Why Data Science Follows the Research Model, Not the Software Development Model

By Michael McGarrah · June 23, 202610 min read

Software development succeeds 4 out of 5 times. Data science succeeds 1 out of 5. That difference is not a quality problem — it is a fundamental difference in the nature of the work. DS follows the academic research model: hypothesis, experiment, documented failure, iterate.

Building a Draft Preview Site for Jekyll — Part 2: Refining the Design

By Michael McGarrah · June 16, 20269 min read

With the options narrowed down to a separate GitHub repo with Staticrypt, I started designing the actual implementation. That's when the interesting problems showed up — Giscus feedback, Staticrypt navigation quirks, leaky RSS feeds, and the question of whether a password even matters when the source is public.

Migrating 16 Domains from SquareSpace to Porkbun — Part 2: The Sprint Finish

By Michael McGarrah · June 10, 20269 min read

With several domains approaching renewal at SquareSpace, I pushed through the remaining transfers in a rush week. Fifteen of sixteen domains now live at Porkbun — only mcgarrah.org remains, deliberately held back until I have a proper weekend to migrate its 25+ DNS records without breaking production services.

Building a Draft Preview Site for Jekyll — Part 1: Exploring the Options

By Michael McGarrah · June 08, 20267 min read

I wanted a way to share unfinished blog posts with reviewers before publishing. Sounds simple — but GitHub Pages has no authentication, serves one site per repo, and my drafts are already in a public repository. Here's every option I considered and why most of them didn't survive.