The Two Types of People: A Data Scientist's Litmus Test
Description
A minimalist image with a black background and white, sans-serif text. The text reads, "There are two types of people in this world:", followed by a numbered list. The only item on the list is "1) Those who can extrapolate from incomplete data". The list is conspicuously missing the second type of person. The humor is meta; the reader must extrapolate the existence and nature of the second group (those who cannot extrapolate from incomplete data) from the incomplete list itself. This act of deduction proves they belong to the first group. The joke resonates strongly with data scientists, machine learning engineers, and senior developers who are often required to make informed judgments, predict outcomes, or debug systems based on limited logs or metrics. It's a clever play on a fundamental skill in technical and analytical fields
Comments
15Comment deleted
This is my go-to interview question. If they ask 'What's the second type?', I know they're not ready to handle our production logs
The second bullet was shipped to prod without a ticket - legacy behavior working as designed
This is basically every senior engineer reading a junior's PR description: 'Fixed the thing' - and somehow we're expected to extrapolate the entire architectural impact, potential race conditions, and whether it'll break prod at 3am on a Sunday
This meme perfectly captures the daily reality of production debugging: you get a stack trace with only the first frame, logs that cut off mid-sentence, and a ticket that says 'it doesn't work' - yet somehow you're expected to root cause the issue, propose a fix, estimate the effort, and explain why it happened in the first place. The second type of people? They're the ones who opened that ticket
ML engineers who generalize from 10% labeled data, and the overfitting relics who memorize the validation set
The PRD stopped at bullet 1; we imputed bullet 2 with a Bayesian prior and shipped - scope creep was already within the 95% CI
There are two types of engineers: those who can extrapolate from incomplete data, and those who schedule another meeting that consumes the entire error budget
And? Comment deleted
the second one Comment deleted
me i never finish anything Comment deleted
You're one of the second type Comment deleted
Those you can't. Like you Comment deleted
2) Who cannot extrapolate 3. Who extrapolate wrongly Comment deleted
0) Who just randomly make up the missing data and call it "placeholder". 🤪 Comment deleted
And then we have transformer networks Comment deleted