Dappos launches xBubble, low-prompt AI for task agents

Dappos launched xBubble and made the product available today. The platform creates, tests and dispatches task-specific AI agents through two subsystems called Bubble Engine and Bubble Pilot, and aims to turn short user requests into finished work for images, video, websites, documents and scheduled tasks.

Users provide a brief goal and xBubble handles model selection, prompt structure, tool use and output testing. Bubble Engine generates and evaluates standard operating procedures (SOPs) for specific tasks. It uses AI coding agents to create solution variants, run test harnesses, combine candidate models and tools, and measure outputs against examples and quality criteria. The strongest solution paths are published as reusable SOPs.

Bubble Pilot reads a user’s request, identifies the task type, checks for a matching SOP and routes the job to the best available execution path. If no specialized SOP exists, Pilot routes the request to a general-purpose agent so the task can still be attempted.

Dappos said xBubble ships with more than 10 core capabilities and two operational modes. Bubble Computer is an end-to-end project workspace that opens a sandbox, loads specialized skills on demand and runs multistep projects from research through verification to final deliverable without the user managing intermediate steps. Bubble Personal is a local-environment mode that gives the system controlled access to a user’s files, browser and apps while heavy compute and installations run in cloud containers and only explicitly authorized actions occur on the user’s machine.

Supported tasks include voice dictation, text-to-speech, talking avatars, deep research, slide and document creation, fact checking, scheduled tasks, poster and image creation, video generation and website development. The product offers a fast mode for simple daily tasks and a work mode that uses SOPs for more stable, professional outputs.

Dappos described how repeated fallback requests steer development: when users frequently request a task that lacks a specialized SOP, Bubble Engine may generate and test a solution for that pattern and publish it for future dispatch. The company stated that as more SOPs are added, a larger share of requests will be routed to task-specific execution paths.

In a company message, Dappos wrote, “We have AI learn AI, and we have AI use AI, so users don’t have to.” The firm described xBubble as an attempt to shift the work of operating AI from users into the platform’s automatic solution-building layer.

Dappos has raised more than $20 million from investors including Polychain, Binance Labs, Sequoia China, IDG Capital and OKX Ventures. The company said it will continue to expand Bubble Engine’s ability to build SOPs for more complex tasks so the dispatch layer can route a greater share of requests to task-specific solutions.