Product & Brand
Halo Alpha
A crypto research and execution product designed 0-to-1 as the only designer, with 4,000+ registered accounts by April 2025.
Overview
Before buying a meme coin, users could check the post, the caller, and the token
Halo helped traders move from a noisy crypto Twitter post to a checked caller record, token data, and a trade. It combined three product areas across Telegram and web. Alpha covered Yaps, Recon, and Tokens. Halo Chat handled research, alerts, and trade actions. Sniper handled time-sensitive swaps.
I was the only product designer. I owned the Halo identity and full product experience. The core team included the CEO, a product manager, five developers, and a marketing manager.
The build ran from mid-January through late March 2025. The wider community reached roughly 7,000 members during the beta period.
The impact
Halo connected current posts, caller history, contract risk, market data, AI analysis, and execution in one product. It moved through beta to public release in about ten weeks.
4,000+
Registered accounts
April 2025
120+
Closed-beta users
Late February to March 2025
2
Shipped surfaces
Telegram mini app and web app
Problem
Crypto Twitter made confidence easy to see and credibility hard to check
Crypto Twitter contains useful market information, but the source can be hard to judge. A caller can build a reputation from one successful token, hide failed calls, remove old posts, or promote an asset after taking a position. Follower count and confidence say little about past performance.
Halo needed to make source history visible without turning an AI verdict into an unquestioned fact. That meant combining fast market information with past-call records, clear score language, human review, and a way to challenge a result.
The product also had to close the gap between seeing a call and acting on it. Users needed to check contract risk, compare market data, see who had mentioned the token, and trade without losing the context they had just reviewed.
Research
Traders wanted the history behind the person making each call
In early 2025, I used Tomo AI Alpha as the direct comparison and Kaito as an adjacent attention product. Tomo had a social feed but lacked caller-performance history, past-call tracking, and a legitimacy score. Kaito tracked attention and creator mindshare. That left room for a product focused on trade-call accountability.
I gathered direct input from roughly 30 people through Discord and Telegram. Most of it came through written interviews and questionnaires, with a few voice discussions during community sessions.
I combined that input with the market review to define three working groups: active traders who needed to check a caller quickly, investors who wanted context before acting, and community members who needed ranks and warnings explained clearly. I used these profiles to set information order and decide how much detail each part of the interface should show.
Research also confirmed that users wanted the past attached to the person making a call. They needed previous calls, performance, promotions, and legitimacy information in one place.
What we learned
Users were not asking for another social feed. They wanted current posts tied to a visible record of past calls, promotions, and outcomes. That finding led to the split between Yaps for live activity and Recon for caller history.
"Before I act on a call, I want to see what this person promoted before and how those calls performed."
Interview Participant
Memecoin Trader
Ideation
Yaps showed the market, Recon checked the caller, and Tokens checked the asset
Separating live activity, caller history, and token evidence
A single feed could show current posts or caller history well. Trying to place token checks in the same stream would make the feed harder to read.
I split Alpha into three connected areas. Yaps handled current meme-coin activity. Recon held the record of the people making calls. Tokens showed the asset, contract, market data, and people discussing it.
Yaps used AI agents to review Twitter activity and filter spam, weak posts, and content from accounts with fraudulent histories. I organized it into Feed for the full stream, Hot Daily Calls for trending tokens, and Favorites for tokens the user wanted to track.
Recon used a short leaderboard of X accounts performing well over the previous 48 hours. Each profile showed a brief, legitimacy score, past performance, and separate tabs for general Yaps and coin-specific Calls. Users could also add or remove an influencer from Favorites.
Keeping token proof next to the call
The Tokens area brought contract risk, charts, market cap, FDV, liquidity, 24-hour volume, supply, and token age into one view. Mentions showed posts referring to the token. Clues showed the influencers who had called it, the time of each post, and the price chart around that call. Users could switch between candlesticks and a standard chart.
Adding human and legal review to the score
A label that marks someone as suspicious can affect a real person. The AI output could not be the final decision.
Halo's agents collected past calls, token promotions, accusations, disputes, and reports from people who said they lost money. Flagged and negative verdicts went through review by the CEO, internal legal team, and an external UAE law firm. The law firm's name was not shared with me. The security lead and CEO owned this process.
I designed the approved result as a numerical score, plain-language label, color tier, performance history, and warning state. I also supported a Twitter contact route so a caller could challenge a score and trigger another review.
Connecting Halo Chat to real actions
Halo Chat answered questions about markets, tokens, callers, and sentiment. Users could attach files or chart screenshots and receive analysis with a suggested action. They could also ask Chat to buy or sell a token, schedule the trade, and set alerts for bullish or bearish activity on favorited tokens.
Before a transaction ran, the user reviewed and confirmed its details. Responses could be copied, regenerated, rated, reported, or shared on social media. The interface also stated that the AI could make mistakes.
Designs
Every score showed the warning, past calls, and performance behind it
I created the Halo logo, logotype, color system, typography, icons, and product language for the compact Telegram interface and the wider web product. I applied that system to Yaps, Recon, ranks, scores, warnings, the AI assistant, community launch material, and later product work.
As Xodex expanded into DeFi and AI research, the Halo name gave the product a separate identity. The governance proposal to separate Halo from Xodex and rebrand received 556 votes.
For Alpha, I designed three connected views. Yaps covered the live feed, daily calls, and favorited tokens. Recon covered the short-term caller leaderboard, profile briefs, legitimacy scores, Yaps, Calls, and influencer favorites. Tokens covered contract risk, market data, charts, mentions, and caller clues.
Halo Chat used the same product language for research, chart analysis, sentiment, alerts, and confirmed trade actions. Sniper carried the research context into a swap flow with token selection, slippage, fees, transaction review, and confirmation.
From mid-January through late March 2025, I created the identity and product design, defined the Alpha model, shaped the scoring and review interface, designed Halo Chat and Sniper, and supported the Telegram and web builds. Yaps, Recon, Tokens, Halo Chat, and Sniper launched publicly in April.
Lessons
A public risk label needs a real review path and a clear way to challenge it
Halo went from concept to public release in about ten weeks. More than 120 people joined the closed beta, and the internal admin dashboard recorded more than 4,000 registered accounts in April 2025.
The three-part Alpha model kept each check clear. Yaps showed what was happening now. Recon showed who was making the call and how they had performed. Tokens showed the contract, market state, mentions, and price movement around each call.
Flagged and negative verdicts went live only after review by the CEO, internal legal team, and external law firm. I designed how the approved result appeared. The security lead and CEO owned the decision behind it.
Once Halo Chat could suggest and execute trades, answer quality became a product risk. Requiring users to review and confirm each transaction kept the final action visible before funds moved.
Twitter support handled the first challenge path. The project showed that public risk labels also need evidence, human review, and a recorded decision inside the product.
