ViralMind

ViralMind is a decentralized AI training marketplace on Solana built around a specific class of AI model called a Large Action Model (LAM). Unlike large language models that generate text, LAMs learn to operate computers directly — moving the mouse, clicking buttons, filling forms, and navigating software interfaces the way a human would. ViralMind's goal is to train and deploy these agents at scale by sourcing the human demonstration data they need through an open, token-incentivized marketplace rather than a closed proprietary pipeline.

The Problem It Solves

Training computer-use agents requires vast libraries of recorded human–computer interactions: real demonstrations of someone completing a workflow inside an application. Collecting that data at the quality and diversity needed to produce reliable agents is expensive and logistically difficult for any single company to do alone. ViralMind turns data collection into a crowd-contributed, permissionless market where anyone can submit demonstrations, earn rewards for quality work, and businesses can commission the specific skills their AI automations need.

How It Works

The platform has three interlocking components.

Training Gym. Contributors open the ViralMind desktop client and record themselves completing digital tasks — navigating a crypto wallet, entering data into a spreadsheet, completing a game level, or operating any other software. Each recording is automatically scored by an on-platform quality agent on four dimensions: clarity (40%), accuracy (30%), usefulness for AI training (20%), and efficiency (10%). The composite score, ranging from 0 to 100, determines how much $VIRAL the contributor earns from the active Training Pool. Public Gym submissions feed an open-source dataset that benefits the broader ecosystem; private Gym submissions belong to the commissioning business.

The Forge. This is the environment where projects and businesses become "Gym owners." A Gym owner defines the skill requirements and objectives they need trained, then funds a Training Pool with $VIRAL, USDC, or their own native token. USDC contributions are automatically converted to $VIRAL. Dynamic pricing ensures contributors are compensated relative to the quality of their work and the size of the pool; any unspent funds return to the pool owner. Trained models produced from private Gyms can be licensed or monetized by the creator.

VM-1 Inference API. VM-1 is ViralMind's flagship LAM, trained on millions of human demonstration data points sourced through the Gym. It is available via API for businesses to deploy autonomous agents that can operate games, complete desktop workflows, and integrate with web applications without requiring manual scripting or pre-defined UI selectors. The team also releases smaller open-source models designed as plug-and-play replacements for traditional OCR tools in developer pipelines.

Key Features

• Quality-weighted rewards. Demonstration submissions are scored automatically, so contributors who produce cleaner, more informative recordings earn proportionally more.
• Multi-user VM support. The platform supports concurrent virtual machine sessions, enabling many contributors to work simultaneously within the same Gym environment.
• USDC on-ramp. Non-crypto businesses can fund Training Pools in USDC, which converts to $VIRAL automatically.
• One-click fine-tuning. Operators can customize models such as GPT-4o on small, task-specific datasets without requiring in-house ML engineering.
• Open-source dataset. Data collected through public Gyms is published openly, contributing to the broader computer-use agent research community.

$VIRAL Token

$VIRAL is the native utility token of the ViralMind ecosystem on Solana, deployed via a fair launch on Pump.fun with no presale and no VC or insider allocations. Total supply is fixed at one billion tokens. The team allocated 3.41% of supply to themselves, locked until 2026 via Streamflow Finance, and self-funded a $50,000 treasury position to sustain early Training Pool incentives. Contributors earn $VIRAL for quality demonstrations; businesses spend $VIRAL to fund Training Pools; and Gym owners receive a share of pool activity. Future roadmap items include DAO governance and staking. The token trades on Raydium as the VIRAL/SOL pair (contract: HW7D5MyYG4Dz2C98axfjVBeLWpsEnofrqy6ZUwqwpump).

Security and Audit Status

No completed third-party security audits are documented in public materials. The Streamflow lock on the team allocation provides on-chain verifiable evidence of the vesting schedule. The Pump.fun fair launch mechanism means there was no presale smart contract through which funds could be misappropriated before the public market open.

Team

ViralMind was founded by three co-founders who have collaborated for roughly eight years.

Dillon Dupont (CTO) graduated from MIT and previously worked as an AI engineer at Microsoft on large-scale model deployment. He created the GPT-4V-Act framework and contributed to OmniParser v2, both directly relevant to computer-use agent research.

Jaxon Heitz (CEO) studied at the University of Hawaii at Manoa, with a background in business development, finance, and growth strategy. He led tokenomics design and manages strategic partnerships.

Morgan Dean (Chief Infrastructure Officer) studied Cognitive Systems at the University of British Columbia, specializing in full-stack AI infrastructure and decentralized systems deployment.

The team's joint track record includes winning the JailbreakMe challenge on Solana — a competition with a prize pool of approximately $120,000 — by reverse-engineering AI model constraints.

Solana Ecosystem Fit

ViralMind is Solana-native at an architectural level: the $VIRAL token is issued on Solana, all Training Pool funding and contributor payouts settle on-chain, and the team's background includes prior Solana competition work. The project sits at the intersection of two of the most active themes in the 2025–2026 Solana ecosystem — AI agents and permissionless token infrastructure — occupying a specific niche (computer-use LAM training data) that is distinct from the broader AI agent chat and trading-bot space. The USDC-to-$VIRAL auto-conversion in The Forge lowers onboarding friction for enterprises that do not hold Solana tokens natively.