Making Memecoin Trading Less Reckless
An AI-assisted social trading app
An AI-assisted social trading app
Client
Speculative / Self-initiated
DELIVERABLES
Competitive Research User Story & Storyboarding Userflow Design Interactive Prototype (React) AI-assisted Development Visual Design System
Client
Speculative / Self-initiated
DELIVERABLES
Competitive Research User Story & Storyboarding Userflow Design Interactive Prototype (React) AI-assisted Development Visual Design System
Year
2026
Role
Product Designer & Builder
Year
2026
Role
Product Designer & Builder
“I use FOMO app. Mostly copy trading people. I don’t really do any research myself until I end up a bag holder 😭.” — $BAGS, memecoin trader. Research conversation
$BAGS is real. His strategy is social trading: find traders who look profitable, copy their positions. He doesn’t research tokens himself. He doesn’t check holder distribution or dev wallets. He picks traders based on what the app shows him — leaderboard rankings, recent wins, follower counts — and rides their trades. Sometimes he catches a 2x. More often the trader he copied exits while he’s still holding the bag. He’s not unusual. A 2024 CHI study on copy-trading platforms found that copiers overwhelmingly follow default leaderboard rankings rather than conducting any due diligence on the traders they’re copying. Interface design, not trader quality, determines who gets copied. Meanwhile, industry data paints a bleak picture for the tokens those traders are buying: 98% of tokens launched on platforms like Pump.fun collapse within 24 hours. Survival rates drop below 8% after 60 days. Social trading apps are built for speed and FOMO. The UI rewards impulsive behavior: flashing green numbers, one-click copy buttons, zero friction between seeing a trader’s position and mirroring it. Nobody is building a social trading experience for the person who wants to copy with confidence, not on faith. Hunch is a speculative concept for an AI-assisted social trading app for memecoins. Same market, same speed, same social mechanics — but the interface works for $BAGS instead of against him. AI handles the due diligence so he can copy with confidence.
My Role
Solo designer and builder. I took this from concept to working interactive prototype in approximately 24 hours. I led the competitive research, defined the product concept and user story, designed the user flows, and built the full interactive prototype using Claude Code (AI-assisted development). This project demonstrates how I use AI to rapidly ideate and prototype without sacrificing design thinking. Every product decision was mapped out by hand before a single line of code was written
What Was Actually Breaking
Trader discovery has no intelligence
Before designing anything, I spent time inside the platforms $BAGS uses: FOMO, Phantom, Terminal, dYdX, Hyperliquid, Binance, and Bybit. Social trading apps show $BAGS a feed of trader activity. Who bought what, when, how much. But every trader looks the same. A wallet that caught one lucky 50x is displayed with the same weight as a wallet with 3 months of consistent 2x returns. The app’s leaderboard does the sorting, and as the CHI 2024 study confirmed, that leaderboard determines who gets copied and not actual trader quality. The study found that on TraderWagon, being on the first page of the leaderboard explained up to 76.7% of a portfolio’s popularity, regardless of actual trading skill. Leaders have strong incentives to game these rankings with tiny positions and inflated ROI percentages, creating moral hazard. The copiers don’t know this. The interface doesn’t tell them. On platforms like FOMO, $BAGS follows wallets and buys what they’re buying. But he doesn’t know if that wallet’s profitability is consistent or a single lucky hit. He doesn’t know their average hold time (if it’s 6 minutes, he can’t copy them fast enough). He doesn’t know if they’re bundled with the token developer.
The gap: no social trading platform tells $BAGS which traders are actually worth following based on on-chain performance patterns, not interface rankings.
Token context is absent when copying
When $BAGS sees a trader buy a token, FOMO shows him the trade. It doesn’t show him the token’s health. Is the holder concentration dangerous? Is the dev wallet still holding? Is liquidity deep enough to exit? These are the signals that determine whether a trade is safe to copy, and none of them are surfaced at the moment of decision. Platforms like Terminal show all of these metrics, but they’re designed for independent traders doing their own research. Social traders like $BAGS discover tokens through people, not through screeners. The information he needs exists on a completely different type of platform from the one he uses. The gap: no social trading app surfaces token safety data at the moment $BAGS is deciding whether to copy a trade.
