Claim ledger

AI Can Take the Practice Without Taking the Job

The practice that builds judgement can disappear while every junior keeps their seat. This audit checks whether it has on your team.

Read the essay Published Tests Law IV: Instruments Over Theory

Claims
67
Verified at the primary source
63
Removed before publication
0
Last checked

67 claims: 63 verified, 4 executed.

Download this ledger: CSV · JSON · Markdown · CC BY 4.0

What the essay claims

Stanford's payroll data puts employment of 22 to 25 year olds in AI-exposed jobs 19% below where it would be had it kept pace with their peers, and the gap comes through hiring, sitting where AI usage substitutes for the work. Experienced workers show no gap, and in jobs that run on tacit knowledge they are growing. Stanford defines tacit knowledge as knowledge acquired through practice, mentorship and repeated exposure to real situations. For most knowledge workers, those situations were the everyday junior tasks AI now does first. So the count everyone is watching measures the visible half. A model of how expertise passes between generations (Ide, 2026) shows the other half: automating entry-level work can slow the spread of expertise even when no entry-level job is lost, because the best experts stop needing anyone beside them. Robotic surgery already ran this experiment: residents kept their places, the robot let the expert operate alone, and only a minority came to competence (Beane, 2019). The reader cannot fix the labour market. They can count the reps on their own team.

The claim ladder

RungClaimEvidence (Object Record IDs; ledger at Stage 2)Whose behaviourWhat it does NOT reach
R1Employment of 22 to 25s in highly AI-exposed jobs is about 19% below where it would be had it kept pace; the same measure read 15% a year earlierC1, C2, C4US workers in ADP payroll dataCausation; non-US workers; ADP gaps run larger than national surveys (C12)
R2The gap comes through hiring, not separations; experienced workers show no comparable gapC5, C6SameWhy firms hire less; whether it lasts
R3Declines sit where AI usage substitutes; the automation link is strongest for the youngest and shrinks with ageC7, C8Occupations scored by Anthropic Economic Index usageAny individual's usage; whether usage causes the decline
R4Jobs that run on tacit knowledge show faster growth for experienced workers, and this survives an education control; the codified result does notC9, C10Occupations scored by a language model on ADP job titlesIndividuals; causation
R5Tacit knowledge, by Stanford's definition, is acquired through practice, mentorship and repeated exposure to real situationsC13A definition, not a findingWhether AI-assisted exposure counts (open; see objection 2)
R6Automating entry-level work can slow the spread of expertise even with no fall in entry-level employment, by moving novices away from the best expertsI1, I2Agents in a theoretical modelAny measured economy
R7Where a technology let experts work alone, trainees kept their places, but only a minority came to competence, and the supply of experts fell relative to demandB1–B3Surgical trainees at five sites plus 13 teaching hospitals, robotic surgery, before generative AIKnowledge work; generative AI. For knowledge work, link our Learn with AI piece (Anthropic's programmer trial) in one line, not as core evidence
R8The Reps Audit gives repeatable calls and separates a junior from a senior task listObject Record § What was run (locked test)Claude Sonnet on two O*NET listsCorrectness for any one reader; other models; a reader's own week

The evidence, row by row

Each row is a claim the essay makes, then the evidence recorded for it and the source it was checked against. For verified rows that is usually the exact words, where they sit and the day they were read. The label says how far it was checked.

What the labels mean
Verified
Checked against the primary source itself by a checker who did not draft the piece, usually with the exact words, where they sit and the date read.
Executed
A run or observation done for the piece: code, a query, a count, or a check made live in an app. The claim is what it returned. Run records are not published.
Checked
Checked against its source by a separate checker in the older ledger format, which recorded no quote, locator or date.
Reported
Not verified word for word against a primary source: carried from a secondary source, from one that could not be opened in full, or supported only in part. The essay words it accordingly.
Struck
Drafted, checked and removed before publication. Kept here, crossed out, with the reason.
Excluded
Considered and deliberately left out of the piece. Kept here with the reason.
Sources (9)
  1. Stanford DEL news post 6 rows
  2. Brynjolfsson 28 rows
  3. Ide 10 rows
  4. Beane 6 rows
  5. UC Santa Barbara news 5 rows
  6. Shen and Tamkin 4 rows
  7. O*NET OnLine 3 rows
  8. First-party runs 4 rows
  9. The Durability Curve 1 row
  1. 1 Verified Read

    L1 Employment of 22 to 25s in highly AI-exposed jobs is 19% below where it would be had it kept pace with less-exposed peers (June 2026)

    "Employment among workers ages 22–25 in highly AI-exposed occupations now stands about 19% below where it would be if it had kept pace with employment among similarly aged workers in less-exposed occupations."

