MacCopilot

🤖Copilot AI app is now live – Explore better ways to interact with large language models 🤔Our thinking LLMs are growing increasingly powerful, and many multimodal models are also maturing, but there is still plenty of room to improve interaction design. Current assistant apps (such as Copilot AI) follow this interaction pattern: when the model gains image understanding capabilities, an image upload button is added; when it can process images and understand voice, a voice input button is added.

1. A better approach may be to infer user intent right at the source of data generation. For example, when a user takes a screenshot, they already have the intent of "needing to understand the image content". Right after the screenshot action is completed, the "understanding" process can be triggered immediately, shortening the path for users to get answers.

2. Additionally, we often encounter lots of small, temporary information lookup needs during work, daily life and creative work, but don't want to get distracted doing detailed searches, and hope to get answers right away to resume our tasks. Our solution is: parse any desktop content + multimodal models.

That's why we developed this app: MacCopilot.

✨Features

• Seamless integration: Press a shortcut to take a screenshot → select the target area → the question input box pops up → get your answer
• Supports OpenAI GPT-4o, Google Gemini, and Claude AI

🔧Use cases

• Leverage powerful multimodal models for more robust, system-wide OCR functionality
• Paper reading assistant for quick lookups of complex concepts
• WeChat reply assistant: take a screenshot to generate replies in your desired tone
• Email reply assistant
• Assistant for all types of language and application websites (e.g. overseas LLC registration)
• Assignment and exam answer assistant