E236｜99%的作业都是AI写的：当代名校生眼里，大学还剩下什么？

Summary

This 硅谷101 episode interviews [[AlfredLinTongyu|Alfred 林童雨]], [[KelentoHouTaiyu|Kelento 侯泰宇]], and [[JackRaoJiewu|Jack 饶街五]] about university life after ChatGPT, Claude, Gemini, Claude Code, Codex, Manus, and Open Claw became normal tools. The guests describe AI as tutor, coding partner, assignment copilot, search layer, and daily infrastructure, while arguing that the remaining value of university moves toward Human Judgment Under AI, trust, peers, teachers, self-discovery, and knowing what problem is worth solving. The episode’s strongest addition is AI Default Learning Environment: younger students may not treat “without AI” as a realistic baseline at all.

Key Claims

• [[AlfredLinTongyu|Alfred 林童雨]] used AI to move from law toward AI and agent engineering, describing AI As Tutor as a personalized teacher for CS, AI theory, law case review, and project retrospectives.
• Alfred’s [[TsinghuaUniversity|清华大学]] experience makes university valuable through high-density peers, office-hour access, and trust-building, not only through lectures or memorized knowledge.
• [[KelentoHouTaiyu|Kelento 侯泰宇]] treats NYU as a resource node for environment, AI research exposure, and social access, while arguing that [[DegreeAsTrustCredential|degrees remain trust credentials]] even if AI weakens knowledge proof.
• Kelento frames judgment as the difference between deciding problems with objective answers and choosing under values, pain, experience, and uncertainty.
• [[JackRaoJiewu|Jack 饶街五]] says about 99% of his assignments or papers had substantial AI participation, especially coding-heavy work and open-ended coursework.
• The episode argues that grades in AI-open environments depend on tool skill plus underlying course understanding; a student who cannot judge AI output can still submit plausible errors.
• Professors’ policies are uneven: some courses still resist GPT use, while others require tools such as Claude Code, Replit, or similar coding assistants from the first class.
• AI University Assessment Reform follows from that unevenness: assessment should move from policing whether AI appeared to checking process, responsibility, judgment, verification, and real-world problem solving.
• Heavy users describe AI as addictive because it supplies execution speed, learning feedback, and ideas; this extends AI Use Pacing from agent queues into school and early-career life.
• The source qualifies productivity hype through the cited “productivity placebo” result: senior engineers may feel faster with AI while verification and interaction overhead can make them slower.
• The guests disagree with a simple “AI makes foundations weaker” story. They say AI can deepen learning when used for explanation and exploration, but also admit that people who began coding with AI may be much less capable when tools disappear.
• Employment anxiety shifts from “AI replaces people” to “people who use AI well replace people who do not,” while legal junior roles, paralegal work, and some software-entry jobs face pressure.
• The closing high-school anecdote pushes the source beyond current college students: some younger learners reject the premise of an AI-free future, making AI Default Learning Environment a generational rather than merely tactical claim.

Key Quotes

“99%” - Jack’s answer for how much schoolwork had AI involvement.

“人不是被 AI 替代，而是被更会使用 AI 的人替代” - Kelento’s job-market formulation.

“为什么要做这个假设，因为我们不会离开 AI 了” - the younger-student response in the host’s closing anecdote.

Connections

• 硅谷101 - show context for the episode.
• [[AlfredLinTongyu|Alfred 林童雨]], [[KelentoHouTaiyu|Kelento 侯泰宇]], and [[JackRaoJiewu|Jack 饶街五]] - three student/early-career guests grounding the discussion.
• [[TsinghuaUniversity|清华大学]], NYU, and [[ColumbiaUniversity|Columbia University]] - university contexts used to discuss peers, teachers, AI research exposure, and credential value.
• ChatGPT, Claude, Gemini, Claude Code, Codex, Manus, Open Claw, Perplexity, and Replit - AI tools and developer surfaces named in the students’ workflows.
• AI As Tutor, AI Default Learning Environment, and AI Shortcut Risk - constructive AI learning use, the default-environment shift, and the risk of bypassed cognitive practice.
• Degree As Trust Credential, College Career Preparation, College Major Choice, and University Opportunity Density - university value and career-preparation branches extended by the source.
• AI University Assessment Reform, Transparent AI Use, First Draft Thinking, and AI Writing Pedagogy - assessment and academic-integrity branch.
• Human Judgment Under AI, AI Coding Verification, and AI Use Pacing - output responsibility, coding correction, addiction, memory, and productivity-overhead concerns.
• Yann LeCun - NYU AI researcher whose presence influenced Kelento’s university choice.

Contradictions

• No direct contradiction found with existing wiki content.
• The source qualifies What do students lose when they rely on AI for homework? and First Draft Thinking: it does not refute the value of a protected first attempt, but it shows that heavy AI-native students may no longer see AI-free work as a realistic default.
• The source also qualifies optimistic AI As Tutor pages by adding a dependency risk: AI tutoring can deepen learning, but students who never built pre-AI foundations may have weaker fallback capacity when tools fail or hallucinate.
• The sample is elite, AI-heavy, and startup-adjacent, so its claims should not be generalized to all college students without additional sources.