https://rounak.taptappers.club/fitness - this is my personal site (recently re-designed). The Fitness page is the highlight. It's still WIP, so quite a number of things are blank.
I crunched Strava data (the Strava MCP project is incredible, and I ended up contributing to it!) and built myself this fitness hall of fame page (and also rejuvenated the remainder of the portfolio). Almost all of the stuff here is vibe coded, very happy how much I could achieve.
Now that this hosted CC is achieved, next up, I think scheduled workflows would be coming. For example, certain open-source repositories host data files scraped regularly from sources via scheduled GitHub Actions, that could be simplified.
I've been using Firefly-III for over three years now. What I'm contemplating to write is a quick util which would auto-import transactions from my emails.
This. Very high-quality writeups. Also, one should read Roots of Progress newsletters - I found the signal:noise ratio extremely high for these two newsletters.
The former is more CS x economy. The latter is more Science x history x economy.
The answer Bard gave is not even very coherent. Very similar results with GPT-4V as well. This makes me very cusrious how exactly do these models "see". Are they intelligently following the route starting from one point all along, or are they just tracing it top-to-bottom-left-to-right? Seemingly, latter is the case.
I expected that the AI would be able to understand that say taking a right turn from a straight road to another sub-road definitely involves crossing (since I specified that one is running on the left of the road). And try answering along those lines.
PS: The entire thing is built using Claude Code.