The author's project involved creating a dashboard to track only "unnecessary" app usage and goal tracking, which they felt was not adequately addressed by Apple's Screen Time feature. They used Python, Streamlit, and data from Apple's knowledgeC.db to create the dashboard. Despite some shortcomings, the author found Claude to be a useful tool for creating software they wanted and plans to continue using the dashboard they created.
Key takeaways:
- The author found Claude 3.5 Sonnet and the artifacts UI to be reliable, fast, and auditable, although it still lacks in design capabilities.
- Using Claude, the author was able to create a screen time dashboard to track usage of unnecessary apps and goal tracking, a project that took about 2 hours to complete.
- Claude 3.5 was found to be a significant improvement over previous AI assistance tools the author had used, with fewer run-time failures, better long-term memory, and the ability to understand requirements rather than just code.
- Despite its strengths, Claude did have some shortcomings, such as failing to generate complete lines when the chat gets long and poor design capabilities.