    Stanford DEL news post, 2026-08-12 (same figure in the paper, p.3: "has since widened to 19% as of June 2026") · section 'Our six main facts', second fact; paper p.3 · digitaleconomy.stanford.edu

    Link
    Cite

    Stanford DEL news post, 2026-08-12 (same figure in the paper, p.3: "has since widened to 19% as of June 2026"), section 'Our six main facts', second fact; paper p.3, https://digitaleconomy.stanford.edu/news/canariesaug26/. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 1, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-1 (source read 2026-10-01).

    Report an error
  2. 2 Verified Read

    L2 The same measure read 15% a year earlier (July 2025 vintage)

    "the kept-pace shortfall was 15% at the July 2025 data vintage"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 2, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-2 (source read 2026-10-01).

    Report an error
  3. 3 Verified Read

    L3 The 13% figure was an earlier regression estimate the study headlined, not the same measure (body: "The 13% you may have seen")

    "Earlier versions of this study headlined regression estimates adjusting for firm-level shocks (a 13% relative decline as of July 2025 data; 16% as of September 2025 data)."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 3, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-3 (source read 2026-10-01).

    Report an error
  4. 4 Verified Read

    L4 In levels, exposed 22 to 25s fell about 11% from Nov 2022 to June 2026 while the least-exposed grew about 10%

    "employment of 22–25-year-olds in the two most exposed quintiles fell about 11% between November 2022 and June 2026, while employment of the same age group in the three least-exposed quintiles grew about 10%."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 4, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-4 (source read 2026-10-01).

    Report an error
  5. 5 Verified Read

    L5 The data is ADP payroll records covering millions of US workers through June 2026

    "high-frequency administrative payroll data from ADP, encompassing millions of workers through June 2026."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.2 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.2, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 5, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-5 (source read 2026-10-01).

    Report an error
  6. 6 Verified Read

    L6 Experienced workers show no comparable gap

    "with no comparable gap for experienced workers"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3, fact (2); news post: 'Experienced workers show no comparable gap.' · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, fact (2); news post: 'Experienced workers show no comparable gap.', https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 6, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-6 (source read 2026-10-01).

    Report an error
  7. 7 Verified Read

    L7 The gap comes mainly through fewer hires of young workers, not more people leaving

    "it operates primarily through reduced hiring rather than increased separations"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3, fact (4) · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, fact (4), https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 7, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-7 (source read 2026-10-01).

    Report an error
  8. 8 Verified Read

    L8 Declines concentrate where AI usage substitutes for the work; where it complements, employment is flat or rising

    "declines are concentrated in occupations where AI usage substitutes for human tasks, while complementary usage is associated with flat or rising employment"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3, fact (5) · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, fact (5), https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 8, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-8 (source read 2026-10-01).

    Report an error
  9. 9 Verified Read

    L9 The automation link is strongest for the youngest (about -0.10 per SD for 22 to 25s) and shrinks with age

    "For the youngest workers, only the automation coefficient is negative and statistically significant—about −0.10 per standard deviation for 22–25 year-olds—and it shrinks monotonically with age."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.19; Table 3 p.20 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.19; Table 3 p.20, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 9, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-9 (source read 2026-10-01).

    Report an error
  10. 10 Verified Read

    L10 Jobs that rely on tacit knowledge show faster employment growth for experienced workers

    "Occupations that involve tacit knowledge see faster employment growth for experienced workers."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 10, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-10 (source read 2026-10-01).

    Report an error
  11. 11 Verified Read

    L11 The codified-knowledge result is not significant with an education control; the tacit result for experienced workers survives it

    "The codified-knowledge gradient is not significant when controlling for college share quintile, demonstrating its overlap with education. The tacit-knowledge gradient for experienced workers, by contrast, survives this control."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.21, footnote 21 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.21, footnote 21, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 11, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-11 (source read 2026-10-01).

    Report an error
  12. 12 Verified Read

    L12 The authors say the findings are descriptive, not a causal estimate of AI's effect

    "We caution that this work does not estimate a causal impact of AI: these are descriptive facts"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.4 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.4, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 12, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-12 (source read 2026-10-01).

    Report an error
  13. 13 Verified Read

    L13 The gap is larger in the ADP data than in national survey benchmarks

    "the divergence is directionally consistent but more pronounced in the ADP panel than in national survey benchmarks"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.4 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.4, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 13, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-13 (source read 2026-10-01).

    Report an error
  14. 14 Verified Read

    L14 The divergence holds when tech firms and computer jobs are excluded and with interest-rate and remote-work controls

    "The divergence remains when we exclude technology firms and computer occupations, when we control for exposure to interest-rate increases and remote work"

    Stanford DEL news post, 2026-08-12 · section 'Is AI causing these changes?' · digitaleconomy.stanford.edu

    Link
    Cite

    Stanford DEL news post, 2026-08-12, section 'Is AI causing these changes?', https://digitaleconomy.stanford.edu/news/canariesaug26/. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 14, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-14 (source read 2026-10-01).

