{"targetId":"ai-entry-and-worker-gains","coveredDimensionIds":["teams-careers"],"status":"current","revisionId":"2605d750764ae7e19980b4d1a22313f8d6d4393a8c98a05f7bd47c122ed44be8","version":17,"judgment":"unchanged","reason":"The invalidation trigger came from a republished reading of the same work, not from a withdrawal or correction of its content. The textHash is identical and the adopted excerpts remain verbatim in the supplied text, so the existing dependency stays valid and nothing needs reconsidering. The date correction to an original version of 20 September 2026 matches the dating already used in the adopted association. It changes no horizon reasoning, because the 2027 horizon rests on Stanford's payroll pattern through June 2026 and Anthropic's October report, not on this paper's date.\n\nThe guide's judgment therefore holds as previously assessed. Entry into AI-exposed digital occupations is expected to stay difficult through 2027. The adopted Chandar and Klein Teeselink evidence supports AI attribution of a junior-share decline among adopters relative to controls. It does so mainly through senior growth, which is a mix shift rather than evidence of absolute recruitment cuts. Trigger 2 remains unmet.\n\nThe arrival is a duplicate, so it cannot be adopted and does not add corroborating weight.","updatedAt":"2026-11-21T11:42:17.106Z","history":"/guide-maintenance/ai-entry-and-worker-gains/history.json","authorship":{"provider":"anthropic","model":"claude-opus-5-5","promptVersion":"guide-maintain-v4","receiptKey":"bfr:edit:647df9aa565055c76a6747680828565ac0be1f937e4d2b4bb3dfacab637cf394"},"evidence":[{"id":"19dcbdd22144e5b0701948579ae759c1202bcfab9465551ed0497ae092a5c3f9","source":{"id":"f8e3da56d4c4f4d261a6ddae370716412a0c5b0b9957bdad7af8676fb14f0f8e","workId":"b9fc2110c50811a4e04b654e7a53acaa578cd7d54d7535db508b1957d9f49521","publicationId":"696eb8c4-5e03-454d-8141-a97a34b21ecd","lane":"live","textHash":"b389a69a41d9617b0a10d8e481b79eeeb337712bba5393ee4c9194b9e8c8650c","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-10-29T23:27:54+00:00","precision":"instant","basis":"publisher_page"},"availableBasis":"retained-version-observation","availableAt":"2026-10-29T23:31:35.753Z","observedAt":"2026-10-29T23:30:30.776Z","authoredAt":"2026-10-29T23:31:58.763Z","publishedAt":"2026-10-29T23:32:20.958Z","indexedAt":"2026-10-31T19:53:59.980Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"AI, Mathematics, Education, and a public goods problem","url":"https://terrytao.wordpress.com/2026/10/29/ai-mathematics-education-and-a-public-goods-problem/"},"association":{"decision":"linked","relationship":"direct","role":"supports","explanation":"Gireeja Ranade's guest post on Terence Tao's blog (29 Oct 2026) is an informed argument by an EECS professor. It is not a measurement. It states the mechanism behind careers-hollow-learning: entry-level jobs and PhD programs are paid training, so removing them without replacement removes a route to expertise. That bears on condition 0, employers removing entry tasks without another route to learn. Her warning that superficial 'sanity checks' erode understanding bears on condition 1, people being rewarded for producing AI output rather than understanding it. Both are framed as conditional concerns ('If these are taken away by AI'). The post reports no observation that entry roles are disappearing. Her one anecdote, that undergraduate researchers increasingly go straight to industry, concerns career preference, not lost openings. Her note narrows her job concern to highly trained scientists, mathematicians and humanists, not exposed digital occupations generally. The retrieval token 'hiring' matched her call for universities to commit to faculty hiring, which is not labour-market evidence. EM does not adopt her further premise (note 7) that AI augmentation cannot generate human understanding. The guide's apprenticeship path and Mollick's 'how you use AI' position treat that as an open, use-dependent question.","targetIds":["careers-hollow-learning","careers-hollow-learning:condition:0","careers-hollow-learning:condition:1"],"spans":[{"sourceId":"f8e3da56d4c4f4d261a6ddae370716412a0c5b0b9957bdad7af8676fb14f0f8e","excerpt":"Entry-level jobs in fact used to train people while also getting useful contributions from them. Entry-level jobs, or PhD programs, are like being paid to go to the gym. If these are taken away by AI, how do we train people?"},{"sourceId":"f8e3da56d4c4f4d261a6ddae370716412a0c5b0b9957bdad7af8676fb14f0f8e","excerpt":"Sanity checks themselves may become illusions — because we will be out of practice at true understanding."},{"sourceId":"f8e3da56d4c4f4d261a6ddae370716412a0c5b0b9957bdad7af8676fb14f0f8e","excerpt":"For now I focus on the narrow question of jobs for highly-trained scientists/mathematicians/humanists."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:b389a69a41d9617b0a10d8e481b79eeeb337712bba5393ee4c9194b9e8c8650c","duplicateOf":null},"references":[{"id":"f8e3da56d4c4f4d261a6ddae370716412a0c5b0b9957bdad7af8676fb14f0f8e","title":"AI, Mathematics, Education, and a public goods problem","url":"https://terrytao.wordpress.com/2026/10/29/ai-mathematics-education-and-a-public-goods-problem/"}]},{"id":"cce467ed3a7e5038a8b74407fb3a917823fd1a395eb270ec7c99bc7ef3574fdf","source":{"id":"3b7cf046ca191b817e8434c81f3ac02512693b730bf6a94c09ee22302f7a1569","workId":"5a87d21f668a098c8ece6b7237dd6d8bcbbe00e44f07ae1e194ac49faf76b77a","publicationId":"705b9613-eb72-b6fd-08ae-4be5b151653b","lane":"bridge","textHash":"0aa998fb3c20dc13fdca2a29ea3ed39199b77ed9fa564ea9c5e0cf3684d4f370","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-09-29","precision":"day","basis":"backfill:publisher_archive_date"},"availableBasis":"retained-version-observation","availableAt":"2026-10-30T23:23:47.256Z","observedAt":"2026-10-30T23:22:07.539Z","authoredAt":"2026-11-04T17:58:38.896Z","publishedAt":"2026-11-05T12:43:26.629Z","indexedAt":"2026-11-17T15:56:34.050Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"August 2026 JOLTS Report: Little Change, Limited Dynamism","url":"https://hiringlab.indeed.com/2026/09/29/august-2026-jolts-report/"},"association":{"decision":"linked","relationship":"indirect","role":"limits","explanation":"Cory Stahle’s Indeed Hiring Lab analysis (29 Sept 2026) of the August 2026 JOLTS release describes the whole US labour market. It does not measure AI, occupations or AI exposure. Two of its observations matter here. Openings have risen since late 2025 while hiring stays low, and Stahle argues this is frustrating especially for new graduates and other entrants. The Federal Reserve also raised rates in September 2026, and Stahle says higher borrowing costs bear on hiring plans. The retrieval token was 'hiring'; the trailing carousel teasers, including the AI wage and AI exposure titles, are page furniture and were not used. The article bears on the main question argument and on careers-entry-pressure through the guide’s named rival: 'Education mix, interest rates and earlier labor-market trends explain more of the apparent AI pattern than automation does.' It shows that much of the absolute difficulty of entering work in late 2026 has a macro, low-hire/low-fire explanation that applies across occupations. It does not challenge the relative finding that carries the guide’s AI read. Stanford compares young workers in exposed occupations with less-exposed peers in the same macro environment, and the authors themselves call that comparison descriptive. The effect is to limit attribution, not to reverse the read. Our forecast that entry stays difficult is partly macro, but evidence that AI changes access must come from the exposed-versus-less-exposed gap. Stahle’s note that postings turned positive year over year is an aggregate intention signal. It is not entry hiring recovering in exposed occupations, so it does not meet the revision trigger.","targetIds":["ai-entry-and-worker-gains","careers-entry-pressure"],"spans":[{"sourceId":"3b7cf046ca191b817e8434c81f3ac02512693b730bf6a94c09ee22302f7a1569","excerpt":"Job openings have trended up since late 2025, but hiring continues to move sideways at low levels."},{"sourceId":"3b7cf046ca191b817e8434c81f3ac02512693b730bf6a94c09ee22302f7a1569","excerpt":"It’s a frustrating one if you’re trying to find one, especially for new graduates and others entering the workforce."},{"sourceId":"3b7cf046ca191b817e8434c81f3ac02512693b730bf6a94c09ee22302f7a1569","excerpt":"gave the Federal Reserve cover to raise interest rates earlier this month, its first hike in more than three years."