There's an old consensus in academia: a productive scholar publishes a dozen papers a year at most. Nicholas G. Polson of the University of Chicago Booth School of Business has stepped well past that line — in the opposite direction.
In late August, Columbia University statistician Andrew Gelman published a short blog post with a single number as its title: 258. That was the total number of papers he had counted under Polson's name for 2026 — and it was only August. On September 29, the post hit the Hacker News front page, drawing over a hundred comments.
📚 Who He Is
Polson is no fringe figure. He is a professor in Booth's econometrics and statistics department, a leading scholar on Bayesian methods in finance, and a frequently cited name at the intersection of risk management and machine learning. The quant finance community knows his contributions well: Markov chain Monte Carlo, stochastic volatility, and Bayesian perspectives on deep learning are all long-standing research areas of his.
OpenAlex records under his name tell the story: 2 papers in 2024, 6 in 2025. Then 2026.
238 papers.
📈 Three Details Beyond the Headline Number
The total alone could be dismissed as a counting artifact, but the breakdown is more intriguing.
First, co-authorship is highly concentrated. Sampling the 2026 papers, the same group of collaborators recurs; two main co-authors each appear in nearly half of the sampled papers. Same people, forty times the output.
Second, the topical breadth is remarkable. Bayesian testing, e-values, goodness of fit, analysis of a chess scandal, the mathematics of turtle trading rules, Bayesian approaches to the Navier-Stokes equations, AI and the nature of the firm — in one year, one person covered the territory of an entire department.
Third, there is no sign of slowing down. Gelman counted 258 on August 27; at that pace, the year was heading toward three hundred.
⚖️ How to Read This
First, the baseline: there is no evidence of misconduct. AI-assisted writing violates no written rules at most journals and preprint servers, and no authorship policy caps how many papers a person can publish per year. This is not a fraud case — it is subtler. Every individual step is compliant; the sum is unsettling.
The unease is about signal failure. Academia has counted papers for decades, and quantity has long served as a proxy for productivity, influence, and quality — tied to promotions, funding, and rankings. The moment large language models pushed the marginal cost of writing to near zero, that proxy broke on the spot. Writing was always the most automatable part of the paper pipeline; nobody automated it before simply because writing by hand was expensive.
Quant finance is one of the fields hit most directly. This community lives on signals — one failed signal in a model means lost money. Scholarly track records are also signals: firms invite professors to endorse models and consult based on the density of their CVs. When density can be produced on an assembly line, the buy side needs new due diligence tools.
Among the most upvoted HN comments was the next question: if writing is automated, how far behind can reviewing be? Papers produced with AI, eventually reviewed by AI — a complete loop when you think about it.
🧭 Why It Matters
The value here isn't gossip. The Polson case is among the first large-scale, publicly documented "AI academic productivity events," turning a hypothesis into a measurement: when the cost of writing goes to zero, how far can individual output go? The answer is roughly three hundred papers a year — two orders of magnitude above peers.
Everywhere still using "paper counts" to make decisions — hiring committees, grant reviewers, journal editors, quant funds paying professor retainers — should print this curve and pin it to the wall. The old ruler isn't fully broken yet, but the markings no longer mean anything.
---
Sources and caveats:
- Sources: OpenAlex author database record A5108378754 (annual counts for 2024/2025/2026 and co-author network, queried September 30; a name-disambiguation split profile accounts for roughly a dozen additional papers); Gelman's blog post of August 27, 2026 (titled "258," verified via the HN discussion); HN post of September 29 (135 points, 95 comments).
- Caveat: 258 is Gelman's personal count and does not exactly match the 238 papers retrieved from OpenAlex; both figures are presented here without adjudicating between them. Polson has not publicly commented on his writing methods, and this post makes no claim about "AI involvement" — it only presents the output structure.