    Report an error
  15. 15 Verified Read

    L15 The gap kept widening after interest rates peaked

    "The divergence has continued to widen through mid-2026, well after interest rates peaked."

    Stanford DEL news post, 2026-08-12 · section 'Is AI causing these changes?', first bullet · digitaleconomy.stanford.edu

    Link
    Cite

    Stanford DEL news post, 2026-08-12, section 'Is AI causing these changes?', first bullet, https://digitaleconomy.stanford.edu/news/canariesaug26/. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 15, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-15 (source read 2026-10-01).

    Report an error
  16. 16 Verified Read

    L16 Stanford's definition of tacit knowledge

    "implicit, experience-based knowledge that is difficult to write down or teach in a classroom. It is acquired through practice, mentorship, hands-on experience, and repeated exposure to real situations."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · Appendix J, pp.133-134 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, Appendix J, pp.133-134, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 16, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-16 (source read 2026-10-01).

    Report an error
  17. 17 Verified Read

    L17 Stanford's definition of codified knowledge

    "formal, standardized, documented knowledge that can be taught through education, textbooks, or written procedures. It is explicit and transferable through text."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · Appendix J, p.133 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, Appendix J, p.133, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 17, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-17 (source read 2026-10-01).

    Report an error
  18. 18 Verified Read

    L18 The authors' proposed mechanism: AI automates the checkable tasks that used to justify entry-level headcount

    "AI may be automating the checkable, process-intensive tasks that historically justified entry-level headcount, while increasing the leverage of experienced staff"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.5 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.5, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 18, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-18 (source read 2026-10-01).

    Report an error
  19. 19 Verified Read

    L19 Stanford's codified/tacit scores were produced by a language model reading job titles and descriptions

    "We query GPT-5.4-mini (OpenAI) with temperature = 0 to score each job independently on two dimensions"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · Appendix J, p.133 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, Appendix J, p.133, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 19, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-19 (source read 2026-10-01).

    Report an error
  20. 20 Verified Read

    L20 The automation-versus-complement split comes from Anthropic Economic Index usage data

    "complementarity (“augmentative”), and overall-usage exposures from the Anthropic Economic Index (Handa et al., 2025)"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · Table 3 caption, p.20 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, Table 3 caption, p.20, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 20, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-20 (source read 2026-10-01).

    Report an error
  21. 21 Verified Read

    L21 Ide: automating entry-level work can lower growth and welfare even with no fall in entry-level employment

    "improvements in entry-level automation increase output upon adoption but can reduce growth and welfare, even without reducing entry-level employment."

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08) · abstract, p.1 · arxiv.org

    Link
    Cite

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08), abstract, p.1, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 21, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-21 (source read 2026-10-01).

    Report an error
  22. 22 Verified Read

    L22 Ide's mechanism: novices get moved away from the most productive experts, so best practice spreads more slowly

    "This occurs when such improvements reallocate novices away from the most productive experts, slowing the diffusion of best practices."

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08) · abstract, p.1 · arxiv.org

    Link
    Cite

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08), abstract, p.1, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 22, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-22 (source read 2026-10-01).

    Report an error
  23. 23 Verified Read

    L23 Ide's premise: novices acquire tacit knowledge by working alongside experts

    "novices acquire tacit knowledge by working alongside experts"

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08) · abstract, p.1 · arxiv.org

    Link
    Cite

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08), abstract, p.1, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 23, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-23 (source read 2026-10-01).

    Report an error
  24. 24 Verified Read

    L24 Ide's result comes from a theoretical model, not measured data

    "I develop a task-based overlapping-generations model"

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08) · abstract, p.1 · arxiv.org

    Link
    Cite

    Ide, Automation, AI, and the Intergenerational Transmission of Knowledge, arXiv 2507.16078v9 (2026-06-08), abstract, p.1, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 24, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-24 (source read 2026-10-01).

    Report an error
  25. 25 Verified Read

    L25 Beane's study: a two-year ethnography at five sites plus interviews at 13 top US teaching hospitals

    "a two-year, five-sited, comparative ethnographic study of learning in robotic and traditional surgical practice, and a blinded interview-based study of surgical learning practices at 13 top-tier teaching hospitals around the U.S."

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09) · abstract (publisher record via Crossref) · doi.org

    Link
    Cite

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09), abstract (publisher record via Crossref), https://doi.org/10.1177/0001839217751692. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 25, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-25 (source read 2026-10-01).

    Report an error
  26. 26 Verified Read

    L26 In robotic surgery only a minority of trainees came to competence, through norm-challenging 'shadow learning'

    "allowed only a minority of robotic surgical trainees to come to competence"

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09) · abstract · doi.org

    Link
    Cite

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09), abstract, https://doi.org/10.1177/0001839217751692. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 26, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-26 (source read 2026-10-01).