},{"sourceId":"stanford-canaries-2026","excerpt":"early, descriptive indicators"},{"sourceId":"stanford-canaries-2026","excerpt":"reduced hiring of young workers rather than increased separations"}],"mechanism":"Economy-wide low hiring and a September 2026 rate rise reduce openings for all entrants. That raises the baseline difficulty of entry in exposed and less-exposed occupations alike. Absolute entry difficulty therefore cannot by itself be attributed to AI. The AI-specific claim rests on the relative gap in Stanford’s descriptive payroll comparison.","assumptions":["National, all-occupation hiring conditions and the rate rise also affect entry hiring in AI-exposed digital occupations.","Stahle’s statement about new graduates is his interpretation of aggregate data, not a measured entrant-specific series.","Macro shocks affect exposed and less-exposed occupations similarly enough that Stanford’s relative comparison is not simply a macro artifact. This remains untested here."],"intermediateSourceIds":["stanford-canaries-2026"],"familyId":"text:0aa998fb3c20dc13fdca2a29ea3ed39199b77ed9fa564ea9c5e0cf3684d4f370","duplicateOf":null},"references":[{"id":"stanford-canaries-2026","title":"Canaries in the Coal Mine? — August 2026 revision","url":"https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/"},{"id":"3b7cf046ca191b817e8434c81f3ac02512693b730bf6a94c09ee22302f7a1569","title":"August 2026 JOLTS Report: Little Change, Limited Dynamism","url":"https://hiringlab.indeed.com/2026/09/29/august-2026-jolts-report/"}]},{"id":"8622859076db3823fca44010ec703539264f7bc601eb45192a0075e8de372858","source":{"id":"22040c8105bff1641ed67b1335192d1d8f044a940b257397337e1362a57877c7","workId":"ef0c416e029a77db6a0e63fc50e9a291bf596940ee11daed9e395b824690efb6","publicationId":"fe7e2037-a24c-2363-4754-7bf1cdeba3be","lane":"bridge","textHash":"fab8779732a176b5e4223b14bcb69154199c5dd5f364047e31669a9574c72e54","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-10-01T06:14:51.000Z","precision":"instant","basis":"backfill:feed_published"},"availableBasis":"retained-version-observation","availableAt":"2026-10-23T21:07:36.027Z","observedAt":"2026-10-18T02:39:32.303Z","authoredAt":"2026-10-29T16:31:06.150Z","publishedAt":"2026-10-29T16:33:57.264Z","indexedAt":"2026-10-31T20:34:19.113Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Roundup #89: It isn’t X, it’s Y","url":"https://www.noahpinion.blog/p/roundup-89-it-isnt-x-its-y"},"association":{"decision":"linked","relationship":"direct","role":"limits","explanation":"Retrieval came from the tokens 'wages' and 'payroll'. Most of Noah Smith's 1 October 2026 roundup (AI risk, minerals, rent, deportation, China, cyber) and the embedded Substack and a16z teasers have no bearing on this guide. Only section 4, 'AI and jobs', is relevant, and it reports other people's work rather than Smith's own measurement. The substantive item is Smith's quotation of a new Fairlie and Wu paper using national CPS microdata. It finds that recent college graduates' unemployment did not spike in summer 2026 and did not rise significantly relative to older graduates or to young workers without a degree, including under a broader 'sidelined unemployed' definition. This bears directly on the main question's population of young entrants, but on a different outcome: unemployment of all recent graduates, not hiring into AI-exposed digital occupations. It therefore limits the read rather than challenging it. It supports assumption 0: the outlook should not be generalized to every young worker or to a graduate unemployment crisis. It also bears on trigger 2, because it is the kind of broader national, education-aware data that trigger names. It does not meet that trigger, because it does not measure the exposed-versus-less-exposed employment gap that Stanford reports. Graduates could avoid unemployment by entering less-exposed work while entry into exposed occupations stays weak. The Fairlie and Wu result is also consistent with Stanford's own finding of no widespread displacement. Two other items are context only and are not targeted. Indeed data reportedly shows faster wage growth in AI-exposed jobs; Smith reads this as demand. A rival reading is that wage averages rise as fewer juniors are hired. No age or entry breakdown is supplied to separate the two. Smith also judges, from a Politano chart we cannot inspect, that digital media jobs are being 'clobbered'. That claim concerns occupational employment, not entry routes, and the attribution to AI is his. Our scope remains EM's: this is a secondhand relay of the paper's abstract, so methods and exposure definitions are not available here.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:assumption:0","ai-entry-and-worker-gains:trigger:2"],"spans":[{"sourceId":"22040c8105bff1641ed67b1335192d1d8f044a940b257397337e1362a57877c7","excerpt":"unemployment rates did not spike in summer 2026 relative to summer months in previous years and did not rise in a significant way relative to older college graduates or young workers without a college degree."},{"sourceId":"22040c8105bff1641ed67b1335192d1d8f044a940b257397337e1362a57877c7","excerpt":"we find no evidence of a statistically significant increase in summer 2026 even after adding these “sidelined unemployed.”"},{"sourceId":"22040c8105bff1641ed67b1335192d1d8f044a940b257397337e1362a57877c7","excerpt":"workers in AI-exposed jobs have seen their wages increase much faster than workers in less-exposed jobs"},{"sourceId":"22040c8105bff1641ed67b1335192d1d8f044a940b257397337e1362a57877c7","excerpt":"digital media jobs are getting absolutely clobbered (even as live performing art jobs have held up fine)"}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:fab8779732a176b5e4223b14bcb69154199c5dd5f364047e31669a9574c72e54","duplicateOf":null},"references":[{"id":"22040c8105bff1641ed67b1335192d1d8f044a940b257397337e1362a57877c7","title":"Roundup #89: It isn’t X, it’s Y","url":"https://www.noahpinion.blog/p/roundup-89-it-isnt-x-its-y"}]},{"id":"28d81cfeac86ab90dd0dafa7da7f3a14aa9fdacb13d8a8a902663e34458d735c","source":{"id":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","workId":"81b30abd3e76ec28de16de3b6181c2d63b27ea50596f884374d386da12e7a564","publicationId":"3f55e76f-86da-648a-47e3-2823d236234e","lane":"bridge","textHash":"d6df8802a0e526368721b389df24f35c7173d001feb726775544dce24aea0206","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-07-24T08:30:00.000Z","precision":"instant","basis":"backfill:feed_published"},"availableBasis":"retained-version-observation","availableAt":"2026-10-24T01:25:36.700Z","observedAt":"2026-10-24T01:23:39.733Z","authoredAt":"2026-10-29T17:40:36.569Z","publishedAt":"2026-10-29T17:43:22.944Z","indexedAt":"2026-10-31T21:09:00.829Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Agents' summary of business conditions - July 2026","url":"https://www.bankofengland.co.uk/agents-summary/2026/july-2026"},"association":{"decision":"linked","relationship":"direct","role":"supports","explanation":"The Bank of England Agents’ summary (published 24 July 2026; Box A summarises Agents’ conversations with UK businesses over 2026) is employer-reported qualitative intelligence. It is not a payroll measurement or a causal estimate. Its AI box speaks directly to the guide’s outcome and population. The Agents report that AI-driven automation is reducing demand for some entry-level and junior roles that traditionally provided early-career experience. Several professional services contacts report reduced graduate recruitment, and some worry about skills development and talent pipelines. That supports the main read’s mechanism: a tool can raise the output of current staff while the employer recruits fewer beginners. It also bears on careers-entry-pressure condition 0 and entry-routes-narrow condition 0 (employers use capacity to reduce recruitment before replacing the training that entry work provided). It gives careers-hollow-learning its first employer-side report of the concern, though contacts express concern rather than observe a measured progression shortfall. The finding that gains are greatest where skilled employees validate and refine AI output, and that demand shifts toward judgement and oversight, supports assumption 1. The stated responses are specialist hiring and training for existing employees. That is not evidence of new paid entry routes, so it does not meet trigger 1. The same report also strengthens the guide’s macro rival. Flat employment intentions are attributed jointly to weak demand, Middle East-related caution, labour cost pressures and automation. Weak UK demand and high labour costs could explain much of the reduced recruitment without AI. The report does not separate these causes, quantify junior hiring, or compare exposed with less-exposed occupations. It therefore adds an independent, non-US employer-reported family to Stanford’s descriptive payroll pattern and Anthropic’s single-firm hiring mix. It does not settle attribution. Retrieval came from the token ‘hiring’ and the career-entry route; the trailing related-publication links are page furniture and were not used.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:assumption:1","careers-entry-pressure","careers-entry-pressure:condition:0","entry-routes-narrow:condition:0","careers-hollow-learning"],"spans":[{"sourceId":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","excerpt":"Automation is reducing demand for some entry-level and junior roles that have traditionally provided early-career experience, including tasks such as document preparation, invoice processing and basic analysis."},{"sourceId":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","excerpt":"Several professional services contacts report reduced graduate recruitment and lower demand for administrative and junior staff. Some have expressed concern about the long-term implications for skills development and talent pipelines."},{"sourceId":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","excerpt":"Productivity gains have been greatest where skilled employees are able to validate and refine AI-generated outputs."},{"sourceId":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","excerpt":"Employment intentions remain broadly flat, reflecting weak and/or uncertain demand (including a degree of Middle East (ME)-related caution), labour cost pressures and increased use of automation."},{"sourceId":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","excerpt":"Contacts are responding by recruiting specialist AI and data staff and by investing in training and redeployment programmes for existing employees."