Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes

URL SCAN: "Failing grades soar as professors see greater AI usage, dwindling math skills in UC Berkeley"
FIRST LINE: The percentage of failing grades in multiple UC Berkeley computer science classes in spring 2026 is significantly higher than past semesters and marks a departure from the department's grading guidelines.

THE DISSECTION

This article is a snapshot of early-stage cognitive deskilling being caught in real-time by an institution that hasn't yet internalized what it's witnessing. The professors are diagnosing symptoms—AI cheating, math unpreparedness, low engagement—as if these are correctable pedagogical problems. They are not. They are the first wave of a structural collapse in what it means to develop cognitive competencies under AI abundance.

The data is stark: CS 10 failure rate jumping from <10% to 35.3% in one semester. CS 61A from <10% to 10.6%. EECS 127 at 16.8%. All far exceeding departmental guidelines of 7% (lower division) and 5% (upper division) for D's and F's.

THE CORE FALLACY IN THE ARTICLE

The professors are treating this as a pedagogical failure—students not putting in "sweat," not attending office hours, cheating with LLMs. Garcia says "something is definitely wrong" and frames it as correctable via better standards and clear thresholds.

Wrong framing. This is the first empirical evidence of what DT predicts: when AI makes cognitive work effortless at the point of production, the feedback loop that builds cognitive competence breaks entirely. You cannot learn to code by having Claude code for you. You cannot develop mathematical intuition when AI does the proof work. The professor is witnessing the precise mechanism by which human cognitive capital formation stalls—and blaming the students for the structural incentive collapse that AI has created.

The phrase "Confusion is the sweat of learning" is pedagogically correct but institutionally irrelevant when the entire incentive structure now rewards delegating the confusion to a language model.

THE HIDDEN ASSUMPTION

The article assumes these are fixable teaching problems within the existing educational paradigm. The professors want better math preparation, reinstatement of SAT/ACT requirements, more engagement. They are trying to solve a structural crisis with pedagogical interventions. This is like installing better air conditioning in a building whose foundation is dissolving.

THE SOCIAL FUNCTION

This article is transition management theater. It documents real dysfunction while leaving the underlying cause unexamined. The article will be read by other professors nodding in recognition, shared on academic listservs, and filed under "AI challenges in education" while the systemic implications are systematically avoided.

The framing—academic dishonesty, lack of preparation, understaffing—channels the anxiety into individually attributable failures (lazy students, inadequate prep, budget constraints) rather than the structural collapse of the learning-competence-employment circuit that the DT framework identifies.

THE VERDICT

What you are watching is the early-stage symptom of P1 (Cognitive Automation Dominance) playing out in real-time inside a university CS department. The failure is not that students cheat with AI. The failure is that the educational model—which depends on students building cognitive competence through struggle—is being rendered incoherent by a technology that eliminates the struggle while leaving the credentialing intact.

The professors are describing exactly the mechanism DT predicts: the break in the feedback loop between cognitive effort and skill formation. When AI does the work, no skill forms. When no skill forms, exams expose the vacuum. The vacuum gets recorded as a failing grade.

The institutional response—better thresholds, more engagement, SAT reinstatement—is cosmetic. The structural problem is that AI has made cognitive delegation the path of least resistance, and there is no pedagogical or institutional fix that can overcome that gravity while the technology keeps improving.

This article is the canary. Spring 2026 is the timestamp. UC Berkeley CS is the subject. The diagnosis is: the feedback loop is broken and getting worse.

ARCHIVAL NOTE: This article belongs in the DT evidence corpus as a primary source documenting real-time cognitive deskilling at the institutional level, spring 2026, tier 1 university. The specific numbers (35.3%, 10.6%, 16.8% failure rates) constitute baseline evidence of the transition dynamics.