Overview

Introduction

Hōkū is the first truly personalized trading intelligence platform for on-chain markets, built on address-level machine learning that understands trader behavior, not just token prices.

Just as TikTok learns your preferences to surface videos you’ll actually watch, Hōkū learns from your on-chain activity to highlight the tokens most relevant to you. It interprets billions of wallet interactions across networks to model how traders behave, cluster similar strategies, and predict which assets are likely to attract attention next.

Instead of generic market data or static watchlists, Hōkū delivers a living, adaptive experience — where every chart, trend, and recommendation is tuned to your unique trading fingerprint. This level of personalization has never existed in crypto before, and it marks a shift from data exploration to data understanding at the individual trader level.

Core functionality and key features

Even in this early version, Hōkū offers crypto traders a series of ground-breaking features to provide a trading edge that no other crypto dapp can offer.

Immersive 3D network exploration with Wallet Cloud

Hōkū’s Wallet Cloud visualizes the entire on-chain trading network as a living, three-dimensional map. Each point represents a wallet, and its position in space reflects how that trader behaves — what tokens they buy, how frequently they trade, and how much risk they tend to take.

Behind the scenes, millions of wallets are analyzed using Deep3 Labs’ proprietary clustering models to detect behavioral similarities invisible in standard blockchain explorers. As you zoom in, you can move from a global market view down to individual addresses, seeing how traders self-organize into distinct ecosystems and discovering which clusters are driving current market trends.

This feature turns raw blockchain data into a map of human behavior, making it possible to explore crypto markets the way you’d explore a social graph.

Personalized token relevance powered by deep learning

At the heart of Hōkū’s personalization is its deep learning relevance model, built using the same class of algorithms that power TikTok’s content recommendations. Instead of videos, Hōkū learns which tokens you’re most likely to care about next based on your wallet’s on-chain behavior and the behavior of millions of others.

It analyzes patterns in what you’ve traded, how long you’ve held, your risk tolerance, and which clusters of traders you resemble. From this, Hōkū assigns relevance scores to tokens you don’t currently hold, ranking them by how well they fit your profile. It's likely the most complex model every trained on blockchain network data.

The result is a truly personalized discovery experience, one that doesn’t just surface popular tokens, but highlights the ones that are most relevant to you.

Smarter trade signal alerts

Most trading alerts focus on price movements or technical indicators but Hōkū’s Trade Signal Alerts are fundamentally different. They predict how market participants themselves are likely to behave next.

Using thousands of behavioral and network features, Hōkū’s models continuously monitor which tokens are trending across clusters of similar traders. When a token begins to attract attention from multiple high-performing groups, our machine learning system forecasts the probability of further buying activity , effectively predicting the flow of capital before it appears on price charts.

This approach transforms traditional signal generation from reactive to predictive. By modeling the psychology and coordination of real traders rather than aggregate price data, Hōkū identifies early moments of collective intent, allowing users to see where the market is about to move, not just where it’s been.

Highly accurate short-term price predictions

Hōkū’s short-term price prediction engine is one of the most accurate on-chain forecasting systems ever deployed, achieving over 90% historical accuracy across millions of backtested predictions.

Instead of relying on price charts or momentum indicators, our models are trained on address-level behavioral data. By learning how specific trader archetypes typically behave before price changes occur, Hōkū can anticipate short-term movements with remarkable precision.

Each time a potential trend is identified, a series of 36 machine learning models per network are run to predict the likelihood of various price increases or decreases over a 1-hour, 4-hour and 24-hour time period. These predictions are continuously updated and evolve with each new block, adapting to changing market dynamics and trader sentiment in real time.

Natural language network search with DeepLeap

DeepLeap lets users interact with blockchain data the same way they would with an AI assistant — through natural language. Instead of writing queries or manually browsing endless wallet lists, you can simply describe what you’re looking for:

“Find wallets that bought AI tokens early and sold for a profit,” or “Show traders who specialize in meme coins.”

