Wink Pings

A Chicken That Doesn't Want to Go to KFC, and the Invisible Scorecard in the AI World

At SGLang Summit SF, Chenyang Zhao presented three judgments: RSI, evaluation, and responsibility. In his year-end compilation, he uses a chicken to discuss meritocracy, and also talks about 140 emails, CMU, SGLang, and leaving school. Some online commenters said this chicken ultimately ended up at an equally high-end brand.

Chenyang Zhao's talk at SGLang Summit SF is now online; the podcast was recorded by the Crossing team. He mentioned that he and Koji share a partial hometown connection, and that the Sichuan-Chongqing AI community is full of capable people. In the talk, he made three judgments, and then compiled the views he had expressed this year into a year-end collection; the first chapter is titled 'A Chicken That Doesn't Want to Go to KFC.'

image

Video podcast on Xiaoyuzhou: 【Year-End Recap ③】Using Kimi to Accelerate Kimi: On the Eve of RSI, 10 Aha Moments from an AI Infra Person | In Conversation with Chenyang Zhao

十字路口Crossing

Of the three judgments, the first is about RSI. Chenyang Zhao said that things we cannot evaluate cannot be optimized. Now we need to add: things that can be evaluated can be optimized. That sounds like engineering common sense, but in the AI industry it stings a little. A commenter, soren law, added a sharp point in reply: once evaluation becomes a reward, optimization easily aims at the leaderboard rather than necessarily at the original problem.

The second judgment is about AI replacing humans in substantive production. Chenyang Zhao sees this as a dangerous future. People who rely on charity for economic assistance would be stripped of most of their freedoms, and society needs to think seriously about how the fruits of labor are distributed. He quoted a line: 'Making decisions for others is arrogance, but being unable to make decisions for oneself is terrifying.'

The third judgment is about responsibility and authority. Before AI can replace humans in bearing responsibility—or before humans reach a consensus to hand all responsibility and power over to AI—the importance of the relationship between responsibility and authority only grows. Chenyang Zhao's words were very direct: in a sense, the fact that humans can still take the blame is an important reason why mass unemployment has not yet happened.

The views in the podcast lean toward industry judgments, while the collection article is more personal. Chenyang Zhao uses the chicken metaphor to unpack meritocracy. If a chicken cannot get into KFC, McDonald's, or Chick-fil-A—or even into Zhengxin Chicken Steak—does a chicken life like that still have value? This is irony. He has seen too many talks about OpenAI salaries, Meta compute, and Google lunches. These companies are great, but he does not believe a person's value needs to be measured solely by whether they join one of them.

The most insidious thing about meritocracy is that it makes a person's value increasingly equivalent to which institution selected them. Which school you got into, which company you joined, which team you are on, how much you make—these were originally just facts about a life, but they gradually become the person themselves. Chenyang Zhao also mentions something his high school teacher Xiong Licheng said: when a person succeeds, many of their original traits—even their flaws—are redefined by others as reasons for their success. Looking forward, meritocracy ranks people by who gets selected; looking backward, it writes the past of the selected person as foreshadowing. Anxiety is written as ambition, and chance is written as vision.

Chenyang Zhao does not exempt himself. He writes that he, too, was once a chicken who really wanted to get into KFC. When choosing a major, he thought about income and respectability, and when he heard that a senior student had gone to a certain AI company, he would quietly wonder whether he could make the cut too. The article contains a series of concrete experiences: in high school he studied for chemistry competitions, and later got into Tsinghua's computer science program; during the summer between sophomore and junior year he missed the chance to join Professor Tang Jie's lab, and later saw someone calculate that the stock from an early Zhipu offer might be worth 100 million yuan—he felt that kind of counterfactual accounting was impossible to do; he sent more than 140 emails applying for summer research and received very few replies; one morning he got a reply from Graham Neubig, jumped out of bed, got dressed, walked to the campus river north of the Zijing dormitory, and shouted out loud; at CMU he worked on Prompt2Model with Graham and Vijay Viswanathan, learning unit testing, CI/CD, small PRs, and review—software engineering practices that undergraduate classes had not taught; later he joined SGLang because he needed an embedding model, emailed the author, completed a small task, and was pulled into a seven-person developer group; he spent 15 months in a PhD program at UCLA and left around June 2025; after interning at RadixArk as a student for more than a month, he became full-time—strictly speaking, he is currently on a leave of absence.

These details are more believable than 'hard work will eventually pay off.' Chenyang Zhao writes that among those 140-plus emails, the one that got a reply was not necessarily the best written, and the professors who did not reply were not necessarily ignoring him because he was not good enough. He did everything he could, and then one morning, luck took his side.

After the article was published, there was a sharp addition in the replies. Yuan Chaofa said that this chicken that did not want to go to KFC ultimately, as a matter of fact, entered an equally high-end brand. If the article had been written by a student who could not find a job for the time being, students on non-mainstream paths would find it more relatable. This reminder is not mean; it points out the position of the narrative: an anti-meritocracy article can also be written by a winner of meritocracy.

At the end of the article, Chenyang Zhao does not give an answer. He says he used to always feel that being able to leave academia was itself a reward; now that this reward has been realized, it has also ended. That was a goal defined by leaving—it told him where he did not want to stay, but not where he wanted to go. What he can say for sure is that he does not want his next reward to be merely another more famous store. A chicken that does not want to go to KFC is not nobler than other chickens; it also needs to eat, it also feels envy, and it also follows the crowd when anxious. It simply does not want its value to be determined by which store it finally enters.

This line can easily be used to justify lying flat, and it can easily be taken as chicken soup. Put back into the concrete experiences of the article, it is more like an unfinished scorecard: evaluations, leaderboards, offers, schools, and companies are all external scales; whether a person treats them as their own reward function is another matter. Chenyang Zhao says that if he later figures out what that reward is, he does not plan to look back and write today's hesitation as foreshadowing.

发布时间: 2026-11-16 23:27