[Digital Today reporter Jinju Hong] DAPPOS has unveiled xBubble, a low-prompt AI agent that selects and links artificial intelligence models and tools on behalf of users. It is designed to perform various tasks with short requests without writing complex prompts or learning the characteristics of each model.

Cointelegraph, a blockchain-focused outlet, reported on May 11 local time that xBubble is an AI agent platform that handles image and video generation, website creation, document writing and schedule-based automation with short natural-language inputs.

DAPPOS pointed out that existing generative AI services have relied too heavily on users’ prompt-design skills to achieve desired results. It explained that xBubble is designed so that instead of users learning AI, the AI understands users’ intent and selects appropriate models and execution paths on its own.

The company said, "To use powerful AI, users should no longer have to study how to use AI." It said, "The goal is to build a structure where AI learns AI on behalf of users, and AI uses AI on behalf of users."

xBubble’s core structure consists of two systems: Bubble Engine and Bubble Pilot. Bubble Engine generates and verifies standard operating procedures, or SOPs, by task. When a specific task request comes in, it tests multiple execution methods using combinations of an AI coding agent and various tools, compares and evaluates the outputs, and accumulates optimized task paths.

This differs from a structure that simply calls prebuilt prompt templates. xBubble is designed to generate the task logic itself and compare multiple execution paths to select the most efficient method.

Bubble Pilot analyzes a user’s short request to determine the task and sends the request through the most suitable execution path among the existing SOPs. If there is no SOP optimized for that task, a general-purpose AI agent takes over and performs it.

DAPPOS stressed that the core of low-prompt AI is "not eliminating the user’s intent, but shifting the operational burden of running AI to the system."

The product is structured around work environments rather than a single function. Bubble Computer is a workspace that processes multiple steps, such as research, writing, design and verification, as a single project. If a user presents only a goal, it calls the necessary functions and tools in a sandbox environment and automatically continues through to the final output.

Another feature, Bubble Personal, is a mode connected to a user’s local environment. It accesses personal files, browsers, apps and schedules to handle personalized tasks such as generating email-based briefings, organizing photos and collecting data. To address security concerns, system-level changes or installation tasks are handled inside a cloud container and then discarded, while the user’s device is restricted to executing only permitted actions.

xBubble runs in Fast mode, suited to quick everyday tasks, and Work mode, focused on producing more stable outputs. The company explained that it is a product aimed at users who do not want to learn how to handle AI but have clear desired results.

DAPPOS plans to further advance Bubble Engine to strengthen its ability to generate AI solutions for complex tasks. The company expects that as SOP data accumulates, reliance on general-purpose agents will decline and use of task-optimized paths will rise, improving performance and response speed.

DAPPOS, a company developing productivity tools that lower barriers to entry for AI, said it has raised more than $20 million from Polychain Capital, Binance Labs, Sequoia China, IDG Capital and OKX Ventures, among others.