Entry has no safety net
When $BAGS decides to copy, every platform gives him a blank order form. Position size? He guesses. Stop loss? He doesn’t set one. Leverage? He cranks it because bigger number equals bigger return. The interface provides no opinion about what a reasonable entry looks like for this specific token at this specific moment. The result: avoidable losses from oversized positions, missing stop losses, and inappropriate leverage — on trades he’s copying from someone who may have a completely different risk tolerance and portfolio size. The gap: no platform pre-loads safe entry parameters based on the current risk profile of a token and the trader’s own balance.
The Decisions I Made
AI-scored trader trust
The principle: make the 20-minute research process happen in 10 seconds.
This is the core of a social trading app. When $BAGS finds a trader, either through the Traders tab or through the Holders tab on a coin’s detail page, Hunch surfaces a trust scorecard instantly. - High Trust (green): Consistent profitability over 30+ days, reasonable hold times, no red flags - Medium Trust (amber): Mixed signals, proceed with caution - Low Trust (red): Short hold times, inconsistent performance, or red flags like bundle detection One trader shows a red “Bundle detected” warning. That’s Hunch telling $BAGS this wallet is likely affiliated with the token developer. On every other platform, $BAGS would never know this. The Traders tab shows a full leaderboard — but unlike existing platforms, the ranking is based on on-chain performance data, not gameable ROI metrics. 30-day consistency, win rate, average hold time, total trades. The CHI study found that ROI-based rankings incentivize leaders to take tiny, risky bets with nothing at stake. Hunch’s ranking rewards sustained, real performance. This directly addresses the core failure of social trading: instead of relying on a leaderboard that rewards gaming, Hunch scores traders on what actually predicts their value to a copier.
The AI-filtered Discover feed
Once $BAGS knows which traders to trust, he needs to see what they’re buying. The Discover feed bridges social signals and token health. Each coin row shows exactly five signals that experienced traders use to make entry decisions: - Holders + trend arrow: Holder count growing fast means real buying, not bots - Top 10 holder concentration: Above 50% means a few wallets control the price - Dev holdings: Above 5% is a red flag unless there’s a strong reason - Volume: Is there real trading activity? - Liquidity: Can you actually exit this position? The Hunch Score synthesizes safety and momentum into one visual: a ring showing a number 1-100. Green means clean and moving. Amber means moving but flagged. Red means stay away. Filter chips above the feed (Rising Fast, New, Clean, High Volume) let $BAGS narrow further. But the default view is already filtered by AI. The “✦ AI filtered” badge makes this visible. This is the information that exists on platforms like Terminal but is completely absent from social trading apps. Hunch puts it where $BAGS actually makes decisions.
AI-assisted Quick Buy with safe defaults
This is the moment $BAGS goes from looking to acting. There are two paths to it: $BAGS taps “Quick Buy” on a coin in the Discover feed. He’s taken to the Coin Detail page where the right panel shows Hunch loading “Hunch is analysing this position… Checking liquidity, holder concentration, dev wallet activity…” then parameters appear, pre-loaded. Position size: 5% of balance (not a guess, a calculated default) Leverage: 1x (Hunch recommends conservative leverage for this token) Stop loss: -25% (auto-protect enabled) Take profit: 2x target Below the parameters, a risk summary: “Holder concentration is healthy at 23%. Dev holds 1.8%, within safe range. Liquidity is strong at $340K. Momentum is rising. This is a moderate-risk entry.” $BAGS can adjust everything. But if he doesn’t, the defaults protect him. He confirms with one tap. The principle: don’t make $BAGS think about risk management. Build it into the default experience.
Perpetual trading for serious positions
Memecoins are volatile by nature. For traders who want to take directional bets with leverage, going long on a coin they believe in or shorting one they think is about to dump, perpetual contracts are the natural instrument. It’s how the market already works on platforms like Hyperliquid and dYdX. Including perps in Hunch means $BAGS doesn’t need to leave for a separate platform when he wants more sophisticated exposure. It keeps his activity, his portfolio, and his risk management in one place. The Perps view is a full trading terminal: left sidebar with real pairs (BTC-USD, ETH-USD, SOL-USD), center chart with live-updating candlesticks, and a right panel with Long/Short order entry, leverage slider, and Auto Close toggles. The layout follows the conventions traders expect while applying Hunch’s design language throughout.
