# I tested Abacus's new SUPERCOMPUTER... (INSANE) Source: https://www.youtube.com/watch?v=J3MHGRAqms0 Recap page: https://rapidrecap.app/video/J3MHGRAqms0 Generated: 2026-07-31T04:35:24.79+00:00 --- ## The Gist Abacus AI SuperComputer is an all in one cloud platform that provides a persistent Linux virtual machine with databases, storage, and over one hundred AI models starting at seven dollars for the first month. Wes Roth demonstrates building four production ready applications inside the environment using plain English prompts. ## Quick Overview Abacus AI SuperComputer delivers a fully functional Linux server with permanent background execution, integrated open source AI agents like Hermes and OpenClaw, and simple natural language deployment tools for ten dollars a month. Wes Roth tests the platform by building a local Qwen LLM chat web app, a 3D multiplayer neon arena game called WesGame, a 24-seven AI streaming TV network called WES TV, and a simulated automated multi agent trading lab. **Key Points:** - Abacus AI SuperComputer provides an always on Linux server running Ubuntu 24.04 LTS with two processor cores and eight gigabytes of memory. - The platform includes integrated support for databases, file storage, GitHub, and secure shell SSH access. - Users can host open source AI models and autonomous agents like Hermes and OpenClaw with single click deployment. - Wes Roth builds a local Qwen 2.5 chat application running entirely on the server without external API calls. - The WesGame project creates a 3D multiplayer arena with live player positioning and persistent scoreboard tracking. - WES TV broadcasts a twenty four seven streaming network featuring six channels generated by AI with ffmpeg loops and avatar presenters. - The trading lab simulates an automated multi agent stock portfolio with specialized research, strategy, risk, and execution roles managing one hundred thousand dollars in paper money. - Pricing starts at seven dollars for the first month and ten dollars per month thereafter. ![Screenshot at 16:15: The automated multi agent trading lab dashboard showing portfolio summary, equity curve, current stock positions, and live agent activity logs.](https://ss.rapidrecap.app/screens/J3MHGRAqms0/00-16-15.jpg) **Context:** Wes Roth tests developer workflows and AI integration platforms on his channel. In this video, he reviews the Abacus AI SuperComputer service to evaluate whether it simplifies the difficult task of hosting, maintaining, and deploying web applications and AI agents in the cloud. ## Detailed Analysis The Abacus AI SuperComputer eliminates traditional infrastructure overhead by letting users build and run persistent web apps, databases, and AI agents through natural language prompts. Wes Roth tests this capability by deploying four distinct applications directly on the cloud server. First, he deploys a local Qwen LLM chat interface that runs privately on the server hardware. Next, he builds WesGame, a 3D multiplayer arcade game featuring real time websocket synchronization and player leaderboards. He then launches WES TV, a twenty four seven streaming television network featuring six channels with custom visual layouts and AI generated video content. Finally, he creates a simulated multi agent trading lab where specialized AI agents handle stock research, strategy formulation, risk management, and order execution. Throughout the testing, all applications maintain background uptime and database persistence without requiring local machine maintenance. ### Platform Architecture and Fundamentals The Abacus AI SuperComputer functions as a real Linux machine hosted in the cloud that never turns off. - The system runs Ubuntu 24.04 LTS with two processor cores, eight gigabytes of memory, and forty eight gigabytes of disk space. - Users manage applications through a web based chat interface or via direct SSH terminal access. - The platform includes built in PostgreSQL databases, Amazon S3 compatible storage, and GitHub integration. ![Screenshot at 20:05: Direct SSH terminal connection displaying system resource usage and active Linux services.](https://ss.rapidrecap.app/screens/J3MHGRAqms0/00-20-05.jpg) ### Hosting Local LLMs with Qwen Users can host open source language models locally on their cloud instance to create private chat interfaces. - The application setup uses Ollama to pull and run Qwen 2.5 models in small and large parameter sizes. - A ChatGPT inspired web UI connects to the local model via port 3012 without relying on external API keys. - The setup script automatically configures systemd background services and Nginx reverse proxy routing. ![Screenshot at 05:19: The local Qwen LLM chat interface running entirely on the cloud server without external internet APIs.](https://ss.rapidrecap.app/screens/J3MHGRAqms0/00-05-19.jpg) ### Building WesGame Multiplayer Neon Arena The platform builds a 3D browser game complete with multiplayer controls and live score tracking. - WesGame features real time websocket communication to synchronize player movement and scoring across multiple browser tabs. - The server handles player input and maintains live and all time leaderboards in the cloud database. - All underlying server infrastructure and game loops run continuously without local hosting. ![Screenshot at 08:14: WesGame 3D neon arena showing two simultaneous players competing and updating the live leaderboard.](https://ss.rapidrecap.app/screens/J3MHGRAqms0/00-08-14.jpg) ### Launching WES TV Streaming Network WES TV demonstrates a twenty four seven streaming television service with six dedicated channels. - The network features live test patterns, AI news desks with avatar presenters, and analytical waveform displays. - FFmpeg scripts generate continuous video loops and on screen text tickers for weather and market indices. - All video streams and channel graphics run persistently as background Linux services. ![Screenshot at 12:22: WES TV channel selection dashboard displaying live preview cards and streaming categories.](https://ss.rapidrecap.app/screens/J3MHGRAqms0/00-12-22.jpg) ### Simulated Multi Agent Trading Lab The trading lab deploys a team of specialized AI agents to manage a paper stock portfolio around the clock. - Agents are divided into research, strategy, risk, and execution roles to evaluate market conditions. - The risk agent automatically vetos trade proposals that exceed the eighty five percent portfolio exposure limit. - All buy and sell orders are logged in a SQLite database and displayed on a dark mode financial dashboard.