Codex Launches In-App Browser Multi-Viewport Testing Feature
Codex now allows developers to test applications under different viewport sizes using the in-app browser (IAB).
It can control the device toolbar, automatically click through the application at various breakpoints for validation and iteration. During long tests, it will take screenshots at critical moments and present the results to the user at the end of the session for verification.
To accelerate testing, Codex supports hiding IAB animations, increasing speed by 1-2 times; it also optimizes annotation sending speed and reduces token consumption.
Market Mechanism: Codex provides automated testing capabilities to developers as an AI coding tool, driven by event-driven AI to assist in development iteration needs, with funding directed towards AI development tool subscriptions and enterprise versions; Codex and AI programming platforms benefit, while traditional manual testing tools and inefficient development processes face pressure.
Source: Public Information
ABAB AI Insight
James Sun's team has previously developed Codex into a programming assistant with multi-round agent capabilities. This IAB integration continues the evolution of AI coding tools from code generation to full-process autonomous testing since 2025, having achieved closed-loop iteration through browser control and screenshot feedback, reducing manual verification costs for developers.
On the capital front, the team aims to enhance developer trust and iteration efficiency in AI-generated applications by concentrating computing power and model invocation resources into real device simulation testing. This is intended to improve paid conversion and retention, upgrading AI coding from an auxiliary tool to a trusted development partner.
Similar cases include the expansion of agent testing features by Cursor and Replit, as well as Devin's early project-level autonomous development attempts; current AI programming tools are at a critical stage of transitioning from code completion to end-to-end application construction and validation.
Structural Judgment: Essentially, this represents a reconstruction of the industry chain driven by technological substitution. The in-app browser combined with screenshot feedback shifts testing pricing power from manual QA and traditional automation scripts to an AI agent visual-operation closed loop. The mechanism significantly reduces testing cycles and token costs through multi-viewport and animation control, allowing AI coding tools to form a complete development flywheel, accelerating software production from labor-intensive to agent-led substitution.
ABAB News · Cognitive Law
The more automated the testing, the more developers dare to generate.
The clearer the screenshot feedback, the more valuable AI trust becomes.
One second of hidden animation is worth more than one hour of manual work.