AI computing power is becoming increasingly expensive, while mobile phones are idle. Acurist wants to connect these two things. Training and reasoning large models require a significant amount of computing power, but not all AI tasks have to run on high-end GPUs in data centers. Content review, spam identification, simple decision-making, lightweight inference, scheduled tasks, API processing, often only require stable, verifiable, and cost controllable computing power. At the same time, there are billions of mobile phones worldwide, with a large number of devices mostly in low load or even idle state. These phones come with hardware level security capabilities (Trusted Execution Environment, TEE) that enable code execution in isolated environments, ensuring integrity while preventing device owners from seeing input data. @What Acurist wants to do is organize these ready-made phones into a decentralized computing network. Developers deploy tasks, mobile phones are responsible for execution, and networks are responsible for matching, scheduling, and settlement. The task runs in the TEE of the mobile phone, the execution result can be verified, and the input is kept confidential from the device owner. No need to build a new data center or rely on traditional cloud providers. Technically, it supports serverless style workloads such as REST APIs, Webhooks, scheduled tasks, confidential computing, and LLM inference. Cargo allows developers to package and deploy using containers, similar to running Linux binary files on a server. For the model, it supports GGUF format and provides a nearly OpenAI compatible interface through integrated llama services. Daily mobile phones can be equipped with Processor Lite to contribute computing power during charging or idle time; Idle phones can be converted into dedicated nodes using Processor Core. The network is a Polkadot parallel chain (mainnet), and supports payment of computing power fees in USDC on chains such as Base. According to recent official public data, the network has connected over 335000 mobile phones, covering more than 175 countries and regions, with over 1.1 billion on chain transactions and a cumulative deployment of over 685000 times. These numbers come from the real network after the main network is running, rather than the pure test network stage. The actual task is also running: At the end of September this year, Acurist deployed the open-source System One decision model Laya on mobile networks. Laya is not good at generating long texts, but rather makes structured judgments: whether emails are in the inbox or spam/phishing, whether chat content is illegal, whether prompt word injection should be blocked, and what to do next in simple games. It runs on a regular mobile phone CPU, with a typical single decision time of approximately 0.2 to 1 second. The official demo page will display which real phone is processing and provide an on chain verification entry. Developers can quickly deploy their own instances using Cargo templates, with sample costs as low as less than 1 ACU per day. This is not a conceptual demonstration, but a running instance that can be directly experienced and verified. ACU is the native token of the network. The initial supply is 1 billion pieces, with a fixed annual inflation rate of 5%, most of which flow towards community related purposes (official caliber close to 70%). Its purpose is quite clear: developers use it to pay deployment fees, network transactions require it as gas, and computing power providers pledge it to promise capacity (failure to meet it will result in proportional confiscation, with most confiscated tokens destroyed), while also receiving rewards. Holders can also participate in governance. Providers can earn directly with their mobile phones. ACU token has just been launched on the Korean exchange buthumb's Korean won trading pair. In the AI x DePIN track, Acurist's positioning is relatively clear. Most DePIN projects either require specialized hardware or only focus on resource coordination without execution verification; Traditional cloud computing is concentrated in the hands of a few vendors, with constraints on cost and data sovereignty. Acurist has brought "verifiable confidential computing" to the edge using existing mobile hardware and TEE. It is currently more suitable for medium to lightweight tasks that require low latency judgment or privacy, rather than the largest scale model training. The Cray (computing cluster) in the roadmap is designed to collaborate multiple mobile phones to run larger models, alleviate the upper limit of single machine capabilities, and is also an exploration and attempt of end-to-end AI. Progress has crossed from the testing network to the main network, with the scale of mobile phones, transaction volume, and actual deployment all accumulating. The follow-up focus is on: Can more real paid tasks continue to come in, can Cargo and cluster capabilities expand the supported workload to another level, and can ACU as a settlement and staking asset keep up with network usage. People interested in this direction can directly visit the official website to see the actual operation situation.