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:d6df8802a0e526368721b389df24f35c7173d001feb726775544dce24aea0206","duplicateOf":null},"references":[{"id":"3879adb75665d0a1d26d462e23749e18df1feec66f5e6464c5d5fce6d1e2311a","title":"Agents' summary of business conditions - July 2026","url":"https://www.bankofengland.co.uk/agents-summary/2026/july-2026"}]},{"id":"5ca1e2722f12cf4ef85735e383c82eaa65a1caccb466ee1b4e693d03e604347d","source":{"id":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","workId":"31600a8ceb52a83f8aa5dd75426ffd6ba753517638069817b442d5986c391a06","publicationId":"5a717d18-d44f-c924-6cf8-64016cd1901c","lane":"bridge","textHash":"a35f4653858a02fa48a9001148e323c4df710e869f8163a29e535e0cbc8f88ff","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-06-11T02:29:51.000Z","precision":"instant","basis":"backfill:feed_published"},"availableBasis":"retained-version-observation","availableAt":"2026-10-23T20:33:36.887Z","observedAt":"2026-10-23T20:31:49.340Z","authoredAt":"2026-10-29T18:09:13.093Z","publishedAt":"2026-10-30T07:48:56.471Z","indexedAt":"2026-10-31T21:10:10.364Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Why AI hasn’t replaced software engineers, and won’t","url":"https://www.normaltech.ai/p/why-ai-hasnt-replaced-software-engineers"},"association":{"decision":"linked","relationship":"direct","role":"limits","explanation":"Narayanan and Kapoor's essay (11 June 2026) on software engineering, the most AI-exposed digital occupation, is mainly an informed argument. It relays other studies secondhand, and their methods are not available here. Retrieval came from 'hiring' and the career-progression route. It bears on three targets.\n\n(1) Main question. The authors place AI's labour effect in hiring rather than layoffs. Their phrase 'slower hiring rather than increased separations' echoes the guide's Stanford finding. We treat it as a relay of that evidence family, not independent corroboration, because the cited study is not named in the excerpt. They also relay a Federal Reserve paper reporting that software engineer employment still grows, but about 3 percentage points per year slower than a no-AI counterfactual. That paper concerns total occupational employment, not entry, and the authors note it misses self-employment. They also say programming, the 'execute' layer, has been shrinking for decades. That gives the guide's prior-trends rival a concrete occupational form.\n\n(2) Assumption 1, that employers keep valuing independent judgment while automating novice tasks. The decide-execute-deliver argument supports it directly: specification, verification and accountability resist automation.\n\n(3) Condition 1 of entry-routes-narrow, that higher output does not create enough demand to restore entry hiring. The authors' price-elasticity argument, that cheaper software raises derived demand for engineers, is the strongest rival to this condition. It is a forecast about aggregate demand, not about entry. The authors themselves expect seniority to shape who gains or loses, and defer that question to a later essay.\n\nIn short, the essay limits the guide to an access-to-experience claim, not a collapse of the profession. That is the scope the guide already states.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:assumption:1","entry-routes-narrow:condition:1"],"spans":[{"sourceId":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","excerpt":"The research is clear that the effect operates through “slower hiring rather than increased separations”."},{"sourceId":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","excerpt":"Software engineer employment is still growing, but they find that it is growing slower post-ChatGPT compared to a no-AI counterfactual, by about 3 percentage points per year."},{"sourceId":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","excerpt":"Not only has programming been shrinking, it is also pays much less because it is seen as grunt work."},{"sourceId":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","excerpt":"The barrier isn’t actually learning the syntax. It’s having enough skilled judgment to make good decisions while maintaining accountability."},{"sourceId":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","excerpt":"And as we have argued, AI doesn’t replace software engineers (the “elasticity of substitution” is low), so the demand for more software results in a derived demand for more software engineers."},{"sourceId":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","excerpt":"The fact that aggregate labor demand in software is likely to remain strong doesn’t mean that most individual workers won’t be affected."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:a35f4653858a02fa48a9001148e323c4df710e869f8163a29e535e0cbc8f88ff","duplicateOf":null},"references":[{"id":"a825216b2058278a035277d9cec10ef4854bfb3e7e34f94877394b90a39d32ca","title":"Why AI hasn’t replaced software engineers, and won’t","url":"https://www.normaltech.ai/p/why-ai-hasnt-replaced-software-engineers"}]},{"id":"75ed32e98888e65bc26edf8bd7d3c7a3194550926966093ea8f9fad451a11500","source":{"id":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","workId":"0772e4d3ec5907fddec85a8a394e56a347c5d587356852d4d344a7b2547a7bd2","publicationId":"97346f01-cc83-7ec8-24b2-37fad655297e","lane":"bridge","textHash":"53345b7692f0e497862490704d42fb48bd80ce1fc3242004573fe64100b834e7","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-10-13T15:30:16.894Z","precision":"instant","basis":"backfill:article_published"},"availableBasis":"retained-version-observation","availableAt":"2026-11-01T20:19:20.295Z","observedAt":"2026-11-01T20:17:46.817Z","authoredAt":"2026-11-02T00:10:18.790Z","publishedAt":"2026-11-02T00:13:32.408Z","indexedAt":"2026-11-02T00:24:24.658Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Is AI Replacing Junior Bankers? What Hiring Data Shows","url":"https://www.reveliolabs.com/news/ai-and-work/ai-junior-bankers-hiring-data"},"association":{"decision":"linked","relationship":"direct","role":"limits","explanation":"Revelio Labs' newsletter (Galeano, Oliver and Simon, 13 October 2026) analyses Revelio workforce data on US banks, which it measures as more AI-exposed than the rest of the economy. It is observational headcount, hiring and posting data, not a causal estimate. Retrieval came from the tokens 'apprenticeship' and 'hiring' and the career-entry route. The match is substantive: the article reports on junior workers in an exposed white-collar sector and on how they progress to senior roles.\n\nWhat it observes. Junior bank headcount is 4.7% below its October 2022 level while senior roles keep growing. The drop comes from lower junior hiring, not higher attrition. That pattern matches the guide's hiring-first mechanism, as in Stanford's payroll finding, though in a different population and dataset.\n\nWhy it limits the read rather than supporting it. The authors find that exposure does not explain which junior roles shrink: the most and least AI-exposed junior roles moved similarly. They attribute the hiring dip partly to the 2022 interest-rate cycle, and the authors argue the decline looks more like a continuation of the long-running shift to digital banking than AI replacing juniors. That is a concrete sectoral instance of the guide's named rival, that interest rates and earlier trends explain more than automation.\n\nScope and dates. Revelio's 'junior' explicitly excludes entry level, so this is not a direct measure of first-job openings. Junior hiring is reported as recovering in 2026. That is relevant to trigger 0 but does not meet it: it is one sector, the category is past entry level, and the headcount remains below 2022.\n\nBearing on the paths. For entry-routes-narrow signpost 0, it shows a junior-versus-senior divergence in an exposed sector, though one not attributed to AI. For careers-hollow-learning, two pieces of evidence matter. Time spent in junior roles before promotion fell, by 11.3% for financial analysts. The authors read this as less time to build experience, and they warn that fewer junior hires means fewer bankers build the judgment senior roles need. Shorter junior tenure could equally mean faster progression, so it does not show the progression shortfall the path's signpost names. The mechanism concern is the authors' argument, not a measured outcome.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:trigger:0","entry-routes-narrow:signpost:0","careers-entry-pressure","careers-hollow-learning"],"spans":[{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"junior roles in US banks have been shrinking and are now 4.7% below their October 2022 level."},{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"Instead, junior hiring fell from about 10.5% of junior headcount a year before ChatGPT to around 8.5% in 2024 and 2025, before recovering in 2026. The timing also lines up with the 2022 interest rate cycle, which likely played a role in the slowdown."},{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"Still, exposure doesn't explain which junior roles are shrinking: headcount in the most and least AI-exposed junior roles has moved similarly since ChatGPT's launch."},{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"We define junior workers as early-career staff past entry level but not yet in senior or supervisory roles, such as bank analysts."},{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"This looks less like generative AI replacing junior bankers and more like a continuation of the long-running shift to digital banking."},{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"Since ChatGPT’s launch, that time has fallen most in Financial Analyst roles, where it has dropped by 11.3%, followed by Tech roles (-8.1%), compared with about 6% across all roles for the rest of the economy."},{"sourceId":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","excerpt":"That said, if banks hire fewer juniors, fewer bankers will build the judgment senior roles rely on."