    Report an error
  27. 27 Verified Read

    L27 Robotic surgery greatly limited the trainee's role in the work

    "the radically different practice of robotic surgery greatly limited trainees’ role in the work, making approved methods ineffective"

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09) · abstract · doi.org

    Link
    Cite

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09), abstract, https://doi.org/10.1177/0001839217751692. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 27, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-27 (source read 2026-10-01).

    Report an error
  28. 28 Verified Read

    L28 Outcome Beane names: a decreasing supply of experts relative to demand

    "a decreasing supply of experts relative to demand"

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09) · abstract, final sentence · doi.org

    Link
    Cite

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09), abstract, final sentence, https://doi.org/10.1177/0001839217751692. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 28, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-28 (source read 2026-10-01).

    Report an error
  29. 29 Verified Read

    L29 The three practices successful trainees used

    "“premature specialization” in robotic surgical technique at the expense of generalist training; “abstract rehearsal” before and during their surgical rotations when concrete, empirically faithful rehearsal was prized; and “undersupervised struggle,” in which they performed robotic surgical work close to the edge of their capacity with little expert supervision—when norms and policy dictated such supervision"

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09) · abstract · doi.org

    Link
    Cite

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09), abstract, https://doi.org/10.1177/0001839217751692. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 29, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-29 (source read 2026-10-01).

    Report an error
  30. 30 Verified Read

    L30 The scene: the robot let the expert operate alone; residents removed smoke and fluids or watched from a second console

    "The robotic surgical system allowed an expert to operate alone — the trainee was optional. Residents either held a laparoscopic tool to remove smoke and fluids from the patient, or watched the surgical action from a second trainee console."

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed · paragraphs 4-5 (Beane quoted) · news.ucsb.edu

    Link
    Cite

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed, paragraphs 4-5 (Beane quoted), https://news.ucsb.edu/2020/020026/shining-light-shadow-learning. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 30, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-30 (source read 2026-10-01).

    Report an error
  31. 31 Verified Read

    L31 Beane's own estimate, in an interview: residents get ten to twenty times less practice

    "“Residents get ten to twenty times less practice,” said Beane"

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed · paragraph 3 (interview estimate, not a figure from the ASQ paper; attribute as his estimate) · news.ucsb.edu

    Link
    Cite

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed, paragraph 3 (interview estimate, not a figure from the ASQ paper; attribute as his estimate), https://news.ucsb.edu/2020/020026/shining-light-shadow-learning. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 31, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-31 (source read 2026-10-01).

    Report an error
  32. 32 Verified Read

    L32 Anthropic's trial: programmers learning a new library with AI help learned less, and some ways of using it preserved learning

    "We find that AI use impairs conceptual understanding, code reading, and debugging abilities, without delivering significant efficiency gains on average."

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245 (2026-01-28) · abstract; second clause for the patterns: 'three of which involve cognitive engagement and preserve learning outcomes even when participants receive AI assistance' (used as a one-line link to our Learn with AI piece) · arxiv.org

    Link
    Cite

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245 (2026-01-28), abstract; second clause for the patterns: 'three of which involve cognitive engagement and preserve learning outcomes even when participants receive AI assistance' (used as a one-line link to our Learn with AI piece), https://arxiv.org/abs/2601.20245. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 32, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-32 (source read 2026-10-01).

    Report an error
  33. 33 Verified Read

    L33 The test task lists are O*NET's published task statements for Market Research Analysts and Marketing Specialists and for Marketing Managers, every task on each page

    "Market Research Analysts and Marketing Specialists"; "Marketing Managers"

    O*NET OnLine (US Department of Labor), details pages, updated 2026 · Tasks table of each saved details page (primaries/onet-13-1161.00.html, onet-11-2021.00.html; both pages carry 'Updated 2026'); every row of each Tasks table extracted to primaries/onet-*-tasks.txt (13 and 21 rows). · onetonline.org

    Link
    Cite

    O*NET OnLine (US Department of Labor), details pages, updated 2026, Tasks table of each saved details page (primaries/onet-13-1161.00.html, onet-11-2021.00.html; both pages carry 'Updated 2026'); every row of each Tasks table extracted to primaries/onet-*-tasks.txt (13 and 21 rows), https://www.onetonline.org/link/details/13-1161.00. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 33, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-33 (source read 2026-10-01).

    Report an error
  34. 34 Executed Read

    L34 Our test: with Stanford's definitions the sort gave identical calls on two reruns (0% flips) and separated the analyst list (54% codified) from the manager list (29%) by 25 points; the bare question flipped 5% to 15% of calls and agreed with the framework 95% of the time

    Result: "P1 separation (framework, A minus B): 25 points -> PASS" / "P2 stability (framework, worst list flip rate): 0% -> PASS" / "P3 framework vs control agreement: 95%"

    First-party run, pre-registered and locked (lock 6c2e9e32760b, HFloyd97/durability-engine#905)

    A run or observation done for this piece. The run record is not published.

