ViralMind

ViralMind (viralmind.ai) was a Solana project that paid people to record themselves using a computer. It paid in its own token, VIRAL. The recordings were meant to train AI agents that click, type and use apps the way a person does. This is not the Viral Mind app for video creators, or the Italian AI firm with the same name.

As of October 2026 the project appears inactive. Its GitHub repositories were archived on July 2, 2025, the main website returns a server error, its last published roadmap stops at Q3 2025, and the VIRAL token trades only sporadically. No shutdown announcement was found, so treat this page as a record of what ViralMind built rather than a guide to a live product.

Is ViralMind still active?

The evidence points to no, though the team never announced a closure that we could find.

• Code archived. Every repository in the viralmind-ai GitHub organisation stopped receiving commits by April 2025. The main viralmind-agents and desktop repositories were archived (made read-only) on July 2, 2025. The desktop app's README still read "Work in progress - coming soon!" when it was archived.
• Website down. viralmind.ai returned HTTP 522 errors (the hosting server not responding) on repeated checks on October 17, 2026. The docs site still loads.
• Roadmap stalled. The published roadmap lists milestones from April to June and a VM-1 flagship model release in Q3, with no later entries. We found no independent report that VM-1 shipped.
• API never launched. The VM-1 inference API page is still marked "Coming soon".
• No recent posts. Our X data source returned no posts from the project's account, @viralmindai, for the past year.
• Thin token market. In October 2026 Bybit listed VIRAL's 24-hour volume at under $10, and our token data showed no trades in the 24 hours before October 17.

How the ViralMind Training Gym worked

The Training Gym was the core product. According to the docs, a contributor connected a Solana wallet, recorded a task, reviewed the replay, and uploaded it. An AI "data quality agent" then scored the recording, and a higher score meant a bigger VIRAL payout.

Each demonstration produced three files: a screen video (recording.mp4), an event log (input_log.jsonl) and a metadata file (meta.json). The event log captured mouse moves, clicks, scrolls and key presses. It also saved the "accessibility tree": a map of the buttons and fields on screen. Video plus that map is what made the data useful for training an AI, not just for watching.

Payouts were small and per task. The docs give an example of an 85% quality score earning about $0.17 in tokens, and the Training Pools page says submissions scoring below 50% earned nothing. A desktop app for Windows, macOS and Linux was in development to record tasks locally, but was archived before a documented release.

The Forge and Training Pools

The Forge let businesses and Web3 projects commission training data for their own tasks. A "Gym" creator funded a Training Pool in VIRAL, USDC or their own token, and contributors recorded the requested tasks for per-task rewards. Pools in USDC were set to auto-buy VIRAL.

The docs list a 20% platform fee on deposits, plus a setup fee of $250 for enterprises or about $100 in SOL for Web3 projects. Gym owners could raise or lower the price per task to attract contributors, and a reputation system steered better-paying tasks to higher-scoring contributors.

VM-1 and the planned inference API

VM-1 was ViralMind's own AI model, to be trained on the community's recordings. The team planned an API for it. You would send a goal and a screenshot, and get back a list of steps to take. Each call was to be paid in VIRAL. The API was never marked as live. The roadmap also promised leaderboards, an agent benchmark and a marketplace for custom-trained agents. There is no evidence these shipped beyond early versions.

Before the Training Gym, ViralMind ran tournaments in which users tried to get AI agents to meet set goals for prize pools held in Solana smart contracts, according to the viralmind-agents README.

VIRAL token

VIRAL is a Solana SPL token with mint HW7D5MyYG4Dz2C98axfjVBeLWpsEnofrqy6ZUwqwpump, which the project's own repository lists as its token address. Live price and liquidity are on our VIRAL token page.

• Launch: a fair launch on Pump.fun, with no presale or venture allocation, per the docs.
• Supply: a maximum of 1 billion tokens. A January 2025 ChainCatcher overview put circulating supply at about 966 million and holders at around 3,000.
• Team tokens: 3.41% of supply was locked on Streamflow "until 2026", and the docs say the team bought a $50,000 position.
• Uses: rewards for demonstrations, funding Training Pools, paying for the planned API, and staking into tournaments. ChainCatcher reported a 5-10% protocol fee on tournament stakes. Governance staking was planned but not documented as launched.

Who was ViralMind for?

ViralMind was built for two groups. The first was people happy to record simple computer tasks for small token payments. The second was teams that wanted training data for their own AI agents, without building a way to collect it.

Today it suits neither. The code is archived and the site is down. There is no working way to earn from new recordings or to fund a Training Pool. This page is mainly for VIRAL holders checking on the project, and for anyone studying early Solana projects that paid people for AI training data.

ViralMind risks

• Likely abandonment. The main risk is that the project has stopped. No team statement explains what happened to the product, the Training Pools or any unspent pool funds.
• Token liquidity. VIRAL trades thinly, so even small orders can move the price sharply, and an exit may not be possible at a quoted price.
• Unfinished product. The VM-1 API was never live and the desktop app was archived as a work in progress, so the token's main planned uses never fully arrived.
• No audit found. We found no published security audit of ViralMind's smart contracts. The docs say tournament prize pools were held in Solana smart contracts, but none were listed in our program data, so we could not check who controls them.
• Name confusion. Several unrelated products use the name Viral Mind, and other tokens use the VIRAL ticker. Check the mint address above before trading.

ViralMind team

The team page lists three co-founders, who it says have worked together for more than eight years:

• Dillon Dupont, CTO: studied at MIT and worked as an AI engineer at Microsoft.
• Jaxon Heitz, CEO: studied at the University of Hawaii at Manoa and led tokenomics, fundraising and business development.
• Morgan Dean, Chief Infrastructure Officer: holds a Cognitive Systems degree from the University of British Columbia and led training and deployment infrastructure.