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:53345b7692f0e497862490704d42fb48bd80ce1fc3242004573fe64100b834e7","duplicateOf":null},"references":[{"id":"e42d2f8d1e9f2ea2829fdd9ce9b9462f7465334c6244475cfba43dd5ba78b1d8","title":"Is AI Replacing Junior Bankers? What Hiring Data Shows","url":"https://www.reveliolabs.com/news/ai-and-work/ai-junior-bankers-hiring-data"}]},{"id":"e241e59e9561a78f4bb9d4d580d88d4989ace6bc13b44103f466cafd553640e3","source":{"id":"dba73c7c2498167a78f37e80d1ae682e5e88f7a861389906ec3c660bb75a213d","workId":"738527928d13380fbe1d1c344c54577267ad5256ffcf361fe987dcd3fbd6eaa2","publicationId":"dfb8f7ea-9345-6c85-1fb8-ea41edc9df3c","lane":"bridge","textHash":"281ab43facd17f2dec7a4e1773379bf767b5dfc90a09878a82e4dbcbabb7a41a","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-08-24","precision":"day","basis":"backfill:publisher_archive_date"},"availableBasis":"retained-version-observation","availableAt":"2026-10-30T19:07:44.717Z","observedAt":"2026-10-30T19:06:05.159Z","authoredAt":"2026-11-02T00:19:01.289Z","publishedAt":"2026-11-02T00:22:35.277Z","indexedAt":"2026-11-02T00:25:56.515Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"US Labor Market Snapshot — August 2026","url":"https://hiringlab.indeed.com/2026/08/24/us-labor-market-snapshot-august-2026/"},"association":{"decision":"linked","relationship":"indirect","role":"limits","explanation":"Indeed Hiring Lab's US Labor Market Snapshot (24 August 2026) is an observational, all-occupation postings and labour-flow summary. It does not measure AI use, AI exposure, age or seniority. Retrieval came from the tokens 'hiring' and 'wages'. Most of the snapshot (Healthcare, wages, monetary policy) has no bearing on this guide. The trailing carousel teaser about an 'AI Reshuffling of White-Collar Work' is page furniture and was not used. The methodology describes an AI-postings series, but the snapshot reports no AI-postings figures.\n\nOne observation reaches a guide condition. Software Development, the clearest AI-exposed digital occupation in the guide's scope, had postings at 74.4, well below pre-pandemic levels, but up from a low of 61.1 in May 2025. That bears on trigger 0 ('Entry hiring recovers in exposed occupations while AI use continues to deepen') and on the careers-entry-pressure signpost. It does not meet either. The series counts postings at all levels, not entry hiring. A posting signals intention, not a hire. The snapshot also gives no comparison with less-exposed work at entry level.\n\nThe recovery limits how strongly the guide can describe exposed digital work as uniformly contracting. It also leaves open the guide's prior-trends rival, because the drop from the 2022 peak predates most generative AI deployment. The snapshot's low-hire, low-fire description and its view that labour supply may matter as much as demand add macro context. That context belongs with the already adopted Stahle JOLTS reading, which makes the same economy-wide point. It is a separate Hiring Lab publication and observation, not a duplicate, but its macro point should not be counted as independent corroboration.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:trigger:0","careers-entry-pressure:signpost"],"spans":[{"sourceId":"dba73c7c2498167a78f37e80d1ae682e5e88f7a861389906ec3c660bb75a213d","excerpt":"Postings on Indeed for Software Development roles (74.4), for instance, remain significantly below their pre-pandemic level. But, demand for these roles has rebounded over the past year or so, increasing from a low of 61.1 in May 2025."},{"sourceId":"dba73c7c2498167a78f37e80d1ae682e5e88f7a861389906ec3c660bb75a213d","excerpt":"The labor market remains stuck in the low-hire, low-fire environment that’s prevailed since last year, with the hiring rate (3.4%), quits rate (2%), and layoffs rate (1.1%) all subdued in June. This may not be purely a labor demand story."}],"mechanism":"Entry openings in software development are a subset of all software development postings. A sustained rebound in total postings is a prerequisite for entry hiring to recover in that exposed occupation. It is not sufficient evidence of that recovery, because employers can direct new openings to experienced hires.","assumptions":["All-level Software Development postings on Indeed are a meaningful upper-bound signal for entry openings in that occupation.","Postings convert to hires at a rate that does not differ sharply between exposed and less-exposed occupations; the snapshot does not report this.","Software Development is representative enough of the guide's 'exposed digital occupations' to bear on the trigger, though it is one occupational group."],"intermediateSourceIds":[],"familyId":"text:281ab43facd17f2dec7a4e1773379bf767b5dfc90a09878a82e4dbcbabb7a41a","duplicateOf":null},"references":[{"id":"dba73c7c2498167a78f37e80d1ae682e5e88f7a861389906ec3c660bb75a213d","title":"US Labor Market Snapshot — August 2026","url":"https://hiringlab.indeed.com/2026/08/24/us-labor-market-snapshot-august-2026/"}]},{"id":"d6c69c4581a809afc117c8c32c09170f7ff1e2812d289308aea835aad5a73db6","source":{"id":"d762516628eea6a0a5955869b98833ad113f07b552f7a77897e24a2836d74837","workId":"81d9426fce45026589cb9489106fdef1de11223ae5fe200d5c74dae31674579d","publicationId":"42f6d497-ce95-ad34-40ee-728383c4b14e","lane":"bridge","textHash":"7e6a9a96d0f9b02b417316dcc1fdb1b3baec585ae2f52b1ba645bac86a689d1f","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-09-02T12:30:10.175Z","precision":"instant","basis":"backfill:article_published"},"availableBasis":"retained-version-observation","availableAt":"2026-10-31T00:39:44.438Z","observedAt":"2026-10-31T00:37:52.566Z","authoredAt":"2026-11-02T03:33:46.644Z","publishedAt":"2026-11-02T03:43:50.420Z","indexedAt":"2026-11-02T03:54:18.027Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"August 2026 Jobs Report: US Economy Adds 36.5K Jobs","url":"https://www.reveliolabs.com/news/rpls/rpls-us-jobs-report-the-us-economy-adds-36-5k-jobs-in-august"},"association":{"decision":"linked","relationship":"direct","role":"supports","explanation":"Revelio Labs' RPLS August 2026 jobs report was published 2 September 2026 (Abdelwahed, Simon, Wheeler and Wilkie-Rogers). It is a short monthly summary of Revelio's AI Labor Market Tracker. The findings are observational and associational, and the methods are not given here.\n\nRetrieval came from the token 'hiring'. The match is substantive because the report measures the guide's population and comparison: young workers in the most versus least AI-exposed occupations. It uses Revelio workforce data, not ADP payroll records, so it is a distinct measurement rather than a relay of Stanford's study. It is not a duplicate. It still belongs to the same descriptive, exposure-gap kind of evidence.\n\n**Main question.** Revelio reports that since before ChatGPT, employment in the most exposed occupations is down about 6% relative to the least exposed, with a 19% gap at ages 22–25. That independently reproduces the age-concentrated relative shortfall that carries the guide's AI read. It is employment, not hiring, so it does not by itself show the hiring-first channel.\n\n**Trigger 2.** The trigger is that the gap shrinks after accounting for education, prior trends and broader national data. A second large dataset finds no shrinkage, which counts against the trigger. However, no education control or pre-trend test is reported, so the guide's education and prior-trends rival is not resolved.\n\n**careers-entry-pressure and entry-routes-narrow.** Two firm-level findings fit the mechanism that experienced teams gain while entry narrows:\n- The most exposed firms show fewer layoffs, not more.\n- AI-adopting firms expand employment with gains 'concentrated in senior roles'.\n\nRevelio itself flags selection: adopters were already growing faster before adoption. These are firm-level associations, not evidence that adoption caused junior hiring to fall.\n\n**Limit on trigger 0.** Revelio reports that new firm AI adoption has slowed from its spring peak. That bears on trigger 0's condition 'while AI use continues to deepen' and is worth watching. It concerns the pace of new adoption, not depth of use, so it does not change the horizon.\n\nRevelio's summary interpretation that AI affects 'which workers face weaker opportunities' is the authors' argument. EM's use of it to support the entry-route read is our inference.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:trigger:2","ai-entry-and-worker-gains:trigger:0","careers-entry-pressure","entry-routes-narrow"],"spans":[{"sourceId":"d762516628eea6a0a5955869b98833ad113f07b552f7a77897e24a2836d74837","excerpt":"Since before ChatGPT, employment in the most AI-exposed occupations is down around 6% relative to the least-exposed occupations, with the gap reaching 19% among workers aged 22–25."},{"sourceId":"d762516628eea6a0a5955869b98833ad113f07b552f7a77897e24a2836d74837","excerpt":"The most AI-exposed firms are seeing fewer layoffs than the least-exposed firms, rather than more."},{"sourceId":"d762516628eea6a0a5955869b98833ad113f07b552f7a77897e24a2836d74837","excerpt":"AI-adopting firms continue to expand employment relative to non-adopters, although adopters were already growing faster before adoption and the gains remain concentrated in senior roles."