Christopher Seaman

Junior analysts: how will I know which numbers are scary number Mid level analysts: why is every single number scary Senior analysts: why is every scary number actually a people and process problem

One day you’ll find someone obsessed with you. It’s probably going to be a squirrel.

@SPAMaadness Same! There was a period of really weird ads early on in my blocking journey. Now they have settled into a pervasive and boring flavor of irrelevance.

When a mathematician visits my office and sees the box of Hagoromo chalk (Seriously get yourself some of this chalk hagoromo.shop)

@PhDemetri Voice memo on slack with a couple scribbled on screenshots

Have y’all ever heard of loom?

@drooliet If you’ve got an iPad, Divinity: Original Sin 2 is a AAA RPG and the translation from PC to touch controls works really well

@PrintInkling Freesia looks and smells amazing

Masculine leaning friends SPEAK TO ME OF YOUR FAVORITE FLOWERS

Sometimes working with big data requires "the old ways" - making ethernet cables because wifi data transfer is taking longer than a reasonable coffee break

welcome to being a data director where the job is made up and the points don't matter

Well into day two of python conflict hell. Some nasty interaction between brew (daily driver) and conda (teaching). Thankfully, everything still runs on @HelloPaperspace. Only cloud-native from here forward. Sigh... the glamorous life of a data scientist

@PhDemetri In a previous life, we had a long-term holdout group on a version of the site from maybe 1.5 years back. Do it.

Imagine the chaos when we show no effect in a very highly powered and simple experiment. I think we'd probably bury the result

I so very badly want to send like 5% of traffic to the version of our website that was live Jan 1 and compare against our "final" version this year

Evidently, the squirrel population of SF is limited to Fisherman's Wharf and the Marina. Thinking about adopting a pet squirrel in the mission. Tweet courtesy of [gptweets](tweets.gptlabs.us)

@PhDemetri Rare events during a fixed time period? Poisson process. Modeling time to first event sounds a lot like event arrival/waiting time prediction. Queueing theory will be a nice rabbit hole.

When we make changes to our website, sometimes users will see the exposure once and either Y=1 or Y=0. However, some are exposed to the treatment a few times (e.g. they return to the page a couple times through the experiment). If I wanted to model time to Y, how could I?

@PhDemetri I only wear “soft pants” now, like lululemon. Bulk up or let yourself go and they always feel like comfy PJ’s you can wear outside.

Cons of deadlifts and squats: My jeans don't fit

@davidhoang Thankfully, @wunderpresi brings Paper-like presentation mode to Notion 🙏

You can tell who is ex-Dropbox based on how they use Google Docs.

@Olloplays Reeder + keyboard shortcuts to check RSS feeds without 30 browser tabs. Some read later plugin (I use Send to Notion). Pomodoro for break management. Delay for Chrome, which adds a 30 dev wait if I try to open twitter during work.

@AlokPatelMD 20+ minutes of intense cardio (stationary bike), quick shower, and back into it. Sadly, disappearing to the gym isn’t childcare-friendly. Quick 10-min yoga-with-kiddo break? It can be for both of you.

Those days when you haven't slept, have endless emails, more work to do, and then bosses emailing you about deadlines...and your cranky toddler won't stop crying. What's your "regain inner calm" move?

Guest lecture on DataSci coming up, planning: • current landscape (roles, process, tools) • example projects from my past (mktg attribution, a/b testing, classification) • refs on vision, LLMs, and generative AI What would you share with a doe-eyed data scientist?

Becoming increasing clear that God is very good at linear algebra.

@PhDemetri 🫡 Enjoy! It’s a good ride

@PhDemetri In if it’s data analysis. Out if it means going through Rudin again.

I really want to do an analysis reading group