From Selector Brittleness to Adaptive Automation

The core shift is from automations that memorize a page to automations that understand a task. Older browser and RPA systems often depended on fixed selectors, XPath chains, and scripted branches, so a small UI change, a renamed button, or a surprise popup could break the workflow and send teams back to developers. That made tools like UiPath powerful for stable, repetitive processes, but expensive to maintain once real world software started changing underneath them.

• Playwright itself documents the problem clearly. CSS and XPath selectors are tied to DOM structure and can fail when the DOM changes. Modern Playwright practice is to use more resilient locators, which shows how central selector brittleness has been in browser automation from the start.
  2 playwright 3 playwright
• UiPath grew out of that same general model. The automation records or defines UI elements, then replays the steps. UiPath now ships selector troubleshooting and self healing features, which is useful evidence that broken selectors remain a real operational problem in production automations.
  1 sacra 4 uipath
• What Asteroid is building on top of Playwright is a different control loop. The agent first completes the task with an LLM, then turns that behavior into a reusable script for later runs. That matters because the product is aiming at insurance and healthcare workflows where forms branch, portals vary, and domain experts need to supervise without writing code.
  1 sacra 5 sacra 6 sacra

The next wave of browser automation will be judged less by whether it can click through a page once, and more by whether it can survive constant UI drift at enterprise scale. That favors platforms that combine hosted browsers, LLM driven recovery, and script reuse, while pushing legacy RPA from rigid recording toward adaptive execution.