I like to think about how to spend money well, with joy, efficiency, and ideally—being from Guangzhou which is literally the most delicious region in China, I take food seriously. Lately, there’s been an interesting bit of culinary gossip about the tech giants behind our takeout orders. I’ve written before about Meituan, the heavyweight in China’s food-delivery duopoly—a platform that has perfected the dark art of rent-seeking. It squeezes the life (and margin) out of both the restaurants and the couriers. The folks ferrying my salad across the city are pulling 10 to 14-hour days for around 6000 RMB a month. Meanwhile, restaurants are bleeding out 30% of their earnings just to be listed. Efficiency is a beautiful thing—until it starts to look like feudalism in a scooter helmet. ...

The art of internet archeology led me to the distinguished lecture given by Sendhil Mullainathan, “Economics in the Age of Algorithms”, at AEA Annual Meeting that took place in Hilton San Francisco Union Square—I’ve been there, several times! The talk took place Jan. 3, 2025. The video is available https://videosolutions.mediasite.com/Mediasite/Play/cb9d64c0274d4aae98b61dd6779791b31d I’ll write about it someday :)

What Is This Poster Even For? My advisor is letting me redesign the poster (the previous version isn’t that bad… isn’t it?) for our paper “Price Stability and Improved Buyer Utility with Presentation Design” (pdf) that’s going to be presented at The Web Conference. Fine. But, to be fair, he made a solid point. Yes, there are endless design guidelines and aesthetic choices—softwares (please, don’t use Powerpoint or Word…), color palettes, font choices, the use of white space (oh the cliche ’less is more’), clever placements of shapes and arrows to guide people’s eyeballs, etc etc etc. But what about the practical question: what do we actually want the poster to do? ...

I do have some (well, a little) Theoretical-CS background. I thought I was learning math for computer science in class—I took notes, solved problems, or may have even nodded sagely once or twice. But only after when I start reading the following books did I realize… ah, this is what they were trying to teach us. Whether you’re into algorithms, theory of computation, or occasionally need to find a quick recap or intermediate result to bypass some calculations yourself, the following are the textbooks that our professors quietly (or loudly) revere—slightly biased, but just to get started. ...

Ah, the sweet smell of competition: Freebies are flying on campus. Verified US students can snag two free months of ChatGPT Plus from March 31 to May 31, 2025 (chatgpt.com/students). And Google’s Gemini Advanced become freely available to US college students all the way through finals 2026 from April 17, 2025 (gemini.google/students/). Why Tech Giants Love Students (and Their Data)? Discounts to college students are the cheapest brand loyalty booster. By getting students accustomed to their ecosystem and tools (AI, cloud services, productivity apps), tech companies help shape the default workflows and preferences of the future workforce, creating built-in demand and fashion. ...

It’s a gift to have sharp music memories—the kind that catch a familiar vibe in something new. The two pieces below don’t share the same melody, but there’s something in the way the instruments blend, the vocals drift, the atmosphere builds. Hit play and feel the resemblance for yourself:

Scoring rules are really cool, and useful. This seminal survey paper has over 6k citations as of April 2025 :o Strictly proper scoring rules, prediction, and estimation tandfonline [pdf] Scoring rules assess the quality of probabilistic forecasts, by assigning a numerical score based on the predictive distribution and on the event or value that materializes. A scoring rule is proper if the forecaster maximizes the expected score for an observation drawn from the distribution F if he or she issues the probabilistic forecast F, rather than G ≠ F. It is strictly proper if the maximum is unique. In prediction problems, proper scoring rules encourage the forecaster to make careful assessments and to be honest. In estimation problems, strictly proper scoring rules provide attractive loss and utility functions that can be tailored to the problem at hand. This article reviews and develops the theory of proper scoring rules on general probability spaces, and proposes and discusses examples thereof. Proper scoring rules derive from convex functions and relate to information measures, entropy functions, and Bregman divergences. In the case of categorical variables, we prove a rigorous version of the Savage representation. Examples of scoring rules for probabilistic forecasts in the form of predictive densities include the logarithmic, spherical, pseudospherical, and quadratic scores. The continuous ranked probability score applies to probabilistic forecasts that take the form of predictive cumulative distribution functions. It generalizes the absolute error and forms a special case of a new and very general type of score, the energy score. Like many other scoring rules, the energy score admits a kernel representation in terms of negative definite functions, with links to inequalities of Hoeffding type, in both univariate and multivariate settings. Proper scoring rules for quantile and interval forecasts are also discussed. We relate proper scoring rules to Bayes factors and to cross-validation, and propose a novel form of cross-validation known as random-fold cross-validation. A case study on probabilistic weather forecasts in the North American Pacific Northwest illustrates the importance of propriety. We note optimum score approaches to point and quantile estimation, and propose the intuitively appealing interval score as a utility function in interval estimation that addresses width as well as coverage. ...

Didn’t know calculating expectation would take this long… Consider $n$ i.i.d. random variables $X_i$ drawn from Power Law distribution with support lowerbound $x_\text{min} = 1$ and the shape parameter $\alpha > 2$—in other words, the pdf of $X_i$ being $$ f_{X_i}(x) = (\alpha - 1) x^{-\alpha }, $$ and $$ F_{X_i}(x) =1 - x^{-(\alpha - 1)} \quad \text{ for }x \ge 1. $$ Btw, for the heavy-tailed distribution family, it can be more comfortable (and general) to understand random variable $X$ as $$ \Pr[X > x] \sim x^{-\alpha}. $$ see Ibragimov and Walden (MS'10)’s Section 3 “Heavy-Tailed Distributions” for a cooler, less intuitive way of seeing it. ...

So are advertisement good or bad? There’s no fixed answer. One view says it’s great cause advertisers have to pay for ads—so only sellers who believe that their own products/services are of good quality will use ads (if ads price and the ads market adjust appropriately)—then ads become informative and useful signals, and they would help consumers distinguish valuable products out of the sea of alternatives. Another view is that ads are more like competitors for attention—and it becomes a bad equilibrium now when seller would have to buy ads to at least get any attention. ...

Four papers (topic) from the C&A (complexity & algorithms) workshop 2025 that I personally found interesting (or, perhaps, just understood), hosted by Sun Yat-Sen University, Guangzhou China. Two EconCS papers: Persuasive Calibration Yiding Feng and Wei Tang, 2025 (Arxiv link) The paper studies a persuasion problem consisting of an information sender and a receiver. It’s somewhat like information design—but calibration replaced commitment to an information policy. The model is restricted to a state in [0, 1] (unknown success probability of a Bernoulli experiment). Calibration means when the sender sends an signal $q$ it’s close to the expected value of the real state. ...