All of this just took less than an afternoon in total, and with Antigravity browser I could ask Gemini for design changes, fixes, etc and it would go and grab the screenshot to analyze.
Sources: github.com/carlossg/tri...

And then I went down the rabbit hole and added better statistics (flights by year/month/airline), flight visualization (like Jetlovers) and countries visited (like Been), but all automated from the Tripit data.

You can try it with sample data at tripit.csanchez.org/For your own data you need to ask Tripit for an export of your data (thanks GDPR). Your data never leaves your browser.

This was already a problem with low latency dynamic changes (ie. infra updates due to user actions). We solved it by dispatching the changes to the clusters and git at the same time. This could be done with agent instructions but it would be brittle

No, but yes, it is spec driven development

If done at the end of the day, I will have the results next morning. Review and iterate on the results

Profit!

Then ask Cursor to split it, given that GitHub Copilot has a limit on how long it runs.
Ask Cursor to create a GitHub issue for each chunk and assign to Copilot, using GitHub MCP

Describe the problem at the high level with some bullet points
Ask Cursor to create a plan, iterate on it until plan looks good