Using Claude Code to Visualize All My Strava Runs from 2025
I ran my first marathon this past year. Over the holidays, I used Claude Code to visualize all my Strava running activities.
my humble digital abode for armchair thoughts and toy projects
I ran my first marathon this past year. Over the holidays, I used Claude Code to visualize all my Strava running activities.
Notion has great UX for manipulating tabular data but getting data formatted into a Notion Database is time consuming. It turns out, using Cursor for writing throwaway data manipulation scripts is a great solution for this.
Natural language is lossy—code isn't. By framing LLM tasks as code completion problems instead of chat prompts, we can preserve far more meaning and unlock better performance.
OpenAI has a surprisingly simple Postgres infrastructure that scales to millions of QPS.
We don’t teach AI — we evolve it. “Training” suggests we’re walking an AI model through a syllabus, but machine learning isn’t education. It's selective pressure. Every choice we make when designing a learning algorithm — the data, the architecture, the loss function — shapes an artificial environment our models adapt to.
A mental model for using AI Coding tools like Cursor—rooted in CSPs and backtracking—to explain why “Vibe Coding” often fails at scale and how engineers can use these tools more effectively by applying core software design principles.
Great Filters and Street Fighting Software Engineering