    Link
    Cite

    A run or observation by The Durability Curve (First-party run, pre-registered and locked (lock 6c2e9e32760b, HFloyd97/durability-engine#905)). Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 34, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-34 (source read 2026-10-01).

    Report an error
  35. 35 Verified Read

    L35 Residents still finished their residencies, licensed to use the robot

    "They are finishing their residencies licensed to use robots in the operating room, but many are graduating without crucial surgical skills"

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed · paragraph 3 (Beane quoted) · news.ucsb.edu

    Link
    Cite

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed, paragraph 3 (Beane quoted), https://news.ucsb.edu/2020/020026/shining-light-shadow-learning. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 35, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-35 (source read 2026-10-01).

    Report an error
  36. 36 Verified Read

    L36 Open surgery needed the trainee because it took four hands

    "In traditional surgery, surgical trainees were literally required for the job because the open procedure required four hands and two people most of the time"

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed · paragraph 4 (Beane quoted) · news.ucsb.edu

    Link
    Cite

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed, paragraph 4 (Beane quoted), https://news.ucsb.edu/2020/020026/shining-light-shadow-learning. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 36, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-36 (source read 2026-10-01).

    Report an error
  37. 37 Verified Read

    L37 Some of the divergence predates generative AI

    "Some differential trends are visible before the widespread use of generative AI."

    Stanford DEL news post, 2026-08-12 · section 'Is AI causing these changes?', reasons for caution · digitaleconomy.stanford.edu

    Link
    Cite

    Stanford DEL news post, 2026-08-12, section 'Is AI causing these changes?', reasons for caution, https://digitaleconomy.stanford.edu/news/canariesaug26/. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 37, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-37 (source read 2026-10-01).

    Report an error
  38. 38 Verified Read

    L38 Ide is an economist at IESE Business School; this version is dated June 2026

    "Department of Economics, IESE Business School" / "June 9, 2026"

    Ide, arXiv 2507.16078v9 · title page, p.1 (author footnote; date line) · arxiv.org

    Link
    Cite

    Ide, arXiv 2507.16078v9, title page, p.1 (author footnote; date line), https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 38, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-38 (source read 2026-10-01).

    Report an error
  39. 39 Verified Read

    L39 O*NET's marketing manager task on pricing, and its market research analyst task on competitor prices

    "Develop pricing strategies, balancing firm objectives and customer satisfaction." / "Gather data on competitors and analyze their prices, sales, and method of marketing and distribution."

    O*NET OnLine (US Department of Labor), details pages, updated 2026 · Tasks tables (saved primaries/onet-*-tasks.txt) · onetonline.org

    Link
    Cite

    O*NET OnLine (US Department of Labor), details pages, updated 2026, Tasks tables (saved primaries/onet-*-tasks.txt), https://www.onetonline.org/link/details/11-2021.00. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 39, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-39 (source read 2026-10-01).

    Report an error
  40. 40 Executed Read

    L40 Our second test: the feeder-tracing prompt was stable (0.86 overlap) and found all 3 expected links, but gave feeders for only 53% of the manager's calls against a 70% bar, so the kit ships that step as a hand-filled worksheet

    Result: "F1 stability (mean Jaccard): 0.86 -> PASS" / "F2 coverage (feeder in both runs): 53% -> FAIL" / "F3 face validity: 3/3 expected pairs in both runs -> PASS"

    First-party run, pre-registered and locked (lock e33328c1c259, HFloyd97/durability-engine#906)

    A run or observation done for this piece. The run record is not published.

    Link
    Cite

    A run or observation by The Durability Curve (First-party run, pre-registered and locked (lock e33328c1c259, HFloyd97/durability-engine#906)). Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 40, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-40 (source read 2026-10-01).

    Report an error
  41. 41 Verified Read

    L41 O*NET's manager forecasting task and analyst trend-tracking task (worked example)

    "Use sales forecasting or strategic planning to ensure the sale and profitability of products, lines, or services" / "Forecast and track marketing and sales trends, analyzing collected data."

    O*NET OnLine (US Department of Labor), details pages, updated 2026 · Tasks tables · onetonline.org

    Link
    Cite

    O*NET OnLine (US Department of Labor), details pages, updated 2026, Tasks tables, https://www.onetonline.org/link/details/11-2021.00. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 41, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-41 (source read 2026-10-01).

    Report an error
  42. 42 Executed Read

    L42 In our feeder run, six of the manager's 15 tacit calls had no feeder in the analyst's work in either run (vendor negotiation, legal disputes, product display, three sustainability items); hiring and training staff got one in one run only

    Result: "No B task involves vendor negotiation or building distribution networks." / "No B task touches legal, copyright, or royalty issues." (rows A10 and A11, marked none)

    First-party run (lock e33328c1c259)

    A run or observation done for this piece. The run record is not published.