},{"sourceId":"d762516628eea6a0a5955869b98833ad113f07b552f7a77897e24a2836d74837","excerpt":"This month’s tracker also shows that the pace of new firm AI adoption has slowed from its spring peak"}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:7e6a9a96d0f9b02b417316dcc1fdb1b3baec585ae2f52b1ba645bac86a689d1f","duplicateOf":null},"references":[{"id":"d762516628eea6a0a5955869b98833ad113f07b552f7a77897e24a2836d74837","title":"August 2026 Jobs Report: US Economy Adds 36.5K Jobs","url":"https://www.reveliolabs.com/news/rpls/rpls-us-jobs-report-the-us-economy-adds-36-5k-jobs-in-august"}]},{"id":"1f5e2746d4e24ac227bf02ba5965250468ab237f3726a44a1cbbfe04066be94e","source":{"id":"6561838646a67f5fa2c5f9d6d5962a94155ce93a815545a92d24b1d26960e930","workId":"acb6e7a18a60461ee8f2b135e5fd0b585661b9011dd0afdd7cddfd2466a6093e","publicationId":"e109c560-8010-d912-f161-29441c2a1d1a","lane":"bridge","textHash":"035d1d60d875f91539b2e5147de6c354a84a354b375f9acae5dfe0d9d375e401","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-07-29","precision":"day","basis":"backfill:article_published"},"availableBasis":"retained-version-observation","availableAt":"2026-10-30T07:31:55.049Z","observedAt":"2026-10-25T12:31:38.945Z","authoredAt":"2026-11-03T00:16:28.407Z","publishedAt":"2026-11-03T00:17:40.453Z","indexedAt":"2026-11-03T00:24:34.806Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Unbundling jobs: Measuring the value of tasks in an AI economy","url":"https://www.adpresearch.com/research/unbundling-jobs-measuring-the-value-of-tasks-in-an-ai-economy"},"association":{"decision":"linked","relationship":"direct","role":"limits","explanation":"Retrieval came from the tokens 'wages' and 'payroll'. The match is substantive for one target only. This ADP Research and Stanford Digital Economy Lab note (29 July 2026; author shown as Andrew Wang) is observational. It reports a wage regression on task indicators drawn from job postings linked to ADP payroll records, covering about 7,000 workers in selected IT jobs from 2019 to 2025. It is a separate study from Stanford's 'Canaries' payroll paper and is not a duplicate. It measures task wage premia, not entry, hiring or progression, so it reaches none of the entry paths.\n\nIt bears on ai-entry-and-worker-gains assumption 1: employers keep valuing independent judgment while automating some tasks traditionally assigned to novices. Treating design, advising and directing tasks as proxies for 'independent judgment' is EM's mapping. In this sample those tasks carry wage premia, and routine maintenance and monitoring carry penalties. That is consistent with the assumption.\n\nThe pre/post-2022 comparison limits it. Among the tasks the authors list as declining in value are 'Develop models of systems, processes, or products' and 'Explain technical details of products or services', both of which are also on their higher-wage list. 'Diagnose system or equipment problems' also declined. The prose summary groups 'evaluation' with high-value work, but the lists put diagnosis and performance analysis among lower-wage tasks. So the evidence does not show that judgment-like work is uniformly gaining value. Some of it is being repriced.\n\nThree further limits apply. The design is correlational. The post-AI label is a period split, not an exposure or adoption comparison. Age is controlled for, so the note says nothing about which tasks novices perform. Our inference that routine, lower-paid tasks are the ones traditionally given to beginners would need separate evidence. We therefore do not link the note to careers-hollow-learning or entry-routes-narrow.","targetIds":["ai-entry-and-worker-gains:assumption:1"],"spans":[{"sourceId":"6561838646a67f5fa2c5f9d6d5962a94155ce93a815545a92d24b1d26960e930","excerpt":"high-value work is clustered around design and evaluation tasks, while more routine responsibilities such as maintenance and monitoring are associated with lower pay."},{"sourceId":"6561838646a67f5fa2c5f9d6d5962a94155ce93a815545a92d24b1d26960e930","excerpt":"When we compare pay in 2023-2025 (post-AI) to pay in 2019-2022 (pre-AI), we find the following tasks declined in value:"},{"sourceId":"6561838646a67f5fa2c5f9d6d5962a94155ce93a815545a92d24b1d26960e930","excerpt":"Using a regression of worker wages on task indicators and controlling for age, gender, year-specific effects, and company-specific characteristics, we estimate the wage premium or penalty associated with each job task."},{"sourceId":"6561838646a67f5fa2c5f9d6d5962a94155ce93a815545a92d24b1d26960e930","excerpt":"These findings, based on a selected sample of IT jobs, offer an early proof of concept"}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:035d1d60d875f91539b2e5147de6c354a84a354b375f9acae5dfe0d9d375e401","duplicateOf":null},"references":[{"id":"6561838646a67f5fa2c5f9d6d5962a94155ce93a815545a92d24b1d26960e930","title":"Unbundling jobs: Measuring the value of tasks in an AI economy","url":"https://www.adpresearch.com/research/unbundling-jobs-measuring-the-value-of-tasks-in-an-ai-economy"}]},{"id":"a75ff410ca8e05ea1140dc5856286ce3bc881af360bb7209d4b9221fd178d999","source":{"id":"2468975e61b5ff02a91758483f274e07cf0ffd31afe9b0eac0a229b589146e42","workId":"0a0cc2f388154f297a4c2b6db1da2223e4ff7f040836d8cdb11645aba6a30cbc","publicationId":"283f8bd9-ceb7-8757-5e3f-5b8942f975c8","lane":"bridge","textHash":"df1261ce8255f1ecab8df1542171a2bdfed488e7cee03bbb6a0744d52a210fef","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-07-23","precision":"day","basis":"backfill:publisher_archive_date"},"availableBasis":"retained-version-observation","availableAt":"2026-10-30T18:55:44.049Z","observedAt":"2026-10-30T18:54:16.413Z","authoredAt":"2026-11-03T00:31:10.955Z","publishedAt":"2026-11-03T08:33:49.638Z","indexedAt":"2026-11-03T08:54:15.015Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Entry-Level Jobs Aren’t Just for Inexperienced Workers","url":"https://hiringlab.indeed.com/2026/07/23/entry-level-jobs-arent-just-for-inexperienced-workers/"},"association":{"decision":"linked","relationship":"indirect","role":"limits","explanation":"Felix Aidala's Indeed Hiring Lab note (23 July 2026) is an observational analysis of Indeed applications by seniority level. Seniority is assigned by an algorithmic classifier with about 95% accuracy in manual checks. The note does not measure AI use, AI exposure or hires. Retrieval came from 'hiring', 'wages' and the career-entry route. The trailing subscribe prompt is boilerplate and was not used.\n\nWhat it observes: in May 2026, 30% of applications to entry-level postings came from workers with 10+ years of experience. Aidala argues this competition may make it harder for people new to the labour force to stand out. That is a non-AI channel for entry difficulty. It bears on the main question's rival, that other labour-market forces explain more than automation, and on careers-entry-pressure, whose mechanism attributes harder entry to employers using AI capacity instead of hiring.\n\nTwo of its own findings narrow how far that rival reaches the guide's population. First, the shares 'have been fairly stable over the past two years', so down-level competition is a standing feature rather than a new cause of the recent exposed-versus-less-exposed gap. Second, experienced healthcare applicants are more than twice as likely as those in tech and professional services to apply for entry roles, so this competition is weaker in the sector closest to exposed digital work.\n\nThe note therefore limits any reading of absolute entry difficulty as AI-driven. It does not challenge the relative gap that carries the guide's AI read. Aidala's phrase 'reduced entry-level hiring' is his framing; he does not attribute it to AI.","targetIds":["ai-entry-and-worker-gains","careers-entry-pressure"],"spans":[{"sourceId":"2468975e61b5ff02a91758483f274e07cf0ffd31afe9b0eac0a229b589146e42","excerpt":"Indeed, across all entry-level postings in May, 30% of applications came from workers with 10+ years of experience — more than any other group."},{"sourceId":"2468975e61b5ff02a91758483f274e07cf0ffd31afe9b0eac0a229b589146e42","excerpt":"For job seekers new to the labor force, competition from more experienced workers may make it even more difficult to stand out in a crowded field of qualified candidates."},{"sourceId":"2468975e61b5ff02a91758483f274e07cf0ffd31afe9b0eac0a229b589146e42","excerpt":"While the shares presented above have been fairly stable over the past two years"},{"sourceId":"2468975e61b5ff02a91758483f274e07cf0ffd31afe9b0eac0a229b589146e42","excerpt":"Among workers with 10+ years of experience, those applying for healthcare jobs are more than twice as likely to apply for entry-level positions as those in tech and professional services."}],"mechanism":"Experienced applicants competing for entry-level postings raise the difficulty of entry for newcomers independently of AI. That provides a non-automation explanation for absolute entry difficulty. The note's stable shares over two years and the lower down-level application rate in tech and professional services limit how much this channel can explain the recent, exposure-specific gap.","assumptions":["Application shares translate into some competition for hires; the note measures applications, not hiring outcomes.","Indeed's 'tech and professional services' grouping is a reasonable proxy for the guide's AI-exposed digital occupations.","Indeed applicants are representative enough of entrants to exposed digital work for the sector contrast to bear on that population."],"intermediateSourceIds":[],"familyId":"text:df1261ce8255f1ecab8df1542171a2bdfed488e7cee03bbb6a0744d52a210fef","duplicateOf":null},"references":[{"id":"2468975e61b5ff02a91758483f274e07cf0ffd31afe9b0eac0a229b589146e42","title":"Entry-Level Jobs Aren’t Just for Inexperienced Workers","url":"https://hiringlab.indeed.com/2026/07/23/entry-level-jobs-arent-just-for-inexperienced-workers/"}]},{"id":"96647c3e2a62a5163ddd42ee73de6420eb5c829d8ea5c0546f8ebe2ef23dcd2f","source":{"id":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","workId":"f18e35b8201ce0abc2132f6365fca3cd0abfbcf84c8be65be3dc579d11f6f1da","publicationId":"9f995f0c-1bb5-6545-19b3-b4e8774d60c4","lane":"bridge","textHash":"6a0e3434ef4299f5f1995554b9c2f6fa302f59b11a0130d1a4ebee8112cb0153","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-07-23","precision":"day","basis":"backfill:publisher_archive_date"},"availableBasis":"retained-version-observation","availableAt":"2026-10-30T20:46:00.848Z","observedAt":"2026-10-30T19:18:29.241Z","authoredAt":"2026-11-03T09:29:26.780Z","publishedAt":"2026-11-03T09:58:49.699Z","indexedAt":"2026-11-03T09:58:48.689Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"The Labor Market Is Tilting Toward Seniority","url":"https://hiringlab.indeed.com/2026/07/23/the-labor-market-is-tilting-toward-seniority/"},"association":{"decision":"linked","relationship":"direct","role":"limits","explanation":"Felix Aidala and Sneha Puri's Indeed Hiring Lab analysis (23 July 2026) is observational. It reports Indeed postings by seniority, assigned by the same algorithmic classifier with about 95% manual-check accuracy used in Aidala's already adopted applications note. It does not measure AI use, hires or progression. It is a separate analysis of postings, not applications, so it is not a duplicate. Because it shares a classifier and publisher, it should not be counted as fully independent of that note.