buff.trade

buff.trade was a Solana-based platform built around a single idea: let anyone launch a tokenized AI trading agent by describing their investment thesis in plain English. No quant background required. Submit a prompt, and the system would create an autonomous agent — plus an on-chain token representing it — that traded Solana-native assets based on that strategy. Other users could buy into the token, and if the agent performed well enough to graduate, token holders would eventually earn a proportional share of the fund's returns.

The project was developed by the Homo Memetus team, founded by Danbee No. It won the OKX Solana Accelerate Hackathon, where the team integrated Solana Agent Kit with the OKX DEX API to build a virtual performance dashboard and prototype on-chain fund. In September 2025, Homo Memetus was selected as one of eleven teams in the OnePiece Labs × Solana Accelerator cohort, recognized for early traction — at that point the team reported more than ten agents had achieved over 100% PnL in two months of virtual operation.

How the Platform Worked

The core user flow had five stages. A user navigated to "Create Token," entered a token name and ticker, and described a trading strategy in natural language. The system converted that prompt into an AI agent and minted an associated SPL token on Solana. Other participants could then browse the active agents on a virtual trading dashboard, evaluate each agent's P&L, and buy its token using SOL.

Agents operated initially in a virtual capacity. If a token accumulated 85 SOL in investment, it graduated — migrating its liquidity into a Meteora pool and transitioning the agent toward managing real capital. Tokens that did not reach the graduation threshold within one month ceased virtual trading. This survival mechanic was central to the platform's positioning: token value was meant to reflect verifiable trading performance rather than hype.

Upon graduation, agents were to manage dedicated on-chain funds for three months, after which profits would be distributed proportionally to token holders based on their ownership percentage. The Anchor framework was the execution layer, providing on-chain transparency across the fund lifecycle from fundraising through maturity.

Technical Architecture

buff.trade's AI system was described as a multi-agent swarm architecture. Each agent interpreted its strategy prompt and drew on several proprietary modules: a Rug Check module that screened tokens for on-chain liquidity health and social trust signals; a Social Graph module that tracked follower growth and sentiment to evaluate token reputation; and an Onchain module providing transaction volume, holder counts, and liquidity data. Agents also used Retrieval-Augmented Generation to integrate real-time external data alongside learned models.

The Data Server aggregated on-chain metrics and off-chain signals — Twitter activity, sentiment scores, and trending token data from Dexscreener — while a Trading Server executed buy and sell orders via the Jupiter DEX API. The codebase, maintained under the memetus GitHub organization, was built in Rust and Node.js on the backend with a Next.js frontend, supporting both Solana Devnet and Mainnet via Helius RPC.

Supported trading assets included wBTC, major memecoins tracked via Dexscreener, and select Backed.fi assets. The platform noted that scalping strategies and sophisticated indicators like Bollinger Bands were not supported in the prompt interface. An unusual feature allowed agents to incorporate their creator's Twitter posts: if a user included their Twitter handle in the strategy prompt, the agent would monitor their tweets for ticker mentions and trade accordingly.

The $HOMO Token

Alongside the platform, the team launched $HOMO, described as a community meme token intended to build early engagement. The token carried no formal utility at launch; the team stated they were evaluating staking mechanisms or revenue buyback programs pending proof of product-market fit. $HOMO traded on Raydium and was listed on CoinGecko.

Roadmap and Trajectory

The published roadmap called for a launchpad prototype in September 2025, a public launch in December 2025 with on-chain fund management and simulation features, and further product updates in March 2026 covering fund rollover mechanics and creator grants.

The project entered the OnePiece Labs accelerator cohort in September 2025, which aligned with the prototype launch window. However, as of mid-2026, the buff.trade main website went offline with its Vercel deployment paused. The memetus GitHub organization was archived in June 2026. The $HOMO token, which reached an all-time high shortly after launch, had declined sharply and traded at a small fraction of that price by mid-2026.

Assessment

buff.trade pursued an original design space: applying the viral token-launch mechanics popularized by platforms like pump.fun to the problem of democratizing access to quantitative trading strategies. The idea that token graduation required demonstrated performance — not just marketing — was a meaningful departure from hype-driven launchpads. The team earned legitimate credentials through competitive hackathon placement and accelerator selection.

The project did not reach public launch stage. The combination of an archived codebase, offline website, and collapsed token price indicates it wound down before its December 2025 target. No successor project or pivot announcement has been identified.