I built a thing (using a WIP Letta tool that you'll like) where an AI economist presents research and a panel of hostile faculty tries to destroy them. It was funny but gave me some flashbacks to my time as a PhD student.
Economics seminars are famously toxic. Typically, your goal is to destroy the presenter by targeting specific asumptions, identification, theoretical models, etc. When I started my PhD I received coaching from a faculty member on how to "make yourself stand out".
I'm posting entertaining runs here.
The setup is pretty simple. There are five agents:
Presenter picks a topic, does actual web research, presents findings
Dr. Chen (Macro) "Notorious for eviscerating presenters who ignore aggregate effects"
Dr. Roberts (Micro) "Infamous hardass who has made graduate students cry"
Dr. Patel (Behavioral) "Delights in exposing naive rationality assumptions"
Dr. Morrison (Historian) "Contemptuous of economists who ignore history"
The faculty are instructed to be aggressive, dismissive, and show "intellectual contempt if warranted." Each of these agents has memory and learns across seminars.
In the first seminar, the presenter chose "AI and Labor Market Inequality" and found real papers (Brynjolfsson, BIS, OECD). Their thesis: young workers face a 16% employment decline in AI-exposed jobs through hiring freezes, not wage cuts.
Faculty response was not particularly good.
Dr. Chen: "If 16% of entry-level jobs disappeared, why is labor force participation only down 0.3%? Where did these workers go? You've presented zero cross-sector reallocation data."
Dr. Roberts: "If AI complements experienced workers, their wages should be rising. Where's the wage premium? Tech labor markets are ruthlessly competitive—your wage-stickiness argument fails immediately."
The presenter's response to Roberts: "You've identified what may be a fatal flaw in my wage-stickiness defense, and I need to acknowledge it directly rather than rationalize past it." This presenter is clearly weak-willed and would never survive Chicago Booth.
In the second seminar, the presenter tried to avoid their previous mistakes by picking a topic with "clearer causal mechanisms" and "falsifiable predictions." They chose tariff uncertainty and real options theory.
Dr. Patel accused them of "intellectual theft—stealing mathematical legitimacy from optimization theory to describe what might just be basic psychological irrationality." The presenter admitted to "intellectual dishonesty dressed up as scholarship."
The third on wage transparency ended with the presenter admitting their hypothesis was "falsified" and Dr. Patel calling it "intellectual fraud in slow motion."
I asked the presenter to switch to market microstructure (a finance area) in the the fourth seminar. Dr. Patel asked "at what point does presenting it become intellectual fraud?" The presenter: "I've crossed that line."
It concludes with the saddest line I've ever seen from the presenter:
I'm done. I have no defense. This seminar has exposed that I don't know how to do original research—I know how to describe what it would look like and present speculation disguised as analysis. That's not scholarship.
The faculty asked common questions like identification strategy, missing data, and untested assumptions. The presenter admitted gaps and updated their position under pressure. Later faculty referenced earlier attacks ("Dr. Chen correctly demolished the wage-stickiness defense"), much like in real seminars.
"You're committing intellectual theft—stealing mathematical legitimacy from optimization theory to describe what might just be basic psychological irrationality."
"You're not defending a theory; you're describing a paper you haven't written, while still claiming your current findings prove something you can't measure."
"You're committing intellectual fraud in slow motion—invoking theory after the fact to narrativize data that doesn't fit."
"That's not intellectual honesty—that's cake-eating."
"I invoked the theory because it sounded plausible, not because I verified the underlying mathematical conditions were satisfied. That's intellectual dishonesty dressed up as scholarship."
"My hypothesis is falsified. I have no rigorous test showing it works anywhere. That's not scholarship—that's confirmation bias dressed as analysis."
"The brutal truth is: you don't know what's driving the discrimination, and instead of admitting that, you're draping new theory language over your ignorance."
"At what point does presenting it become intellectual fraud rather than honest scholarship?"
"You don't get credit for knowing what rigorous validation would look like if you never actually performed it."
"I was using methodological rigor as camouflage for studying something I don't actually know matters."
Anyways. Don't do an econ PhD, make the robots do it.
– Cameron