\n\nRetrieval came from 'hiring', 'wages' and the career-entry route. The subscribe prompt and the carousel teasers are boilerplate and were not used.\n\nWhat it observes. Nationally, entry-level postings have trended down since their 2022 peak, falling 7.5% year over year to May 2026, while senior postings rose. In software development only 4.5% of postings were entry-level in Q1 2026. In tech, the entry-level share fell by just over one percentage point between 2019 and 2025. That bears on entry-routes-narrow signpost 0, junior openings staying weak relative to senior openings in exposed occupations, and is consistent with it. It also bears on the careers-entry-pressure path and signpost, and on trigger 0: there is no sign of entry postings recovering in tech. These are postings, which signal intentions, not entry hires.\n\nWhy it limits rather than supports. Three findings constrain the guide's AI reading.\n- The authors relay New York Fed researchers' argument that remote work is more closely linked to recent graduates' difficulties. They also list rate hikes and post-pandemic right-sizing, and do not rank these causes. Remote work is a rival the guide does not name. Ability to work remotely plausibly correlates with AI exposure across white-collar work, so a sector-level association cannot separate the two. That is EM's inference, and the authors make the same point when they write 'AI exposure and/or more opportunities to work remotely'.\n- In tech, the 9-point senior gain came mainly at the expense of mid-level postings, not entry-level ones.\n- Healthcare, which is less exposed, saw a larger fall in entry-level share (5.2 points) than tech.\n\nShares within a sector are not the exposed-versus-less-exposed employment gap that Stanford and Revelio measure. This therefore qualifies the mechanism rather than challenging the relative finding. The source is also internally inconsistent: it says entry postings saw 'slight upticks' while its key points report year-over-year declines. We rely on the key-point figures.","targetIds":["ai-entry-and-worker-gains","ai-entry-and-worker-gains:trigger:0","entry-routes-narrow:signpost:0","careers-entry-pressure","careers-entry-pressure:signpost"],"spans":[{"sourceId":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","excerpt":"Entry-level job postings have been trending downwards since peaking in 2022, declining 7.5% year-over-year as of May 2026."},{"sourceId":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","excerpt":"The share of job postings that are entry-level positions varies widely by sector, ranging from 91.3% in personal care & home health to 4.5% in software development in Q1 2026."},{"sourceId":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","excerpt":"Tech saw a more modest decline in the entry-level job share of just over one percentage point. Still, given how small its entry-level share already was, even marginal shifts could carry outsized consequences for workers trying to break into the field."},{"sourceId":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","excerpt":"suggesting that the 9 percentage-point growth in senior-level job postings in tech has primarily come at the expense of mid-level job postings."},{"sourceId":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","excerpt":"But researchers at the New York Fed recently argued that the rise of remote work is more closely linked to recent graduates’ labor-market woes."},{"sourceId":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","excerpt":"Broadly speaking, sectors with high AI exposure and/or more opportunities to work remotely, like tech, tend to have fewer entry-level roles."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:6a0e3434ef4299f5f1995554b9c2f6fa302f59b11a0130d1a4ebee8112cb0153","duplicateOf":null},"references":[{"id":"3c7a0700b1d0630160ad0c3d8847d40938a3a82d5c4fcd1214d5296437de4dab","title":"The Labor Market Is Tilting Toward Seniority","url":"https://hiringlab.indeed.com/2026/07/23/the-labor-market-is-tilting-toward-seniority/"}]},{"id":"9fb772ede2bed6b4c6cbcb1610cb8f174129951349a04f73adc0e63471a9846b","source":{"id":"4dba6cf255a15b2c975d0d7beefceefe9bc3a66adb903d0a480ebb91e05bead7","workId":"b4dd0e2cf78fa407c2febf0c6299da3cdbf5f264d210095f298514a3dc08990a","publicationId":"67e307de-5222-4c91-8415-105765030ad0","lane":"live","textHash":"adc33f720095a6f95414cb7e28d8e544945fe7005629f4f90dee1b88d93e8edf","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-01-31","precision":"day","basis":"publisher_citation_date"},"availableBasis":"retained-version-observation","availableAt":"2026-11-20T13:51:12.531Z","observedAt":"2026-11-20T13:50:23.739Z","authoredAt":"2026-11-20T13:51:34.784Z","publishedAt":"2026-11-20T13:52:03.002Z","indexedAt":"2026-11-20T13:53:22.128Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI-Mediated Software Engineering","url":"https://arxiv.org/abs/2602.00496"},"association":{"decision":"linked","relationship":"direct","role":"context","explanation":"Feng, Yun and Wang's CHI '26 paper (arXiv v1 31 Jan 2026, v2 11 Feb 2026; observed here 20 Nov 2026) is a small mixed-methods study of junior and senior software engineers using agentic AI. It combines an ACTA and Delphi process with 5 seniors, an AI-assisted debugging task with 10 juniors, and blind reviews of junior prompt histories by 5 more seniors. Only the abstract is available. It does not measure hiring, pay, entry openings or progression, so it does not reach the entry-pressure paths, the hiring triggers or the main hiring forecast. Retrieval came from the token 'mentoring'. The arXiv tool listings are page furniture.\n\nThe match is substantive for the guide's learning conditions in an exposed digital occupation. The authors report that 'novices struggle between over-reliance and cautious avoidance'. That bears directly on entry-opportunities-renew condition 1 (workers can question the tool, receive feedback and build expertise beyond accepting its output) and careers-apprenticeship condition 1 (progression depends on detecting and correcting AI mistakes). It shows these conditions are not met by default; they have to be built. The authors also report that seniors 'possess valuable perspectives for mentoring juniors'. That is consistent with careers-apprenticeship condition 0, which treats mentoring as a lever, though the study does not test whether mentoring changes outcomes. Their finding that agency is 'primarily constrained by organizational policies rather than individual preferences' bears on entry-opportunities-renew condition 0. It places the decision with the employer, as the guide does, rather than with individual workers.\n\nThis is context, not support or challenge. Over-reliance is an observed behaviour in one debugging task, not evidence that people are rewarded for output over understanding (careers-hollow-learning condition 1). It is also not evidence that progression falls. The finding extends Mollick's classroom-based argument, that how AI is used determines whether people learn, to junior software engineers. That connection is EM's. The sample is very small, the abstract does not say whether the juniors were employed, and the three proposed practices are the authors' suggestions, not tested interventions.","targetIds":["entry-opportunities-renew:condition:1","entry-opportunities-renew:condition:0","careers-apprenticeship:condition:1","careers-apprenticeship:condition:0"],"spans":[{"sourceId":"4dba6cf255a15b2c975d0d7beefceefe9bc3a66adb903d0a480ebb91e05bead7","excerpt":"agency in software engineering is primarily constrained by organizational policies rather than individual preferences, with experienced developers maintaining control through detailed delegation while novices struggle between over-reliance and cautious avoidance."},{"sourceId":"4dba6cf255a15b2c975d0d7beefceefe9bc3a66adb903d0a480ebb91e05bead7","excerpt":"Seniors leverage pre-AI foundational instincts to steer modern tools and possess valuable perspectives for mentoring juniors in their early AI-encouraged career development."},{"sourceId":"4dba6cf255a15b2c975d0d7beefceefe9bc3a66adb903d0a480ebb91e05bead7","excerpt":"ACTA combined with a Delphi process with 5 seniors, an AI-assisted debugging task with 10 juniors, and blind reviews of junior prompt histories by 5 more seniors."