I have another project where there are three years of decisions made on layers, two refactors that never saw the light of day, and a misc directory that I dare not open. Last month I deliberately tried Cursor and Claude Code on it for two purposes.
For deciphering the mess, the terminal agent proved superior. It told me why a background task executed twice and tracked down the problem through five files without my even looking at any of them. For minor changes in user interface, the editor was superior since inline diffs allowed me to accept one line and reject the next.
Seeing big tech companies secure whole power plants for AI compute is a reminder that running these large models takes serious resources.
It made me rethink the common assumption that model costs will infinitely drop over time.
I started adding a column in my metrics sheet to track exact AI costs per user interaction. Turns out one fancy background feature was burning through tokens without adding real value to the user experience.
Are you actively monitoring token costs per feature in your app or relying on price drops to keep your margins safe?
We used to believe every SaaS needed a free tier, but hard to use apps usually lose free signups right at onboarding. Without built-in community templates or direct guidance most people drop off within minutes.
Switching to a short paid trial forces users to have skin in the game making them more likely to walk through setup examples and discover real value.
