vaderai.ai

Vader, often searched as "VaderAI by Virtuals", is a data project for robots. It pays people in its VADER token to make the data robots learn from. You can film yourself doing everyday tasks (EgoPlay). Or you can steer a fake humanoid robot in your web browser (SimPlay). Robot builders, which Vader calls partners, pay VADER to use that data. They also get a list of open models the data can train. The token lives on Base, and a wrapped copy trades on Solana. This is not the Darth Vader voice in Fortnite, which shows up in the same searches.

VaderAI by Virtuals: from AI agent token to physical AI data

VADER launched in late 2024 as an AI agent token on Virtuals VIRTUAL$0.505-4.6%, the agent launchpad on Base. Back then it was a different product. Gate Learn's January 2025 guide called it a social media "proxy KOL" and an investment DAO manager. IQ.wiki describes staking that earned Virtual Genesis Points.

The token address has not changed. Base trading pages still list it as "VaderAI by Virtuals". The current Vader litepaper names the same address as $VADER. So the old agent token is the new data project's token.

The turn to robot data began in September 2025. That month EgoPlay entered closed beta as a game: users did daily tasks with smart glasses for Vader Points. Today's litepaper drops the investment DAO, the staking and the voting. Guides that tell you to stake VADER are out of date.

How the Vader protocol works

The litepaper splits the protocol into two layers joined by one token:

• Data layer. Contributors produce training data through EgoPlay and SimPlay. Each submission that passes validation earns VADER.
• Model infrastructure layer. This is a reference catalog that matches open physical AI models to the data the protocol collects. Partners pay VADER to access both the datasets and the catalog.

The bet is a loop. Partner payments fund contributor rewards, rewards bring in more contributors, and more data draws more partners. The weak point is the partner side. As of October 2026, neither the site nor the litepaper names a paying partner. That means rewards are not yet shown to come from outside demand.

EgoPlay: earn VADER for first-person task videos

EgoPlay is a task app. You film set daily tasks from your own eyes, then upload the clips to be checked. The idea is simple. A robot sees the world from its own point of view, so video shot the same way is useful for training it.

The Vader homepage shows what one sample holds: the video at 30fps, a hand-tracking file, labels for each clip and a scene map. You are paid per clip that passes the check. The litepaper warns that work you send "may not be accepted, rewarded, or used".

SimPlay: operate a Unitree G1 humanoid in your browser

SimPlay puts a Unitree G1 humanoid robot in your browser. It runs inside MuJoCo, a physics engine. You guide the robot through tasks, and each run is saved as motion data. The homepage lists 250Hz physics, with friction, weight and light changed between runs. In October 2026 it showed more than 50,000 humanoid runs recorded.

Only one scene, an office, was live in October 2026. Home, factory and hospital scenes are marked "coming soon". So is a separate robot arm game for bench tasks.

SimPlay costs money. Each run uses computer time, so you buy tries in VADER. You earn VADER back only for runs that pass. A rejected run is a loss.

Vader model infrastructure: WAM, VLA and task-specific policies

The model layer is a list, not a training service. The litepaper says it "does not run training itself". It sorts models into three groups, with named examples for each on the homepage:

• World Action Models (WAM): generalist models of how actions play out in physical space. Examples include LingBot-VA 14B, Fast-WAM, WorldVLA and DreamZero 14B.
• Vision Language Action models (VLA): policies that turn camera input and spoken or written instructions into robot actions. Examples include OpenVLA 7B, OpenVLA-OFT, Physical Intelligence's pi0 and RoboVLMs.
• Task-specific policies (TSP): small models for narrow jobs, trained with methods such as ACT, Diffusion Policy, 3D Diffusion Policy and RL fine-tuning.

The homepage shows moving demo panels for uploads, training and "deployed" models. Treat them as pictures of the idea. The litepaper is clear that Vader does not train models for partners.

VADER token: contract address, Base and Solana

The litepaper gives VADER four uses. It pays rewards, buys SimPlay tries, and buys partner access to the data and the model list. There are "no complex derivative mechanisms, governance layers, or staking reward structures", it says. Max supply is 1 billion, per IQ.wiki.

• Base (main chain): 0x731814e491571A2e9eE3c5b1F7f3b962eE8f4870, as named in the litepaper.
• Solana: a copy wrapped by Wormhole W$0.011-4.4%, VADER, listed as "VaderAI by Virtuals (Wormhole)". Our token data showed almost no liquidity for it in October 2026. Most trading is on Base, so check the token page before you buy on Solana.

For more on how Virtuals agent tokens reach Solana, see our report on Virtuals opening AI agent ownership to Solana users.

Who Vader is for

• Contributors with time and a camera. EgoPlay suits people willing to film routine tasks for token rewards that are uncertain. SimPlay suits people who will pay VADER up front to try for approved runs. Neither works as a dependable income, because the litepaper lets Vader change or stop rewards at any time.
• Robotics teams looking for egocentric or humanoid simulation data. The datasets and model catalog are aimed at them. Teams should check data volume, licensing and quality directly, since no customers or dataset terms are public.
• Solana users. The protocol and its main token market are on Base. Solana is a bridged sideline, not a native home.

Is VaderAI safe? Risks before you contribute or buy

• No audits cited. Neither the litepaper nor the site names an independent smart contract audit. The litepaper warns that contracts "may contain errors, be subject to exploits, or change significantly".
• Discretionary rewards. Vader "reserves the right to withhold, modify, or discontinue rewards", and that includes rewards for SimPlay runs you have already paid for.
• Unproven demand. No paying partner is named. Until one is, the flywheel depends on token rewards rather than outside revenue.
• Bridge and liquidity risk on Solana. The Solana token is a Wormhole wrapper, so it relies on that bridge and trades thinly.
• Shifting identity. The token has been an agent token, an investment DAO and now a data protocol. Older guides, staking pages and airdrop write-ups describe features that are gone.
• Thin public communication. Between late September and late October 2026, the @VaderResearch X account posted robotics industry commentary but no product updates.

Team and company behind Vader

The litepaper names Vader Fun Corporation as the company behind it. We found no team names in its own pages. The @VaderResearch account posts under the name "mete". In September 2026 it posted about trips to Chinese robot makers such as AgiBot, Ubtech and EngineAI.