Ken's Nearest Neighbors

The episode contains a handful of genuinely useful observations - the dual-correctness problem in data science vs. software engineering, the argument that MLOps is a narrow slice of a broader productionisation problem, and the social pressure voter-outreach finding - but they are buried under extensive career-path storytelling, language-war banter, and mutual affirmation. Insight rate is low relative to runtime.

“it's very difficult to write testing for actual results in data science because usually like you wouldn't be doing it if you knew what the answer was”

“MLOps is just this tiny, narrow slice of the pie which is like, once you've built a machine learning model, how do you serve that machine learning model?”

The archaeology-vs-architecture analogy and the dual-correctness framing for data science are moderately fresh, but most content - R vs. Python is preference-based, LLMs are risky for junior devs, communication matters - recycles widely-held views without adding a genuinely contrarian or first-principles twist.

“the analogy I draw is like if, if uh, software engineering is like architecture, data science is like archaeology”

“there's a second sense in which it needs to work, which is like the answers need to be correct and they need to be what you meant them to be”

Alex Gold is a genuine practitioner - managed data science teams, administered production R environments, worked in political data science, now leads solutions engineering and technical support at Posit, and has written a practitioner-oriented book. Solid domain credibility, though not a scaled-company operator or widely-recognised industry figure.

“I ended up leading a team in that role”

“I found myself managing an RStudio server because somebody needed to do it. And so what I found was I actually kind of enjoyed that part of it”

The voter-file and randomised-outreach experiments are usefully specific (NLSY surveys named, RCT design described, social-pressure technique cited), and the team composition estimate (70/30 data science vs. IT admin) is concrete. However, the Scikit-learn lasso anecdote is conspicuously vague ('it was like the penalty value wasn't what people thought'), and most DevOps discussion stays at framework level without named tools, timelines, or customer metrics.

“each secretary of state for every state keeps a voter file, which is a list of all the people who are registered voters in that, in that state, um, and when they have voted”

“about 70, 30, 60, 40 between people who have a data science background and, and people have like an IT administration background”

The host asks reasonable redirecting questions and occasionally shares relevant experience that moves the conversation forward, but consistently defaults to agreement ('I 100% agree with that sentiment') rather than probing or challenging. The LLM exchange is the only moment of genuine back-and-forth tension; elsewhere, claims go unchallenged and the host's personal anecdotes frequently displace guest insight time.

“I think, uh, I, I 100% agree with, with that sentiment”

“Are there any key factors for improving voter turnout that are you just like. Yeah, like hugely correlated with improving”