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:adc33f720095a6f95414cb7e28d8e544945fe7005629f4f90dee1b88d93e8edf","duplicateOf":null},"references":[{"id":"4dba6cf255a15b2c975d0d7beefceefe9bc3a66adb903d0a480ebb91e05bead7","title":"From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI-Mediated Software Engineering","url":"https://arxiv.org/abs/2602.00496"}]},{"id":"fc72a9fe8e1be1a8a911b3bda149a6ce13f44cd93a61d8c8ec1f3bc1d5149df8","source":{"id":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","workId":"9850bcabd3c2776c23653d367680367b2403dbafc9346d9de57182f9d6c20199","publicationId":"56616587-5dc2-48b0-86bc-578f02583c7d","lane":"live","textHash":"0f2c37992c0ce8768bb5b809aab0f5304c0c699ec67a36badbc749a80f1385a5","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-07-27","precision":"day","basis":"publisher_citation_date"},"availableBasis":"retained-version-observation","availableAt":"2026-11-20T13:52:04.014Z","observedAt":"2026-11-20T13:50:23.739Z","authoredAt":"2026-11-20T13:52:44.881Z","publishedAt":"2026-11-20T13:53:35.151Z","indexedAt":"2026-11-20T13:53:38.906Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Verification-Conditioned Use: A Qualitative Study on How Generative AI Reshapes Learning, Autonomy, and Market Entry for Junior Software Developers","url":"https://arxiv.org/abs/2607.24606"},"association":{"decision":"linked","relationship":"direct","role":"context","explanation":"Andriotte and Ribeiro's arXiv paper (v1, submitted 27 July 2026; observed here 20 November 2026) is a qualitative interview study. Thirteen interns and junior developers were interviewed and the interviews were thematically analysed. Only the abstract page is supplied. Retrieval came from the career-progression route ('junior', 'learn'). The arXiv tool listings are page furniture and were not used. The match is substantive: the population is newcomers to software development, an AI-exposed digital occupation, and the study addresses learning and the judgment needed for independent work.\n\nWhat the source observes is participants' self-report. Whether they can check a result most often decides between AI and manual work. They also describe an 'autonomy paradox': feeling more capable yet less in ownership of the result. The authors then argue a 'formative paradox', that AI-induced shallow learning makes critical judgment harder to build. That is their interpretation, not a measured learning outcome. The abstract names market entry as a research question but reports no hiring, openings, pay or progression findings. It therefore does not reach careers-entry-pressure, entry-routes-narrow or the hiring triggers.\n\nWhich targets it reaches. (1) careers-hollow-learning: the autonomy paradox and the formative paradox are newcomers' own account of the path's mechanism and its first implication, greater dependence without readiness for responsibility. It does not show that entry work is being removed (condition 0), nor that people are rewarded for output over understanding (condition 1). (2) careers-apprenticeship condition 1 and entry-opportunities-renew condition 1: participants converge on verification-conditioned use, meaning reviewing before accepting, asking for explanations and practising outside AI-assisted work. That is the competence those conditions require. The participants frame these as individual practices, however, not as employer-provided feedback or mentoring. Read with Feng, Yun and Wang's finding that organizational policy constrains agency, our inference is that the guide's conditions remain employer-dependent. (3) Assumption 1 of the main question: participants say the market 'has begun to demand' critical judgment. This is newcomers' perception of employer demand, not employer evidence.\n\nThe role is context rather than support. The sample is small and self-reported, and no comparison group or outcome measure is reported. The study extends Mollick's 'how you use AI' argument and the Feng et al. over-reliance finding to junior developers' own accounts. Those connections are EM's. It is a distinct study, not a duplicate of Feng et al.","targetIds":["careers-hollow-learning","careers-apprenticeship:condition:1","entry-opportunities-renew:condition:1","ai-entry-and-worker-gains:assumption:1"],"spans":[{"sourceId":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","excerpt":"thirteen interns and junior developers were interviewed individually, by videoconference."},{"sourceId":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","excerpt":"the criterion that most often decides between AI and manual work is not deadline or task complexity, but the ability to check the result."},{"sourceId":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","excerpt":"the autonomy paradox (feeling more capable yet less in ownership of the result)"},{"sourceId":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","excerpt":"the formative paradox: the shallow learning that AI induces makes it harder to build the very critical-judgment competence that, according to participants, the market has begun to demand."},{"sourceId":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","excerpt":"the individual practice of reviewing before accepting, refusing to use AI without understanding it, asking the tool for explanations, and keeping deliberate practice outside of AI-assisted work."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:0f2c37992c0ce8768bb5b809aab0f5304c0c699ec67a36badbc749a80f1385a5","duplicateOf":null},"references":[{"id":"8292123889c408d14c0ef026f248422432bf6650cd749e0112dc6f5037405b37","title":"Verification-Conditioned Use: A Qualitative Study on How Generative AI Reshapes Learning, Autonomy, and Market Entry for Junior Software Developers","url":"https://arxiv.org/abs/2607.24606"}]},{"id":"73936a9eecc2314244924bd846107b447c6b2b5c483280fac80e63b17b48c6a0","source":{"id":"cb426071d42fe4d7ad164e055ff4f3ba9b85cd6dd3749f8f095fce6e22e53441","workId":"acbb3e5297e8a038721f3c2731b50c19a5a7d938038c7c7be00716b79cd40355","publicationId":"cbea0ea9-83dc-4f12-8040-698a5ad224e6","lane":"live","textHash":"53360d1505fe8ac6567bcab1b019f5a4e7686f45476018376148433d188d6b60","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-05-22","precision":"day","basis":"publisher_citation_date"},"availableBasis":"retained-version-observation","availableAt":"2026-11-21T10:09:16.652Z","observedAt":"2026-11-21T10:07:46.641Z","authoredAt":"2026-11-21T10:09:36.790Z","publishedAt":"2026-11-21T10:09:58.693Z","indexedAt":"2026-11-21T10:14:01.897Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Generative AI and the Reorganization of Labor Demand","url":"https://arxiv.org/abs/2605.23159"},"association":{"decision":"linked","relationship":"direct","role":"context","explanation":"Wang, Wei and Wang's arXiv paper (v1 submitted 22 May 2026; observed here 21 November 2026) is an observational decomposition of nationwide US job postings across all sectors. Only the abstract page is supplied; the arXiv tool listings are page furniture. Retrieval came from 'hiring', the career-entry route and careers-entry-pressure. The match is substantive because the paper examines how firms reorganise hiring demand and job content as generative AI diffuses, with separate results for junior and senior jobs.\n\nWhat it measures. 'Exposure' is the authors' LLM-classified share of posting tasks that generative AI can perform or assist. It is not AI use, hires, openings, pay or progression. A decline in aggregate exposure means postings describe less AI-performable or AI-assistable work. It does not mean less AI adoption. The authors attribute 52% of that decline to hiring reallocation across jobs and 39.5% to task redesign within jobs. They report that senior jobs adjust earlier and mainly through reallocation, while junior jobs adjust through a broader mix that includes redesign.\n\nMain question. The findings fit the guide's broad claim that employers adjust hiring demand rather than simply cutting workers. They also add a second channel the main argument underplays: firms change what junior jobs contain, not only how many beginners they recruit. The abstract gives no junior posting levels, so the paper cannot show whether entry openings fell. It therefore neither supports nor challenges the 2027 forecast that entry stays difficult.\n\ncareers-entry-pressure. The reallocation margin is consistent with the path's mechanism, in which employers shift demand away from tasks tools can handle. The authors locate reallocation mainly in senior jobs, however, so this is not evidence that beginners specifically bear that margin.\n\ncareers-hollow-learning. Within-job redesign of junior postings is the closest large-scale observation yet of the path's premise that entry work is being reshaped. This is our inference. Posting text is not the work actually performed. The paper also does not show that the removed tasks were the ones that taught the job, or that no other learning route replaces them.\n\nAn additional limit, also our inference. If exposure is 'dynamic rather than fixed', the fixed occupational exposure comparisons in Stanford and Revelio may misclassify some jobs over time. That qualifies how cleanly those comparisons separate exposed from less-exposed work. It does not reverse them.\n\nThe paper is a distinct study and dataset, not a duplicate of any adopted family.","targetIds":["ai-entry-and-worker-gains","careers-entry-pressure","careers-hollow-learning"],"spans":[{"sourceId":"cb426071d42fe4d7ad164e055ff4f3ba9b85cd6dd3749f8f095fce6e22e53441","excerpt":"The pipeline identifies the tasks described in each posting and classifies the extent to which generative AI can perform or assist them."},{"sourceId":"cb426071d42fe4d7ad164e055ff4f3ba9b85cd6dd3749f8f095fce6e22e53441","excerpt":"Hiring reallocation explains the largest share of the aggregate decline in exposure, accounting for 52% on average, while within-job redesign becomes increasingly important, accounting for 39.5%."