin traders deal with hundreds of new tokens an hour, with signals scattered across Telegram and X and only seconds to act. I designed Trady from scratch, from auth to signal filtering to multi-wallet execution, as the only designer on a team with 1 PM, 1 frontend and 1 backend developer, and 1 QA.

55%Faster to the relevant signal

2.5×Daily users, about 400 to 1,000

+38%

First-session retentionFrom about 15% to about 21%, after we redesigned onboarding and the initial screen.

The problem

Existing tools showed everything and filtered nothing. Watching how traders worked, the pattern was constant jumping: the token list, a detail page, Telegram, X, then back to buy. In a market that moves in seconds, each jump can cost the trade.

So we set one constraint for the whole product: 10 seconds from discovering a token to buying it.

The approach

Keep the trade in the list

The familiar pattern is a detail page for every token. It’s the easiest to build, but it adds a step exactly where time runs out. A side panel keeps the list visible, but still takes a click and shrinks an already dense table.

I put actions and data directly in the list instead: Quick Buy & Sell in every row, token data inline, wallet and chain selectors fixed in the layout. The cost is density, and one-click buying makes mistakes easier, so the buy amount is set once and shown on the button, and secondary links stay in tooltips until hover.

Amount set once, buying right in the rowBuySOL

Pair infoMCLiquidityVolumeTXNsToken infoQuick Sell & Buy

Invented tokens: the amount is set once, then buying and selling happen right in the row. No real trades.

Building it

Custom Radar

Every trader looks for something different, so Radar lets them set personal filters by chain, source, market cap and holder count, with AI-suggested presets based on past trading.

Max market cap Min holders

Invented tokens: a preset, a removed filter or a moved slider, and the list follows.

KOL Alpha

Checking who was buying a token meant leaving the list. AI now tracks KOL wallets across chains in real time and shows who is accumulating as inline avatar stacks. It became the second most-used tab after Trending.

KOLs buying now

Pair infoMCKOLsLiquidityVolume

Invented tokens and traders: stacks grow as KOLs buy in, and no tab-switching is needed to check who.

Multi-wallet settings

Power users split buys across wallets, which takes several settings that affect each other. Buy Variance, TX Delay, Auto-Cover Shortfall and Sell Deduction Order all live in a single modal, so they can be changed between trades.

Buy Variance0%25%50%75%100%Max % difference per wallet when splitting buys. TX Delay0ms250ms500ms750ms1sDelay between each wallet TX. 0ms = simultaneous. Auto-Cover ShortfallOther wallets cover what one can’t afford.

Invented wallets: a setting changes the split, then the buy runs. No real trades.

What we cut

Three ideas were dropped to protect focus. A Social Feed pulled attention from the token list at the worst moment. Notifications would have needed a volume that overwhelmed people, and a wrong alert at 3am is worse than none. A full Portfolio Page turned out to be something traders didn’t want: they wanted to act, not reflect, so we shipped a balance widget in the dropdown instead.

Impact

In the first 6 weeks after launch we tracked DAU and session recordings, and interviewed 12 active traders.

55%Faster to the relevant signal

2.5×Daily users, about 400 to 1,000. A shared result with marketing, so I don’t attribute it to design alone.

15→21%

First-session retentionAfter we redesigned onboarding and the initial screen.

Takeaways

What I’d changeRun a holdout group to separate the effects of design and marketing.

What I’d changeTest radar presets with traders before launch, not after.

Speed wasn’t a feature, it was the product.

AI was the infrastructure that made it possible.