Portfolio with AI health scoring
The Portfolio view goes beyond showing bags. Two AI-powered stat cards sit above the holdings table: Portfolio Health: A Hunch Score for your entire portfolio. Are you over-concentrated? Is your risk spread? Is your exposure appropriate? Market Mood: Is the memecoin market currently bullish, bearish, or choppy? Volume trends tell $BAGS the conditions before he trades.
Give $BAGS context about his own position, not just the market.
How AI Showed Up in This Process
This case study is as much about process as product.
Research: AI helped me synthesize competitive findings across six platforms into three actionable problem statements in under two hours. Product thinking: I used AI as a sounding board for product decisions — the storyboard, the userflow, the navigation architecture. It pushed back on weak ideas (like building a chat interface when a contextual panel was stronger) and I pushed back on its assumptions (like defaulting to a chart-first layout when the storyboard made clear that $BAGS browses people before he browses tokens). Development: Claude Code built the full React application from a spec document I authored. Every screen, component, and interaction was specified in that document. I directed, reviewed, and corrected every output. The code was the implementation, not the design. The key insight: AI doesn’t replace design thinking. It compresses the time between having a judgment and seeing it realized. What would normally take weeks of back-and-forth between design and engineering, I prototyped and iterated on overnight. Not by skipping the thinking, but by accelerating everything that comes after the thinking is done.
Reflection
Hunch is a speculative design project. Built in ~24 hours as a product thinking and rapid prototyping exercise.
The hardest constraint wasn’t time. It was scope. 24 hours forces brutal prioritization. I had to constantly choose between “what would make this better” and “what would make this done.” The order book didn’t make it. The skeleton loaders didn’t make it. But the three AI touchpoints — trader trust, filtered feed, safe entry — all shipped. Those three are what make Hunch a social trading product, not a UI exercise. One decision defined the whole project. When I was mapping the layout, I had two strong references pulling in different directions: a full Terminal-style data grid, or Phantom’s cleaner, more consumer-friendly layout. The answer wasn’t in the references. It was in the storyboard. $BAGS doesn’t start with a token. He starts with a person. That one insight — that social trading is about people first, coins second — set the entire architecture. This is a V1. The goal is to make $BAGS feel more confident when he copies trades. I’ll continue iterating on it — tightening the AI trust scoring, improving the connection between trader activity and token health, adding the order book, and testing with real traders to see where confidence actually breaks down.
https://hunch-v1.netlify.app/
“I use FOMO app. Mostly copy trading people. I don’t really do any research myself until I end up a bag holder 😭.” — $BAGS, memecoin trader. Research conversation
$BAGS is real. His strategy is social trading: find traders who look profitable, copy their positions. He doesn’t research tokens himself. He doesn’t check holder distribution or dev wallets. He picks traders based on what the app shows him — leaderboard rankings, recent wins, follower counts — and rides their trades. Sometimes he catches a 2x. More often the trader he copied exits while he’s still holding the bag. He’s not unusual. A 2024 CHI study on copy-trading platforms found that copiers overwhelmingly follow default leaderboard rankings rather than conducting any due diligence on the traders they’re copying. Interface design, not trader quality, determines who gets copied. Meanwhile, industry data paints a bleak picture for the tokens those traders are buying: 98% of tokens launched on platforms like Pump.fun collapse within 24 hours. Survival rates drop below 8% after 60 days. Social trading apps are built for speed and FOMO. The UI rewards impulsive behavior: flashing green numbers, one-click copy buttons, zero friction between seeing a trader’s position and mirroring it. Nobody is building a social trading experience for the person who wants to copy with confidence, not on faith. Hunch is a speculative concept for an AI-assisted social trading app for memecoins. Same market, same speed, same social mechanics — but the interface works for $BAGS instead of against him. AI handles the due diligence so he can copy with confidence.
