Convergence

The Credential Paradox

AI masters the specialist's knowledge while eroding the specialist's brain—and the only shelter is the combination no algorithm would recommend.

Convergence

THE CREDENTIAL PARADOX

EDUCATION

1.1

Test scores rose 1.1 percentiles more in year 2 of FL cell ban.

NBER WP 34388 (2025): 2nd yr cell ban boosted spring test scores 1.1 percentiles in high-use FL schools vs low-use.

NEUROSCIENCE

83%

ChatGPT users unable to recall their own essay content.

83.3% of ChatGPT group couldn't quote essays vs 11.1% in brain-only/search groups (Kosmyna et al., 2025).

ECONOMICS

56%

Double majors have 56% protection against earnings shocks vs. single majors.

Double majors experience 56% protection against earnings shocks, robust across model specifications and datasets (NBER WP 32095, 2024).

ARTIFICIAL INTELLIGENCE

89.4%

GPT-5 pro scores highest on PhD-level science questions.

GPT-5 pro (with tools) achieves 89.4% on GPQA Diamond, PhD-level science questions (2026).

THE CREDENTIAL PARADOX

As AI scores 89.4% on PhD-level science questions and heavy AI use drops neural connectivity nearly in half, the value of deep single-domain expertise is collapsing from both ends—machines replicate it while humans lose the cognitive capacity to maintain it.

EDUCATION

1.1

Test scores rose 1.1 percentiles more in year 2 of FL cell ban.

NBER WP 34388 (2025): 2nd yr cell ban boosted spring test scores 1.1 percentiles in high-use FL schools vs low-use.

NBER WP 34388 (2025): Suspensions rose 25% in 1st month of ban, 30% for Black males in affected schools.

NBER WP 34388 (2025): Unexcused absences fell post-ban, explaining ~50% of test score gains.

NEUROSCIENCE

83%

ChatGPT users unable to recall their own essay content.

83.3% of ChatGPT group couldn't quote essays vs 11.1% in brain-only/search groups (Kosmyna et al., 2025).

High-alpha brain connections dropped from ~79 (brain-only) to ~42 (ChatGPT) during writing (Kosmyna et al., 2025).

EEG showed weakest neural connectivity in ChatGPT group vs strongest in brain-only across 32 regions (Kosmyna et al., 2025).

ECONOMICS

56%

Double majors have 56% protection against earnings shocks vs. single majors.

Double majors experience 56% protection against earnings shocks, robust across model specifications and datasets (NBER WP 32095, 2024).

Unrelated field double majors gain 64% protection vs. 36% for related majors compared to single majors (NBER WP 32095, 2024).

In ACS data (2009-2019), 10% of 145,826 participants had double majors; 70% in unrelated fields (NBER WP 32095, 2024).

ARTIFICIAL INTELLIGENCE

89.4%

GPT-5 pro scores highest on PhD-level science questions.

GPT-5 pro (with tools) achieves 89.4% on GPQA Diamond, PhD-level science questions (2026).

GPT-5 pro (with tools/reasoning) scores 42% on Humanity’s Last Exam expert-level questions (2026).

GPT-5-high ranked better than or on par with industry experts 40.6% of the time across jobs (2025).

See these connections in the Knowledge Graph →

Enjoyed this?

Get new stories in your inbox.