Tools I rely on, May 2026

Tool lists age fast, so I dated this one. It's what I run in May 2026 as a geospatial data engineer working across GIS, AWS pipelines, and a steady amount of writing. I built the list on one rule: a tool earns a slot only if removing it would slow me down.

Note-taking — Obsidian + Bases

My notes are plain Markdown files on disk, which is why I trust them. Obsidian is the editor on top; if it shut down tomorrow the files would still open in any text editor.

Bases changed how I use it. It queries a folder of notes as a table, so I filter and sort my career history, project log, and a running "decisions" file instead of scrolling them. I write the entries in prose and read them back as a small database.

AI pairing — Claude Code

Claude Code lives in the terminal with access to my files and shell, so it works on the real repository instead of a pasted snippet. The skill and MCP layer compounds: I write an instruction once ("draft a project page from this achievement note") and reuse it instead of re-explaining the task every time.

Context is the limit. Long sessions drift, so I keep each one to a single task and commit before switching. A bad skill produces bad output on repeat, so I review skills the way I review code.

Personal site — Astro + Tailwind 4

This site runs on Astro. It ships static HTML by default and hydrates only the interactive parts. The home page is mostly server-rendered markup with two small islands; every other page ships zero JavaScript. Tailwind 4 handles styling without a separate config file. For a portfolio that needs to load fast, stay crawlable, and host for free, that combination wins.

Daily Python on Windows

Most of my data and geospatial work runs on Python: pandas, GeoPandas, scikit-learn, the AWS SDK. I run it on Windows, which forces one discipline. The interpreter is python, not python3. Most cross-platform tutorials assume the Unix name, so it stays a recurring papercut.

For environments I stay conservative: a virtual environment per project, requirements pinned, no global installs. A script I wrote six months ago should still run today.

What I dropped this year

• Heavy all-in-one IDEs for quick scripts. For a one-file utility, a terminal and an editor beat opening a full project.
• Manual note cross-linking as a system. Bases queries replaced most of the by-hand linking I used to do.
• Bookmarking tools. Anything worth keeping becomes a note with a sentence on why it matters. If it's not worth a sentence, it wasn't worth saving.

None of this is a recommendation to adopt my stack. Read it as a prompt to ask which of your own tools you would miss if they vanished tomorrow.