Test an iPhone game and report what passed
Cognition had Devin on GPT-6 Astra test the game Otter Run in a simulator, returning a recording and a report of checks passed and untested areas. Task: Suggested implementation: Use Devin (Cognition's cloud agent) running on GPT-6 Astra to test an iPhone game (e.g., Otter Run) within a simulator. The agent executes test scenarios and generates two outputs: a video recording of the gameplay and a report detailing passed checks and untested areas. Engineers review these outputs to assess software behavior and identify coverage gaps.; tools: Devin. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include access to Devin, GPT-6 Astra, and a compatible iOS simulator environment. The agent's report explicitly lists untested areas, requiring human engineers to manually review these gaps. Source reliability is low as the original URL returned a 403 error, and no specific human roles or execution metrics were disclosed..
Tools
Published on Agent Work Atlas:
Overview
Cognition had Devin on GPT-6 Astra test the game Otter Run in a simulator, returning a recording and a report of checks passed and untested areas.
Workflow
Suggested implementation: Use Devin (Cognition's cloud agent) running on GPT-6 Astra to test an iPhone game (e.g., Otter Run) within a simulator. The agent executes test scenarios and generates two outputs: a video recording of the gameplay and a report detailing passed checks and untested areas. Engineers review these outputs to assess software behavior and identify coverage gaps.
Reported result
Actual outcomes are unknown as the source provides no pass counts, timings, or verification of success. The case is presented as a potential step toward reducing manual code review, but Cognition has not confirmed this outcome. A measurable acceptance check would be: 'Verify that the generated report accurately lists at least 90% of executed test cases against the simulator logs.'
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include access to Devin, GPT-6 Astra, and a compatible iOS simulator environment. The agent's report explicitly lists untested areas, requiring human engineers to manually review these gaps. Source reliability is low as the original URL returned a 403 error, and no specific human roles or execution metrics were disclosed.
Try it
Act as a QA engineer using Devin to test a mobile game. 1. **Inputs**: [Game Build URL], [Test Scenario List], [Simulator Config]. 2. **Steps**: - Configure Devin to run on GPT-6 Astra. - Instruct Devin to launch the [Game Build] in the iOS simulator. - Execute the [Test Scenario List] and record the session. - Generate a report listing passed checks and untested areas. 3. **Deliverables**: A video recording file and a Markdown report. 4. **Quality Check**: Compare the report's 'passed' items against the video timestamps. 5. **Safeguards**: Perform a dry run on a sample level first. Do not publish or spend credits until a human confirms the report's accuracy. 6. **Missing Inputs**: If the game build is unavailable, ask for a local path. If the simulator config is missing, use default iPhone 15 Pro settings.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.