AB Testing

The episode contains some substantive ideas, particularly around Jevon's Paradox and its application to AI-driven code commoditization. However, the conversation is heavily padded with banter, personal anecdotes, and meandering asides that dilute the insight density. The core insight about how automation of code production creates new opportunities rather than eliminating jobs is valuable, but it takes 20+ minutes to get there and the supporting analysis is somewhat repetitive.

“code is a commodity...things that no one bothered coding before, there's now going to be a return on that investment”

“there's not only a long tail, but a very long tail of things that could be improved by code that haven't been because the cost is too high”

The application of Jevon's Paradox to modern AI and code production is somewhat novel in framing, but the underlying economic principle is not new. The discussion about quality as a system property, domain experts creating code, and the transition of roles is thoughtful but not particularly contrarian or first-principles thinking. Much of this echoes existing conversations in the software development community.

“Jayvon's paradox is this phenomenon where for certain types of goods, the elimination or the automation of how we produce that good actually doesn't end up removing jobs”

“the purpose of the role will persist, the tools will change”

Brent is an experienced practitioner at Microsoft with significant depth in AI, testing, and organizational dynamics. He brings real operational experience and has tangible insights about team transitions and LLM limitations. However, this is a casual conversation between two hosts/former colleagues rather than a formal guest interview, which reduces the value of his caliber somewhat. He has domain credibility but limited external validation of specific accomplishments.

“we have to very rapidly adapt and move in a different direction or we're not going to be here in 6 to 12 months”

“I had built a really sweet...dashboard...very responsive”

While Brent references Gartner reports and mentions specific examples (the Power BI dashboard anecdote, the college debate group project), most claims lack concrete numbers, timelines, or named companies. The Gartner statistic about rehiring 50% by 2027 is cited but not detailed. The dashboard story is concrete but brief. Most of the discussion remains at an abstraction level without sufficient grounding in data or specific case studies.

“Gartner claims it is and that they further predict that by 2027 many...of the companies other than hyperscalers...are going to end up hiring back at least 50% of the people they laid off”

“I had used LLM...very responsive dashboard...bringing in signals near real time”

Alan frequently plays the role of skeptic and pushes back (e.g., 'Because it's my job...this of course it's inevitable. But is it happening? No.'), which creates some productive tension. However, much of the conversation devolves into tangential banter about pickleball, microphones, and personal jokes. Many of Alan's follow-ups are genuine but the episode lacks tight questioning structure. The hosts seem more interested in riffing than systematically exploring claims.

“Why are you being a dick?”

“Because it's my job. Because you're...talking like this has already happened and it's not”