    Link
    Cite

    A run or observation by The Durability Curve (First-party run (lock e33328c1c259)). Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 42, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-42 (source read 2026-10-01).

    Report an error
  43. 43 Verified Read

    L43 Young workers' employment fell in occupations that involve codified knowledge

    "We find employment declines for young workers in occupations that involve codified knowledge."

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.3 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.3, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 43, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-43 (source read 2026-10-01).

    Report an error
  44. 44 Verified Read

    L44 In Ide's model the best experts pass their expertise to several novices at once

    "only the most knowledgeable experts manage production and transmit their expertise to multiple novices"

    Ide, arXiv 2507.16078v9 · abstract, p.1 · arxiv.org

    Link
    Cite

    Ide, arXiv 2507.16078v9, abstract, p.1, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 44, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-44 (source read 2026-10-01).

    Report an error
  45. 45 Verified Read

    L45 Anthropic's study was a randomised experiment on developers learning a new library

    "We conduct randomized experiments to study how developers gained mastery of a new asynchronous programming library with and without the assistance of AI."

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245 (2026-01-28) · abstract · arxiv.org

    Link
    Cite

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245 (2026-01-28), abstract, https://arxiv.org/abs/2601.20245. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 45, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-45 (source read 2026-10-01).

    Report an error
  46. 46 Verified Read

    L46 The authors' hypothesis: AI substitutes more effectively for codified knowledge

    "One possibility is that AI more effectively substitutes for knowledge that has been digitized and codified"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.5 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.5, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 46, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-46 (source read 2026-10-01).

    Report an error
  47. 47 Verified Read

    L47 Ide frames the concern: by letting senior workers do more on their own, AI-driven automation may cut entry-level opportunities and the labour-for-training exchange

    "By allowing senior workers to accomplish more tasks independently, AI-driven automation may reduce entry-level opportunities and threaten the implicit contract in which younger workers exchange their labor for training and the prospect of future promotion"

    Ide, arXiv 2507.16078v9 · p.2, Introduction, second paragraph (line-wrapped "se-nior" in extraction) · arxiv.org

    Link
    Cite

    Ide, arXiv 2507.16078v9, p.2, Introduction, second paragraph (line-wrapped "se-nior" in extraction), https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 47, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-47 (source read 2026-10-01).

    Report an error
  48. 48 Verified Read

    L48 In open surgery, trainees learned by doing more of each operation over time, using approved methods, and it worked

    "I found that learning surgery through increasing participation using approved methods worked well in traditional (open) surgery, as current literature would predict."

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09) · abstract (publisher record via Crossref) · doi.org

    Link
    Cite

    Beane, Shadow Learning, Administrative Science Quarterly 64(1):87-123 (online 2018-01-09), abstract (publisher record via Crossref), https://doi.org/10.1177/0001839217751692. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 48, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-48 (source read 2026-10-01).

    Report an error
  49. 49 Verified Read

    L49 Education is the one control that shrinks the estimates, which may be an alternative explanation or the channel AI works through

    "estimates attenuate when controlling for occupational education levels, which may reflect either an alternative explanation or the very channel through which AI operates"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · pp.3-4 (sentence crosses the page break; matches once the page number is removed); also p.22: 'Education is the one control that attenuates these estimates' · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, pp.3-4 (sentence crosses the page break; matches once the page number is removed); also p.22: 'Education is the one control that attenuates these estimates', https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 49, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-49 (source read 2026-10-01).

    Report an error
  50. 50 Verified Read

    L50 In occupations with a high share of college graduates the difference between more- and less-exposed jobs is muted

    "muted differences between more-exposed and less-exposed occupations compared to our main results"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · Section 4, p.25 (occupations with a high share of college graduates) · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, Section 4, p.25 (occupations with a high share of college graduates), https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 50, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-50 (source read 2026-10-01).

    Report an error
  51. 51 Verified Read

    L51 Ide's slowdown does not require lost practice or lost jobs; a worse pool of mentors is enough

    "this slowdown does not require an improvement in automation to reduce entry-level employment or deprive novices of hands-on practice; a compositional deterioration in the pool of mentors is sufficient"

    Ide, arXiv 2507.16078v9 · Introduction, p.4 · arxiv.org

    Link
    Cite

    Ide, arXiv 2507.16078v9, Introduction, p.4, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 51, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-51 (source read 2026-10-01).

