AI-Assisted Coding Experiences
AI coding tools drastically reduced development time from a month to a day, but tool quality differences were highlighted. - One developer went from building skeleton/base classes first to starting with prompts; AI handled most implementation. - The feeling was 'one person is a team'. - Another user tried to hide Twitter's blue bar using luna terra deepseek pro, but three approaches all failed. - Switching to 'sol' and clearing previous code allowed the feature to be completed in one attempt. - The conclusion: 'you get what you pay for,' suggesting tool quality correlates with cost.
Key insights
AI Coding Tools Compress Month-Long Projects Into a Single Day
A developer reported that a project which previously required a month of work, with extensive skeleton and base class setup, now can be completed in one day by starting with prompts. The AI handles most implementation after tests pass, leading to the sensation that 'one person is a team'.
AI Coding Tool Quality Varies Significantly, Cost May Correlate with Effectiveness
An attempt to hide Twitter's blue notification bar using luna terra deepseek pro failed after three different methods, leading the user to call the AI 'too dumb.' After switching to 'sol' and clearing the previous code, the same feature was successfully built in one attempt, leading to the conclusion that 'you get what you pay for.' This shows that different AI tools have different capabilities, and higher-cost tools can offer better results for specific tasks.