The trouble is that thought and culture are not the sort of thing that can have distinct units. They do not have a granular structure for the same reason that ocean currents do not have one — namely, because they are not stuffs, but patterns.
— Mary Midgley, Myths We Live By
💽🌿 I love building stuff with (and for) data and AI.
A hyper-responsive data platform → super tidy and well-documented data models for agents and humans → clear narrative analyses with just the right level of snappy interactivity and gorgeous artifacts — this stack does something to my brain that makes me very happy.
🌌🐴 As my fascination with data ranges all over — and I'm just as interested in building dev tools for the work as the work itself — I've never been able to stay cramped in a rigid box. Which one am I: an Analyst or Data Engineer (or Software Developer, Developer Experience Advocate, Teacher, Consultant, Manager...)? The answer is 'yes'.
Hence: Data Cowgirl. Ride around my trusty stallion Bracket and make sure all the data is1 thriving. Nurture it, help it grow strong. When the leaves turn, go shear it, then spend the long winter weaving it into tapestries. That kind of thing.
To be less...poetic2 about it — the way this manifests changes all the time, making labels a bit difficult. I've spent professional stretches:
- building industry-leading enterprise SaaS platforms, and lovingly crafted OSS CLIs
- constructing bomb-proof streaming data for cutting-edge factories, and tiny data warehouses pro bono for education non-profits (carefully designed to never tick over BigQuery's generous free tier)
- weaving diverse data sources into a cohesive world, and passionate human beings into cohesive communities
- designing web apps, and resilient patterns for data work
- managing analytics for teams (a lot of Marketing, Finance, DevOps, Sales), and teams of my own
- writing in-depth docs on Python SDKs, and explorations of truth in data work
- also, making the best memes in the data space 😌
🎷🐄 You get the idea — data cowgirl.
No matter the particulars, the mission is always:
- 🌱🍄🟫 Nurture data ecosystems
- ✨🕸️ Get people excited about their ability to do amazing things within it
- 🗺️🌊 Work together to understand this world a little better every day 🫶
Giddyup!
👩🏻🌾🐚 Terminals, shells, command lines, Neovim, and spending waaay too much time gardening my dotfiles. You can check them out here (although I'm very close to launching a new repo based on chezmoi). I'm really stoked about the terminal renaissance we're seeing from Claude Code! Seeing more people reach for the precision, speed, and control of the command line is so welcome after a decade of trying to wrap everything in a GUI.
🍦🦕 Writing cute and useful CLIs and TUIs in Go with the Charm libraries is a favorite of mine — although recently I've been using Deno a lot as well, and sometimes build multi-threaded Python tools with Typer when scale or portability is not a concern. I'm planning to spend more time building in Rust this year though, as it's clearly become the gold standard for modern command line tools.
- DuckDB
- Daft and DataFusion running on platforms with fantastic developer experiences like Modal as the new way to handle massive datasets
- The convergence on Parquet + Arrow lakes with flexible compute and query engines accessing them is just generally a huge improvement (speaking as somebody who has migrated a lot of data between BigQuery ⇄ Snowflake ⇄ Redshift ⇄ Databricks)
- Marimo, whose cloud platform runs on Modal
- Hex, hands-down best BI tool in the world (maybe ever? probably ever.)
- Open standards emerging around metrics definitions in these kinds of tools (as opposed to the LookML days)
- MotherDuck, the best warehouse for 90% of businesses not operating at massive scale
- Most importantly, I think what we understand broadly as analytics engineering, the analytics engineering skillset, is what's driving the best AI data systems — context engineering for analytics agents is just a new category of analytics engineering
If you like my recommendations and want more: my collecting/scrapbooking/journaling/organizing/labeling instinct is powerful, so I am a dedicated GitHub Stargazer — you're welcome to explore my lists!
If you're still here, I guess it's safe to get a little spicy. I'll leave you with some of my more heterodox beliefs:
-
Data visualization is really overrated! Don't get me wrong, I love great data visuals, and there are tons of analyses where they help communicate in a way words can't (I'm a girl with the complete works of Tufte on my shelf who gets very excited when a new issue of The Pudding drops) — but in 95% of BI work, visualization has historically been treated as the obvious standard output. In spite of half a decade lamenting the deficiencies of dashboards, they remain the final destination of all this engineering and modeling effort in most people's minds. We're past due to realize that a summary of interesting shifts in this week's revenue metrics might work better as a short paragraph. Blessedly, AI is finally getting us to the post-dashboard promised land.
-
You're looking at too much data. The reality is, your brain cannot track a dashboard of 50 metrics hour-by-hour, day-by-day in a way that is actually useful and effective. It's the data equivalent of operating as if you had no object permanence. A child figures out how a bouncy ball moves, then can track when it disappears behind the couch — we need to get there with our metrics. All the brain power you expend keeping tabs would be better spent building that intuition for the physics of your business. While it is cool to pounce on your CEO's pop quiz at the board meeting with a ratatat of punchy metrics that would make Aaron Sorkin proud — I absoluetly get that — it's theater, not impact. Save chart watching for the chaotic crises.
🃏🧚🏼 In parting, my favorite card from Brian Eno's famous Oblique Strategies deck...
∧＿∧
ｒy´・ω・｀ヽっ
`! i
ゝc＿c_,.ノ
（
）
.∧＿∧.(
(´・ω・ ∩
o ,ノ
Ｏ＿ .ノ
.(ノ