    Report an error
  52. 52 Verified Read

    L52 Ide's own policy conclusion: look beyond entry-level headcount, because welfare can fall with the number of entry-level jobs unchanged

    "policy assessment should look beyond aggregate entry-level employment: the deployment of new technologies may reduce welfare even if the total number of entry-level jobs remains unchanged, provided it reallocates novices away from the most skilled experts"

    Ide, arXiv 2507.16078v9 · Conclusion (line-end hyphenation 'wel-fare' in extraction; verbatim otherwise) · arxiv.org

    Link
    Cite

    Ide, arXiv 2507.16078v9, Conclusion (line-end hyphenation 'wel-fare' in extraction; verbatim otherwise), https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 52, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-52 (source read 2026-10-01).

    Report an error
  53. 53 Verified Read

    L53 Stanford's authors cite Ide for the entry-level mechanism

    "while increasing the leverage of experienced staff (Ide, 2025; Garicano and Rayo, 2025)"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.5 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.5, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 53, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-53 (source read 2026-10-01).

    Report an error
  54. 54 Verified Read

    L54 In a field study of 5,172 support agents, AI help improved less experienced workers' speed and quality and facilitated learning

    "using data from 5,172 customer support agents" / "Less experienced and lower-skilled workers improve both the speed and quality of their output" / "AI assistance facilitates worker learning"

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161) · abstract (all three phrases) · arxiv.org

    Link
    Cite

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161), abstract (all three phrases), https://arxiv.org/abs/2304.11771. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 54, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-54 (source read 2026-10-01).

    Report an error
  55. 55 Verified Read

    L55 Stanford treats experienced workers as a comparison group

    "consistent with the placebo role of experienced workers"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.12 · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.12, https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 55, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-55 (source read 2026-10-01).

    Report an error
  56. 56 Verified Read

    L56 Ide: technology that puts more novices beside the most productive experts strengthens knowledge transmission and raises growth

    "By contrast, technological improvements that increase the number of novices learning from the most productive experts strengthen knowledge transmission and raise growth."

    Ide, arXiv 2507.16078v9 · abstract, p.1 · arxiv.org

    Link
    Cite

    Ide, arXiv 2507.16078v9, abstract, p.1, https://arxiv.org/abs/2507.16078. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 56, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-56 (source read 2026-10-01).

    Report an error
  57. 57 Verified Read

    L57 Stanford measures tacit reliance partly through required experience and on-the-job training, and calls the codified/tacit estimates non-causal and purely suggestive

    "We proxy reliance on tacit knowledge with required work experience and on-the-job training" / "We caution that these estimates are non-causal and purely suggestive"

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL · p.19 (second sentence crosses the p.19/20 break) · digitaleconomy.stanford.edu

    Link
    Cite

    Brynjolfsson, Chandar, Chen, Canaries in the Coal Mine? (Aug 2026 revision), Stanford DEL, p.19 (second sentence crosses the p.19/20 break), https://digitaleconomy.stanford.edu/app/uploads/2026/08/Canaries_August2026.pdf. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 57, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-57 (source read 2026-10-01).

    Report an error
  58. 58 Verified Read

    L58 By November 2022 the exposed young workers' position had returned to roughly its pre-pandemic level, so the later decline is new

    "By November 2022, the relative position of exposed young workers had already returned to roughly its pre-pandemic level."

    Stanford DEL news post, 2026-08-12 · section 'Is AI causing these changes?', reasons AI may be playing a role · digitaleconomy.stanford.edu

    Link
    Cite

    Stanford DEL news post, 2026-08-12, section 'Is AI causing these changes?', reasons AI may be playing a role, https://digitaleconomy.stanford.edu/news/canariesaug26/. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 58, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-58 (source read 2026-10-01).

    Report an error
  59. 59 Verified Read

    L59 Beane: residents lacked skills because they got too little practice or mentorship; the console time they did get was lower quality because surgeons 'helicopter taught' and retook control

    "because they did not receive enough practice or mentorship" / "surgeons “helicopter taught” — giving frequent and very public feedback to residents at the console and periodically retaking control of the robot from them"

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed · paragraphs 3 and 5 · news.ucsb.edu

    Link
    Cite

    UC Santa Barbara news, 'Shining a Light on Shadow Learning' (2020-09-15), Beane interviewed, paragraphs 3 and 5, https://news.ucsb.edu/2020/020026/shining-light-shadow-learning. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 59, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-59 (source read 2026-10-01).

    Report an error
  60. 60 Executed Read

    L60 Our sort marked the manager's pricing task tacit in both reruns

    Result: "Balancing objectives and satisfaction requires nuanced judgment from experience."

    First-party run (lock 6c2e9e32760b)

    A run or observation done for this piece. The run record is not published.

    Link
    Cite

    A run or observation by The Durability Curve (First-party run (lock 6c2e9e32760b)). Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 60, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-60 (source read 2026-10-01).

    Report an error
  61. 61 Verified Read

    L61 Of the 5,172 agents in the data, 1,636 are observed after getting the AI assistant

    "In total, we observe the conversation text and outcomes associated with 3 million chats by 5,172 agents. Within this, we observe 1.2 million chats by 1,636 agents in the post-AI period."

