Cursor in the Vibe Coding Olympics: Cursor Tries Its Best, But Is That Enough?

Key Concepts

• Agent Mode: An operational mode for AI coding assistants.
• Sonat 3.7: A specific model used by the AI coding assistant.
• Dolly 2, Dolly 3, Dolly 3 HD: Different versions of the Dolly AI model.
• GitHub Pages: A static site hosting service offered by GitHub.
• API Key: A code used to authenticate and authorize access to an API.
• Context Switching: The process of changing focus between different tasks or applications.
• Friction: Resistance or difficulty encountered while using a tool or process.

Corser's Olympic Run: A Detailed Analysis

Initial Setup and Prompting

The video documents an attempt to use Corser, an AI coding assistant, to generate an image gallery application. The process begins with creating a new folder specifically for the Corser project. The initial prompt instructs Corser to operate in agent mode using the Sonat 3.7 model. The prompt is "I'm interested in model Can we use sonat 37 and that's it And go in agent mode".

Code Generation and Model Selection

Corser successfully creates the initial files for the application. However, it initially makes an error by using older Dolly models (Dolly 2, Dolly 3, Dolly 3 HD) instead of the specified Sonat 3.7 model. This is despite Corser attempting to read documentation about available models. The AI assistant regenerates the code multiple times, seemingly trying to correct the model selection.

User Interaction and Correction

The user intervenes to correct the model selection, explicitly requesting the use of the newer model. The user provides an API key, but Corser doesn't seem to remember it across interactions, requiring the user to re-enter it.

Functionality and Output

After the model correction, Corser successfully generates images using the correct model. The application displays images, but the process is not without its issues.

GitHub Integration

The user then prompts Corser to prepare the project for deployment on GitHub. Corser adds a license file, a .gitignore file, and updates the README.md file. It also attempts to create a GitHub Pages configuration.

Terminal and Git Operations

Corser then attempts to use the terminal and Git to create a new repository and push the code. However, it starts pushing into the same repository used in previous tests, which is not the desired behavior. The user stops the process to prevent overwriting the existing repository. Corser commits the code, adds files, and pushes to GitHub. It also enables GitHub Pages and provides instructions for setting it up.

Performance and Cost Analysis

The video analyzes the cost associated with using Corser. Corser Pro costs $20 per month for 500 premium requests. The test run used six requests, resulting in an estimated cost of 24 cents. This is comparable to other AI coding assistants tested in the series.

Observations and Conclusion

The user notes that Corser feels more intimidating and requires more effort compared to other tools, even though the actual amount of work might be similar. Despite this, Corser had the least amount of interruptions during the test. However, it didn't perform as well as some other tools in the VIP coding Olympics, ranking fourth overall. The user concludes that Corser's performance was approximately as expected, given their biases.

Notable Quotes

• "It's interesting that it's reading documentation after it actually generated the document and now it's regenerating it. It's a little bit wasteful."
• "So far they all feel similar but there is definitely I don't know for me there is something more intimidating about how this all looks in code editors even though I work in code editors for decades."
• "Honestly, it's approximately to my expectations. Maybe it's my bias speaking but at least this is the result I got with it."

Technical Terms Explained

• Agent Mode: A mode where the AI assistant takes more initiative in solving the problem, requiring less explicit instruction.
• API Key: A unique identifier used to authenticate requests to an API, ensuring that only authorized users can access the service.
• .gitignore: A file specifying intentionally untracked files that Git should ignore.

Logical Connections

The video follows a logical progression: setup, prompting, code generation, error correction, GitHub integration, and cost analysis. Each step builds upon the previous one, demonstrating the capabilities and limitations of Corser in a real-world coding scenario.

Synthesis/Conclusion

Corser is a capable AI coding assistant, but it has some drawbacks. It can generate code and integrate with GitHub, but it may require user intervention to correct errors and ensure the desired behavior. The cost is comparable to other AI coding assistants. The user experience feels more intimidating compared to other tools.