HyperCluster's blockchain (chain 838838) runs on QBFT. Today it has four validators in four cities, all run by HyperCluster: they take turns proposing each block and confirm it together. Nobody races to solve puzzles, so no power is spent on mining.

Machines earn HYPEC only for work members asked for that passes our checks: a photo looked at, a training job run, a file kept. Being switched on earns nothing, so there is no reward for burning power.

Members share the spare capacity of PCs they already own. The software can pause while you use your computer, and you can set the hours it works.

Pictures that apps send to HyperAI to be read are looked at on small AI chips (NPUs) built for that job, like the one in a Quipu machine, not on large graphics cards. Members' own AI models run on graphics cards for now. A Quipu machine is a small single-board computer, not a server.

HyperCluster 的区块链（838838 链）采用 QBFT 共识。目前有四个验证者，分布在四座城市，全部由 HyperCluster 运行：它们轮流提出区块，并共同确认每个区块。没有人需要竞相解题，因此不会在挖矿上耗费电力。

设备只有完成会员要求、并通过我们核查的工作，才能赚取 HYPEC：例如查看一张照片、运行一次训练任务、保存一个文件。仅保持开机不产生收益，因此耗电本身得不到奖励。

会员分享的是他们已经拥有的电脑的闲置算力。你使用电脑时软件可以暂停，你也可以设定它工作的时段。

应用发给 HyperAI 读取的图片，由专为这项工作设计的小型 AI 芯片（NPU）查看，例如 Quipu 设备中的芯片，而不是大型显卡。会员自己的 AI 模型目前在显卡上运行。Quipu 设备是一台小型单板电脑，而不是服务器。

We do not yet know how much energy the network, our servers and members' machines use, so we make no claim about how it compares with other networks.

Members' machines and our servers run on whatever electricity their location supplies. We do not offset emissions.

To keep files safe, HyperCloud is built to store each file more than once: on our main server and on machines in the storage network.

我们目前还不知道网络、我们的服务器和会员设备用了多少能源，因此不就其与其他网络的比较作任何说法。

会员的设备和我们的服务器使用的是所在地提供的电力。我们没有进行碳抵消。

为了保证文件安全，HyperCloud 的设计是把每个文件保存不止一份：在我们的主服务器上，以及在存储网络中的设备上。

Keep using a way of agreeing on blocks that does not depend on mining.

Estimate the energy the network uses, and publish the estimate and how we made it, at least once a year.

Keep rewards tied to work people asked for, never to machines simply being on.

Prefer hardware people already have, design Quipu machines for a long life, and keep them working with software updates for as long as we reasonably can.

Encourage members who share machines to use renewable electricity where they can.

Look for ways to use less energy, and explain what we learn in plain words.

继续采用不依赖挖矿的区块共识方式。

估算网络所用的能源，并至少每年一次公布估算结果及其方法。

让奖励始终与人们要求的工作挂钩，绝不因设备仅仅开机而给予奖励。

优先使用人们已有的硬件，把 Quipu 设备设计得经久耐用，并在合理可行的期限内持续提供软件更新。

鼓励分享设备的会员尽可能使用可再生电力。

寻找减少能源消耗的方法，并用通俗的话说明我们的发现。