Commoditization of Browser Automation
Asteroid
This risk says Asteroid may be building on a capability layer that foundation model labs can absorb into the model itself. If OpenAI, Anthropic, and Google make browser control a standard model feature, then basic clicking, typing, and navigation stop looking like a standalone product. In that world, value shifts away from raw computer use and toward workflow reliability, vertical playbooks, hosted scale, credentials, and human review for high stakes tasks.
-
Asteroid already sits on top of third party models and Playwright, and routes work between DOM based automation and screenshot based computer use. That means part of its product is orchestration around model capabilities, not a proprietary control model from first principles.
-
The closest comparison is Browserbase. It sells browser infrastructure to developers, while Asteroid aims at non technical operations teams with hosted browsers, live supervision, and workflow setup. If model vendors bundle browser use, both layers get squeezed, but the higher level workflow product usually has more room to survive.
-
The product wedge is repeated enterprise work, not one off browsing. OpenAI launched Operator in January 2025, later folded it into ChatGPT agent, while Anthropic and Google also pushed official computer use products. That confirms the direction of travel. Browser use is becoming a native model feature at the same time Asteroid is trying to sell it as software.
The likely end state is that browser automation platforms either move up the stack into full workflow systems for insurance, healthcare, and similar legacy sectors, or they get flattened into thin wrappers around model APIs. As foundation labs keep shipping better browser agents, standalone value will concentrate in deployment, compliance, monitoring, and vertical specific know how.