My Role
Solo designer and builder. I took this from concept to working interactive prototype in approximately 24 hours. I led the competitive research, defined the product concept and user story, designed the user flows, and built the full interactive prototype using Claude Code (AI-assisted development). This project demonstrates how I use AI to rapidly ideate and prototype without sacrificing design thinking. Every product decision was mapped out by hand before a single line of code was written
What Was Actually Breaking
Trader discovery has no intelligence
Before designing anything, I spent time inside the platforms $BAGS uses: FOMO, Phantom, Terminal, dYdX, Hyperliquid, Binance, and Bybit. Social trading apps show $BAGS a feed of trader activity. Who bought what, when, how much. But every trader looks the same. A wallet that caught one lucky 50x is displayed with the same weight as a wallet with 3 months of consistent 2x returns. The app’s leaderboard does the sorting, and as the CHI 2024 study confirmed, that leaderboard determines who gets copied and not actual trader quality. The study found that on TraderWagon, being on the first page of the leaderboard explained up to 76.7% of a portfolio’s popularity, regardless of actual trading skill. Leaders have strong incentives to game these rankings with tiny positions and inflated ROI percentages, creating moral hazard. The copiers don’t know this. The interface doesn’t tell them. On platforms like FOMO, $BAGS follows wallets and buys what they’re buying. But he doesn’t know if that wallet’s profitability is consistent or a single lucky hit. He doesn’t know their average hold time (if it’s 6 minutes, he can’t copy them fast enough). He doesn’t know if they’re bundled with the token developer.
The gap: no social trading platform tells $BAGS which traders are actually worth following based on on-chain performance patterns, not interface rankings.
Token context is absent when copying
When $BAGS sees a trader buy a token, FOMO shows him the trade. It doesn’t show him the token’s health. Is the holder concentration dangerous? Is the dev wallet still holding? Is liquidity deep enough to exit? These are the signals that determine whether a trade is safe to copy, and none of them are surfaced at the moment of decision. Platforms like Terminal show all of these metrics, but they’re designed for independent traders doing their own research. Social traders like $BAGS discover tokens through people, not through screeners. The information he needs exists on a completely different type of platform from the one he uses. The gap: no social trading app surfaces token safety data at the moment $BAGS is deciding whether to copy a trade.
Entry has no safety net
When $BAGS decides to copy, every platform gives him a blank order form. Position size? He guesses. Stop loss? He doesn’t set one. Leverage? He cranks it because bigger number equals bigger return. The interface provides no opinion about what a reasonable entry looks like for this specific token at this specific moment. The result: avoidable losses from oversized positions, missing stop losses, and inappropriate leverage — on trades he’s copying from someone who may have a completely different risk tolerance and portfolio size. The gap: no platform pre-loads safe entry parameters based on the current risk profile of a token and the trader’s own balance.
The Decisions I Made
AI-scored trader trust
The principle: make the 20-minute research process happen in 10 seconds.
This is the core of a social trading app. When $BAGS finds a trader, either through the Traders tab or through the Holders tab on a coin’s detail page, Hunch surfaces a trust scorecard instantly. - High Trust (green): Consistent profitability over 30+ days, reasonable hold times, no red flags - Medium Trust (amber): Mixed signals, proceed with caution - Low Trust (red): Short hold times, inconsistent performance, or red flags like bundle detection One trader shows a red “Bundle detected” warning. That’s Hunch telling $BAGS this wallet is likely affiliated with the token developer. On every other platform, $BAGS would never know this. The Traders tab shows a full leaderboard — but unlike existing platforms, the ranking is based on on-chain performance data, not gameable ROI metrics. 30-day consistency, win rate, average hold time, total trades. The CHI study found that ROI-based rankings incentivize leaders to take tiny, risky bets with nothing at stake. Hunch’s ranking rewards sustained, real performance. This directly addresses the core failure of social trading: instead of relying on a leaderboard that rewards gaming, Hunch scores traders on what actually predicts their value to a copier.
The AI-filtered Discover feed
Once $BAGS knows which traders to trust, he needs to see what they’re buying. The Discover feed bridges social signals and token health. Each coin row shows exactly five signals that experienced traders use to make entry decisions: - Holders + trend arrow: Holder count growing fast means real buying, not bots - Top 10 holder concentration: Above 50% means a few wallets control the price - Dev holdings: Above 5% is a red flag unless there’s a strong reason - Volume: Is there real trading activity? - Liquidity: Can you actually exit this position? The Hunch Score synthesizes safety and momentum into one visual: a ring showing a number 1-100. Green means clean and moving. Amber means moving but flagged. Red means stay away. Filter chips above the feed (Rising Fast, New, Clean, High Volume) let $BAGS narrow further. But the default view is already filtered by AI. The “✦ AI filtered” badge makes this visible. This is the information that exists on platforms like Terminal but is completely absent from social trading apps. Hunch puts it where $BAGS actually makes decisions.
