Selling the Future
Some Simple Economics of AGI
Christian Catalini, Xiang Hui & Jane Wu
MIT Working Paper, February 2026
"[T]he binding constraint on growth [will be] human verification bandwidth: the scarce capacity to validate outcomes, audit behavior, and underwrite meaning and responsibility when execution is abundant. We model the transition toward AGI as the collision of two racing cost curves: an exponentially decaying Cost to Automate, driven by compute and accumulated knowledge, and a biologically bottlenecked Cost to Verify, bounded by human time and embodied experience...Rents migrate to what remains scarce: verification-grade ground truth, cryptographic provenance, and liability underwriting — the ability to insure outcomes rather than merely generate them."
The Impact of LLMs on Online News Consumption and Production
Hangcheng Zhao & Ron Berman
University of Pennsylvania Working Paper, February 2026
Abstract:
Large language models (LLMs) change how consumers acquire information online; their bots also crawl news publishers' websites for training data and to answer consumer queries; and they provide tools that can lower the cost of content creation. These changes lead to predictions of adverse impact on news publishers in the form of lowered consumer demand, reduced demand for newsroom employees, and an increase in news "slop." Consequently, some publishers strategically responded by blocking LLM access to their websites using the robots.txt file standard. Using high-frequency granular data, we document four effects related to the predicted shifts in news publishing following the introduction of generative AI (GenAI). First, we find a moderate decline in traffic to news publishers occurring after August 2024. Second, using a difference-in-differences approach, we find that blocking GenAI bots can be associated with a reduction of total website traffic to large publishers compared to not blocking. Third, on the hiring side, we do not find evidence that LLMs are replacing editorial or content-production jobs yet. The share of new editorial and content-production job listings increases over time. Fourth, regarding content production, we find no evidence that large publishers increased text volume; instead, they significantly increased rich content and use more advertising and targeting technologies. Together, these findings provide early evidence of some unforeseen impacts of the introduction of LLMs on news production and consumption.
Social media advertising loads as prices
George Beknazar-Yuzbashev et al.
Columbia University Working Paper, February 2026
Abstract:
Most digital platforms are funded through advertising rather than direct payments. Why? We argue that three main factors could explain this prevalence: users are more sensitive to monetary prices than to ad loads, microtargeting improves the match quality between users and ads, and platforms can personalize ad loads and thus price discriminate. We conduct a field experiment on Facebook with 1,810 users who install a browser extension that (i) hides nearly all ads or (ii) replaces microtargeted ads with untargeted ones. We find that hiding 82% of ads increases time on the platform by only 6%, showing that users are highly insensitive to ad loads. Removing targeting sharply reduces ad clicks and long-run engagement, indicating that targeting increases the match quality between users and ads. Finally, two-thirds of ad-load variation occurs across users, consistent with ad-load discrimination. Counterfactual simulations indicate that an ad-funded model performs at least as well as a subscription model in terms of profits and delivers higher consumer surplus. The key mechanism is that users are much less sensitive to ad loads than to monetary prices, making advertising a relatively efficient revenue source.
The Semantic Risk Premium
Rouzbeh Rezaei Sanjabi
University of Indianapolis Working Paper, February 2026
Abstract:
Prediction markets aggregate information through prices, yet their efficacy depends critically on the clarity of contract language. We introduce the Semantic Risk Score (SRS), a novel measure of linguistic ambiguity in prediction market contracts derived from large language model predictions of dispute probability. Using dispute data from the UMA Protocol (N = 804) to validate our measure, we demonstrate that SRS correlates significantly with actual oracle dispute variance (ρ = 0.157, p < 0.001). Applying this validated instrument to Polymarket contracts (N = 246), we find that semantic risk manifests through liquidity withdrawal: a one-standard-deviation increase in SRS is associated with a substantial collapse in trading volume (t =-4.41, p < 0.001). Paradoxically, high-ambiguity markets converge faster to extreme prices, which we attribute to adverse selection: sophisticated traders exit ambiguous markets, leaving homogeneous participants who rapidly reach false consensus. Our findings extend the Market for Lemons framework to prediction markets and have implications for contract design, portfolio optimization, and decentralized oracle governance.
Stopping Shopping at Stop and Shop? How Temporary Disruptions Affect Store Choice
Julia Levine & Sylvia Hristakeva
Cornell Working Paper, January 2026
Abstract:
Shopping patterns in retail markets are highly persistent, with households patronizing the same stores over time. Whether this persistence reflects unobserved heterogeneity or a causal effect of past choices through state dependence remains an open question. We study an 11-day strike that effectively closed 240 Stop & Shop grocery stores, using a novel identification strategy to isolate the strike's long-term effects on consumer demand through state dependence. We find that the strike caused households to make 9.9% fewer trips to S&S after the strike's resolution, simply by displacing planned visits during the strike. The reduction is observed immediately in the period after the strike's resolution and attenuates only gradually over time. The effect of trip displacement is larger for households who, during the strike, visit a store that they had not previously visited, suggesting that state dependence in store choice is partially driven by search and learning frictions. These results support an economically meaningful role of state dependence in grocery store choice, suggesting that temporary supply disruptions, and marketing tactics that induce consumer switching, can have long-term effects on profitability.