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161) · Section 3 (data); Table 1 'Treated, Post' agents 1,636 (full PDF, primaries/blr-2304.11771.pdf) · arxiv.org

    Link
    Cite

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161), Section 3 (data); Table 1 'Treated, Post' agents 1,636 (full PDF, primaries/blr-2304.11771.pdf), https://arxiv.org/abs/2304.11771. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 61, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-61 (source read 2026-10-01).

    Report an error
  62. 62 Verified Read

    L62 The learning result is stated as evidence, not a firm finding

    "We also find evidence that AI assistance facilitates worker learning"

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161) · abstract; full paper: the durable-learning evidence uses outage periods and the authors call those estimates noisy (Section 5) · arxiv.org

    Link
    Cite

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161), abstract; full paper: the durable-learning evidence uses outage periods and the authors call those estimates noisy (Section 5), https://arxiv.org/abs/2304.11771. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 62, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-62 (source read 2026-10-01).

    Report an error
  63. 63 Verified Read

    L63 The trial had 52 participants and was run through the Anthropic Fellows Program

    "In our main study, 52 participants completed the task, 26 for each of the control and treatment groups." / "Work done as a part of the Anthropic Fellows Program"

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245 · Section 4 (study design); title-page footnote (full PDF, primaries/anthropic-2601.20245.pdf) · arxiv.org

    Link
    Cite

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245, Section 4 (study design); title-page footnote (full PDF, primaries/anthropic-2601.20245.pdf), https://arxiv.org/abs/2601.20245. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 63, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-63 (source read 2026-10-01).

    Report an error
  64. 64 Verified Read

    L64 Estimates adjusting for firms' overall hiring point the same way but shift with modelling choices

    "estimates that account for overall changes in firm hiring are directionally consistent while becoming more sensitive to specification choices"

    Stanford DEL news post, 2026-08-12 · section 'Is AI causing these changes?', reasons for caution · digitaleconomy.stanford.edu

    Link
    Cite

    Stanford DEL news post, 2026-08-12, section 'Is AI causing these changes?', reasons for caution, https://digitaleconomy.stanford.edu/news/canariesaug26/. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 64, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-64 (source read 2026-10-01).

    Report an error
  65. 65 Verified Read

    L65 Our earlier piece on reps for individual judgement

    "It runs on reps, and every choice you hand over before trying it yourself is a rep you do not take."

    The Durability Curve, You Reach Before You Think (2026-08-12) · body, section before 'Why the reps are the thing' · harryfloyd.substack.com

    Link
    Cite

    The Durability Curve, You Reach Before You Think (2026-08-12), body, section before 'Why the reps are the thing', https://harryfloyd.substack.com/p/you-reach-before-you-think. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 65, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-65 (source read 2026-10-01).

    Report an error
  66. 66 Verified Read

    L66 The support-study authors suggest the assistant may capture and spread the behaviour of the most productive agents

    "This finding suggests that generative AI systems may be capable of capturing and disseminating the behaviors of the most productive agents."

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161) · Introduction (full PDF, primaries/blr-2304.11771.pdf) · arxiv.org

    Link
    Cite

    Brynjolfsson, Li, Raymond, Generative AI at Work, arXiv 2304.11771 (NBER w31161), Introduction (full PDF, primaries/blr-2304.11771.pdf), https://arxiv.org/abs/2304.11771. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 66, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-66 (source read 2026-10-01).

    Report an error
  67. 67 Verified Read

    L67 In the trial, participants who fully delegated the coding got some productivity gain at the cost of learning the library

    "Participants who fully delegated coding tasks showed some productivity improvements, but at the cost of learning the library."

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245 · abstract · arxiv.org

    Link
    Cite

    Shen and Tamkin (Anthropic), How AI Impacts Skill Formation, arXiv 2601.20245, abstract, https://arxiv.org/abs/2601.20245. Checked in: Claim Ledger, "AI Can Take the Practice Without Taking the Job", The Durability Curve, row 67, https://durabilitycurve.com/claims/the-reps-nobody-counts/#c-67 (source read 2026-10-01).

    Report an error

Cite and reuse

A single claim
Use the row's Cite button. It names the original source first, then this row, which carries its own link.
This ledger
Floyd, Harry (2026). Claim Ledger: AI Can Take the Practice Without Taking the Job. The Durability Curve. https://durabilitycurve.com/claims/the-reps-nobody-counts/
The essay
Floyd, Harry (2026). AI Can Take the Practice Without Taking the Job. The Durability Curve. https://durabilitycurve.com/blog/the-reps-nobody-counts/

The ledger data is licensed CC BY 4.0: reuse it, quote it or train on it, with credit to The Durability Curve and a link. Read the licence. Machine copies: JSON, CSV, Markdown.