AI-assisted Quick Buy with safe defaults
This is the moment $BAGS goes from looking to acting. There are two paths to it: $BAGS taps “Quick Buy” on a coin in the Discover feed. He’s taken to the Coin Detail page where the right panel shows Hunch loading “Hunch is analysing this position… Checking liquidity, holder concentration, dev wallet activity…” then parameters appear, pre-loaded. Position size: 5% of balance (not a guess, a calculated default) Leverage: 1x (Hunch recommends conservative leverage for this token) Stop loss: -25% (auto-protect enabled) Take profit: 2x target Below the parameters, a risk summary: “Holder concentration is healthy at 23%. Dev holds 1.8%, within safe range. Liquidity is strong at $340K. Momentum is rising. This is a moderate-risk entry.” $BAGS can adjust everything. But if he doesn’t, the defaults protect him. He confirms with one tap. The principle: don’t make $BAGS think about risk management. Build it into the default experience.
Perpetual trading for serious positions
Memecoins are volatile by nature. For traders who want to take directional bets with leverage, going long on a coin they believe in or shorting one they think is about to dump, perpetual contracts are the natural instrument. It’s how the market already works on platforms like Hyperliquid and dYdX. Including perps in Hunch means $BAGS doesn’t need to leave for a separate platform when he wants more sophisticated exposure. It keeps his activity, his portfolio, and his risk management in one place. The Perps view is a full trading terminal: left sidebar with real pairs (BTC-USD, ETH-USD, SOL-USD), center chart with live-updating candlesticks, and a right panel with Long/Short order entry, leverage slider, and Auto Close toggles. The layout follows the conventions traders expect while applying Hunch’s design language throughout.
Portfolio with AI health scoring
The Portfolio view goes beyond showing bags. Two AI-powered stat cards sit above the holdings table: Portfolio Health: A Hunch Score for your entire portfolio. Are you over-concentrated? Is your risk spread? Is your exposure appropriate? Market Mood: Is the memecoin market currently bullish, bearish, or choppy? Volume trends tell $BAGS the conditions before he trades.
Give $BAGS context about his own position, not just the market.
How AI Showed Up in This Process
This case study is as much about process as product.
Research: AI helped me synthesize competitive findings across six platforms into three actionable problem statements in under two hours. Product thinking: I used AI as a sounding board for product decisions — the storyboard, the userflow, the navigation architecture. It pushed back on weak ideas (like building a chat interface when a contextual panel was stronger) and I pushed back on its assumptions (like defaulting to a chart-first layout when the storyboard made clear that $BAGS browses people before he browses tokens). Development: Claude Code built the full React application from a spec document I authored. Every screen, component, and interaction was specified in that document. I directed, reviewed, and corrected every output. The code was the implementation, not the design. The key insight: AI doesn’t replace design thinking. It compresses the time between having a judgment and seeing it realized. What would normally take weeks of back-and-forth between design and engineering, I prototyped and iterated on overnight. Not by skipping the thinking, but by accelerating everything that comes after the thinking is done.
Reflection
Hunch is a speculative design project. Built in ~24 hours as a product thinking and rapid prototyping exercise.
The hardest constraint wasn’t time. It was scope. 24 hours forces brutal prioritization. I had to constantly choose between “what would make this better” and “what would make this done.” The order book didn’t make it. The skeleton loaders didn’t make it. But the three AI touchpoints — trader trust, filtered feed, safe entry — all shipped. Those three are what make Hunch a social trading product, not a UI exercise. One decision defined the whole project. When I was mapping the layout, I had two strong references pulling in different directions: a full Terminal-style data grid, or Phantom’s cleaner, more consumer-friendly layout. The answer wasn’t in the references. It was in the storyboard. $BAGS doesn’t start with a token. He starts with a person. That one insight — that social trading is about people first, coins second — set the entire architecture. This is a V1. The goal is to make $BAGS feel more confident when he copies trades. I’ll continue iterating on it — tightening the AI trust scoring, improving the connection between trader activity and token health, adding the order book, and testing with real traders to see where confidence actually breaks down.
https://hunch-v1.netlify.app/