DeepLeap parses these plain-English prompts and searches across millions of on-chain behavior profiles to return precise results in seconds. Powered by Deep3 Labs’ proprietary language-to-data mapping models, it bridges human intuition with machine-scale blockchain analytics, turning complex behavioral datasets into something instantly accessible and actionable for every user.

Cluster-level real-time trends

Hōkū’s Cluster Trends system shows what every type of trader group across the network is doing in real time. Each cluster represents a community of wallets that share similar behaviors, risk preferences, and trading styles.

Each cluster gathers similar groups of traders and Hōkū continuously tracks what’s trending within each group in real time. Because this insight is based on live wallet behavior, not aggregated market data, it reveals momentum shifts hours before they appear in traditional “trending token” tools.

By monitoring early activity from influential clusters, users can spot new narratives and tokens before they go mainstream, acting on trader behavior rather than waiting for price movement.

Underlying technology

Hōkū's platform is built entirely on real-time AIML inference in order to provide groundbreaking trading insights. By integrating high-speed data feeds, advanced algorithms, and comprehensive trading tools, Hōkū offers traders an edge you can't find anywhere else.

Real-time chain data

Unlike most analytics platforms that rely on third-party APIs or delayed indexers, Hōkū runs on Deep3 Labs’ own blockchain nodes. This allows it to stream new transactions directly from the network in real time — faster than traditional explorers can display raw block data. Every trade, wallet interaction, and liquidity movement is captured, processed, and analyzed within seconds.

Address-level AI/ML

At the core of every Hōkū network integration are nearly 50 custom machine learning models, trained on billions of wallet-level events. These models don’t just analyze token prices, they learn the behavior of the traders themselves. Each wallet is treated as a unique data point with its own history, risk profile, and strategy. From this, Hōkū predicts what those wallets are likely to do next, forming the foundation for features like personalized token recommendations and real-time behavioral trends.

Use and impact

Hōkū transforms the overwhelming world of crypto trading into something intuitive and personal, the same way Netflix helps you find your next movie. Instead of scrolling through endless token lists or reacting to trends too late, Hōkū’s AI understands your on-chain behavior and curates tokens that align with your unique trading style and risk profile.

For new users, this means the barrier to entry in crypto is dramatically lowered. You no longer need deep market knowledge or constant monitoring to find promising opportunities, Hōkū guides you there automatically.

For experienced traders, it provides a powerful edge: the ability to see what’s trending among similar wallets, spot early moves before the broader market reacts, and adapt strategies in real time.

Ultimately, Hōkū just makes discovering, evaluating, and acting on crypto opportunities easier.

Roadmap

Hōkū was built not just for today’s traders, but for the next generation of autonomous trading agents. Every component, from our real-time data pipeline to our wallet-level machine learning models, was designed to power agentic decision-making.

Today’s trading bots rely on lagged data: price feeds, sentiment indexes, and social media trends. Hōkū’s infrastructure delivers something far more powerful — a live, high-resolution stream of on-chain behavior down to individual trades and traders. When connected to autonomous systems, this becomes the ultimate competitive advantage: agents that see the market forming as it happens, not after.

Beyond trading, Hōkū’s personalization framework applies to any category of smart contract. Tokens are just the beginning. The same architecture can recommend on-chain games, social experiences, NFTs, or real-world asset opportunities, all with the same adaptive intelligence that understands user behavior and intent.

Over time, Hōkū will evolve into the discovery engine of Web3 — the interface where both humans and autonomous agents navigate an ever-expanding on-chain world.

Summary and next steps

Personalization transformed how people interact with the internet, powering the rise of platforms like TikTok, Netflix, and Spotify. Hōkū brings that same breakthrough to Web3, using real-time blockchain data and address-level AI to create the first truly personalized trading experience.

As we continue to expand, the next step is clear: bring Hōkū to every major network. The more ecosystems it connects, the smarter and more comprehensive its recommendations become, accelerating on-chain discovery for millions of users and setting the stage for a personalized, intelligent Web3 economy.