The Analytics Engineering Podcast

The episode contains a handful of genuinely useful practitioner insights - most notably the 40%→90% reliability jump from fixing data modeling rather than adding more context, and the org-level memory governance problem - but these are diluted by lengthy personal backstory, a consulting-industry tangent, and a meandering French-accent opener that together consume roughly a third of the runtime.

“I was stuck at like 40% reliability and I was just noticing that the agent was failing because you know, there was some ambiguity between two columns or there were ah, there was a metrics in two different table that was slightly different numbers. And so I just redid some parts of the data model, wrote some data documentation and that got me to like 90% reliability.”

“when you're at the company level you want to make sure that people don't teach wrong stuff to the agent. And that's where the data team still needs to approve what gets into the global memory of the company.”

The file-system-as-context-layer insight borrowed consciously from IDE tooling is a genuinely non-obvious transfer, and the analogy between 'plug agents directly into raw data' and the early-2010s 'BI plugged into production DB' era is a sharp framing; elsewhere the episode recycles widely circulating ideas about context, iteration, and open-source monetization without adding much new.

“agents just have like a bunch of text and files and they can just like grep and search things and that goes very fast and that scales very well. So that's what we learned from it and we wanted to put the same in our product.”

“we are the phase where people are like, oh, let me just connect cloud code to my smells like mcp and what could go wrong? And it's the same as when you had your BI plugged to the database”

Claire is a genuine builder - she actually ran the data stack at a 20→300 person startup and is now 6 months into a real open-source product with 80 companies in production - but she is an early-stage founder rather than a practitioner who has done context engineering at significant scale, which caps how much hard-won operational depth she can offer.

“we have like a thousand, three hundred stars, something like that after like six months of launch. Um, and we have, I would say like 80 companies with now in production.”

“I was the only data people and I didn't know data engineering, I didn't know the tools of the data stack. So I just built an ETL by myself.”

The 40%→90% reliability benchmark anchored to concrete failure modes (column ambiguity, duplicate metrics) and the Ramp Research example (Ian, dentist-usage question) are the strongest specifics; most other claims are framed as 'some data teams' or 'many companies' without named examples, real timelines, or dollar figures.

“I was stuck at like 40% reliability and I was just noticing that the agent was failing because you know, there was some ambiguity between two columns”

“The reason that I was thinking about that is that I just recently spoke to um, Ian from Ramp, um, and they have this kind of process up and running”

The host is domain-expert enough to ask genuinely sharp questions - pushing on whether a product is actually needed here and threading in Anthropic's blog post findings and metric-flow debate as real follow-ups - but he also leads witnesses heavily, lets several vague answers pass unchallenged, and spends the first ten minutes on personal biography and French-accent banter rather than substance.

“do you think that there is a, a product to be built here? I mean, there certainly is work to be done like practitioners. I mean the idea of a context engineer, I think clearly has legs. But do you think that there's a product needed in this space or is it just a set of best practices”

“if there's a golden path metric, then use it. Otherwise, you know, try to figure it out yourself”