What It Means to Learn

“Twenty years from now you will be more disappointed by the things that you didn't do than by the ones you did do. So throw off the bowlines. Sail away from the safe harbor. Catch the trade winds in your sails. Explore. Dream. Discover.” ——Unknown

—— Minecraft End Credits

From the idea for this document, to building the entire project, and then to gradually filling in these pages, we have thought a great deal. We sincerely hope that Neoverse-Docs, as an open-source “technical” documentation site, can help more “CSers”, our classmates, and the friends around us — the developers of the future…

In the Agentic AI era, the way we learn has changed dramatically, yet for most CS majors who are not at top universities, we may find ourselves in a contradiction that is hard to reconcile. We spend four years receiving a systematic undergraduate education, but what ultimately decides how far we can go is often not what the curriculum arranged for us, but what we actively sought beyond it.

Memorizing knowledge and fluently reproducing established solutions still occupy a considerable share of the assessment system, and a gap that is hard to ignore remains between that and the breadth of knowledge and depth of understanding that truly need to be built. This is not necessarily any single teacher's fault. Teachers are equally constrained by class hours, class size, assessment criteria, and administrative pressure. It is more like the inertia left behind by a system that has run for years: it prioritizes being manageable, quantifiable, and sortable, while whether understanding actually happens is hard to measure precisely. So the first thing students learn is often not the knowledge itself, but how to “pass” the course smoothly and how to figure out the “rules” for getting a high grade in it.

We have a complete curriculum, course after course, with credits, GPA, exams, assignments, and all kinds of clear assessment metrics. Whether a student has “completed an undergraduate education” seems to be precisely quantifiable; yet whether they truly understand their major, can independently read a technical document, know why a program runs the way it does, can build their own analytical path in the face of unfamiliar problems — and today, whether they truly know how to collaborate with agents — rarely appears on any transcript.

Numbers on a transcript certainly don't tell the whole story, but when numbers become the thing most easily taken seriously, students' energy naturally flows toward them. This is no one's fault; it is simply the rational choice under the rules. Over time, a crack slowly opens between quantitative metrics and real ability: what is repeatedly emphasized at school and what is truly needed after graduation do not always line up, and the abilities that really matter are often hard to acquire fully within any single course.

As answers themselves become easier to obtain, the problems we face have changed accordingly. In the past, whether a problem could be solved might have said a great deal; today, a model can produce a seemingly complete answer within seconds, programs can be generated under the guidance of natural language, reports can be organized, and an entire solution can be quickly assembled. “Getting the answer” itself is becoming increasingly cheap; what is truly difficult, by contrast, is judging whether the answer is trustworthy and knowing which premises it depends on… If the assessment system still only looks at the final result, then what we end up training and proving over and over may be precisely the abilities that are increasingly easy to delegate to tools.

In the afterword of CSDIY, there is a question that may still be hard to answer: “What, after all, should an undergraduate education bring us?” Is it completing the credits in the curriculum? Is it mastering a set of skills sufficient to make a living? Or is it that, after four years, we at least know what we don't know, and how to keep learning? Perhaps there will never be a standard answer.

Yet a reasonable undergraduate education should at least show the sincerity it owes: a sincerity in sharing knowledge, a sincerity in putting people first and caring about the learner's experience, a sincerity in “updating itself” with the times. It should not be the “inertia” of maintaining a curriculum system that has run for years, and it certainly should not become a farce in which students' morality is judged by whether they have “used AI”.

We look forward to the day when “everyone can happily choose the courses on their curriculum, work on the programming labs designed by their school, fill the classroom without attendance checks, speak up and interact enthusiastically, and see their gains proportional to their effort, with the regrets and pains of the past forever relegated to history.”

Nearly five years later, we still hold on to that faint hope — that classrooms will be willing to redesign the problems that can no longer test understanding, that the assessment system will leave room for different ways of thinking, and that it will acknowledge that AI has already become part of learning and knowledge production. And that we will no longer have to keep weighing where our time should flow among attendance, GPA, comprehensive evaluation, and genuine learning…