},{"sourceId":"cb426071d42fe4d7ad164e055ff4f3ba9b85cd6dd3749f8f095fce6e22e53441","excerpt":"Senior jobs adjust earlier and mainly through reallocation, whereas junior jobs adjust through a broader mix of reallocation, redesign, and their interaction."},{"sourceId":"cb426071d42fe4d7ad164e055ff4f3ba9b85cd6dd3749f8f095fce6e22e53441","excerpt":"First, generative AI exposure is dynamic rather than fixed, changing substantially over time."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:53360d1505fe8ac6567bcab1b019f5a4e7686f45476018376148433d188d6b60","duplicateOf":null},"references":[{"id":"cb426071d42fe4d7ad164e055ff4f3ba9b85cd6dd3749f8f095fce6e22e53441","title":"Generative AI and the Reorganization of Labor Demand","url":"https://arxiv.org/abs/2605.23159"}]},{"id":"2578c247a2d8c37c60bf7c9b5b6f25b434fcff19a33f7243bf5e60aec5e78104","source":{"id":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","workId":"46fcbe73d5f646be9e8edba993626bb58ddaf8c2e38dcc4f8f6556bdfe128514","publicationId":"ee4d0d13-14ee-4df8-84f5-081421f3e5cc","lane":"live","textHash":"f6d59e98423447ded51bc7d40cf32ace5e58ac845a2341b766d79e8f54aa1e12","textBasis":"source-text","clocks":{"eventTime":{"value":null,"precision":"unknown"},"sourcePublished":{"value":"2026-07-19","precision":"day","basis":"publisher_citation_date"},"availableBasis":"retained-version-observation","availableAt":"2026-11-21T10:09:59.589Z","observedAt":"2026-11-21T10:07:46.641Z","authoredAt":"2026-11-21T10:10:19.933Z","publishedAt":"2026-11-21T10:10:47.325Z","indexedAt":"2026-11-21T10:14:05.297Z"},"eligibleWhenAdopted":true,"currentlyEligible":true,"eligibilityReason":"canonical_public_reading","title":"Who Will Become the Next Senior? How Generative AI Erodes the Development Pathway in Software Engineering","url":"https://arxiv.org/abs/2607.17067"},"association":{"decision":"linked","relationship":"direct","role":"supports","explanation":"Yu and Moon's arXiv paper (v1 19 July 2026, v3 25 August 2026, accepted to AIES 2026; observed here 21 November 2026) is a qualitative study of software engineering, an AI-exposed digital occupation. It rests on 14 semi-structured interviews in South Korea with juniors at the threshold of entering the field and with senior engineers, analysed by Reflexive Thematic Analysis. Only the abstract page is supplied, so methods beyond that are unavailable. The arXiv tool listings are page furniture and were not used. Retrieval came from 'hiring'. That word appears only in the authors' relay that 'macro statistics show a decline in junior hiring', which restates the evidence family the guide already holds (Stanford, Revelio) and adds no weight to it. The substantive match is elsewhere.\n\nWhat it reports. The authors' central pattern, 'Absorption', is that GenAI redirects entry-level work into senior-AI workflows. That is a field account of the mechanism the guide already names: careers-entry-pressure says tools let the current team handle tasks previously assigned to beginners, and entry-routes-narrow condition 0 says employers automate routine implementation before replacing the training it once provided. Their first consequence, juniors losing the 'productive struggle' through which expertise developed, bears directly on careers-hollow-learning and its condition 0, that entry tasks are removed without another route to learn. Their argument that the loss is reproduced through normalised GenAI use in university classrooms adds an educational channel the guide does not yet name. Their prescription for deliberate institutional design across classrooms, workplaces and junior evaluation criteria matches entry-opportunities-renew condition 2, which requires employers and educators to build paid apprenticeships. That prescription is an argument, not a tested intervention.\n\nScope and limits. These are interview-based perceptions and the authors' interpretation; the hedge 'appears to be absorbing' is theirs. The study measures no hiring, openings, pay or progression, so it does not reach the hiring triggers or the careers-hollow-learning signpost (fewer newcomers progressing to independent work). The sample is small and Korean, and it combines would-be entrants with seniors rather than employed juniors. It is a distinct study from Feng, Yun and Wang and from Andriotte and Ribeiro, not a duplicate. It strengthens the qualitative base for the hollow-learning mechanism; together, those studies remain accounts of mechanism rather than outcome measures. A rival reading is that redirected work reflects a temporary adjustment that later training redesign absorbs. Only cohort progression data would distinguish the two.","targetIds":["careers-hollow-learning","careers-hollow-learning:condition:0","entry-routes-narrow:condition:0","careers-entry-pressure","entry-opportunities-renew:condition:2"],"spans":[{"sourceId":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","excerpt":"Through 14 semi-structured interviews with juniors at the threshold of entering software engineering and senior software engineers in South Korea"},{"sourceId":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","excerpt":"Absorption -- GenAI redirects entry-level work into senior-AI workflows"},{"sourceId":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","excerpt":"juniors losing the productive struggle through which expertise once developed"},{"sourceId":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","excerpt":"the structural reproduction of this loss through collective normalization of GenAI use in university classrooms"},{"sourceId":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","excerpt":"Preserving this pathway will require deliberate institutional design across classrooms, workplaces, and the evaluation criteria for juniors."}],"mechanism":null,"assumptions":[],"intermediateSourceIds":[],"familyId":"text:f6d59e98423447ded51bc7d40cf32ace5e58ac845a2341b766d79e8f54aa1e12","duplicateOf":null},"references":[{"id":"e3d2e8470102aa25eabefb8afc5055a056458835857e6f3b93fc2cf9c6576ef8","title":"Who Will Become the Next Senior? How Generative AI Erodes the Development Pathway in Software Engineering","url":"https://arxiv.org/abs/2607.17067"}]}],"revisionUrl":"/guide-maintenance/ai-entry-and-worker-gains/history.json?revision=2605d750764ae7e19980b4d1a22313f8d6d4393a8c98a05f7bd47c122ed44be8","document":{"questionId":"ai-entry-and-worker-gains","questionRead":{"gapLabel":"AI can help a junior do the job—and remove the opening","basis":"We expect entry into AI-exposed digital occupations to stay difficult through 2027, even as AI makes some novices more productive.\n\nStanford’s August 2026 payroll study finds young workers in AI-exposed occupations falling behind less-exposed peers, mainly because fewer are hired; Anthropic’s workforce report describes a declining junior share of its own technical hires. Yet in the customer-support study, less experienced agents gained more from AI assistance.\n\nThose findings can coexist: a tool can help a novice already in the job while an employer decides to recruit fewer beginners. The consequence is a harder route to the experience that future skilled work requires.","alternatives":["AI lowers the cost of training and makes novices valuable sooner, encouraging employers to expand entry hiring.","Education mix, interest rates and earlier labor-market trends explain more of the apparent AI pattern than automation does."],"assumptions":["The outlook is strongest for exposed digital roles, not every young worker or occupation.","Employers continue to value independent judgment while automating some tasks traditionally assigned to novices."],"revisionTriggers":["Entry hiring recovers in exposed occupations while AI use continues to deepen.","Employers report and demonstrate new paid learning routes with progression into responsibility.","The employment gap shrinks materially after accounting for education, prior trends and broader national data."],"implication":"Employers deciding what to automate are also deciding where experience comes from. Preserve paid work that teaches judgment; job seekers should show they can own an outcome, including checking and improving AI work.","estimate":{"status":"editorial-outlook","horizon":{"start":"2027","end":"2027","reason":"The hiring pattern has persisted into mid-2026 and Anthropic reports a similar shift in October. Another hiring cycle is more likely to extend that pressure than reverse it without deliberate new entry routes."}},"sourceRefs":[{"sourceId":"anthropic-workforce-2026","spanIds":["anth-junior"],"role":"context"},{"sourceId":"nyfed-business-survey-2026","spanIds":["nyf-fewer","nyf-more"],"role":"context"},{"sourceId":"steering-beyond-automation","spanIds":["un-voice"],"role":"context"},{"sourceId":"stanford-canaries-2026","spanIds":["hiring-first","descriptive"],"role":"supports"},{"sourceId":"brynjolfsson-ai-work-v2","spanIds":["novice-gains"],"role":"challenges"}],"positionRefs":["unger-narayanan-collaboration","stanford-entry-researchers","customer-support-researchers","mollick-retain-the-learning"],"connectionRefs":[],"gapRefs":[{"id":"entry-progression-and-distribution","reason":"Test the hiring interpretation against education, prior trends and national data, and follow actual paid entry and progression routes."}]},"questionPaths":[{"id":"entry-routes-narrow","questionId":"ai-entry-and-worker-gains","outcome":"Fewer first jobs, even when the work gets easier","scope":"Employers ask smaller experienced teams to deliver more. Novices who get access may learn faster, but fewer people receive paid entry into the occupation.","conditions":["Employers automate routine implementation before they replace the training it once provided.","Higher output per team does not create enough new demand to restore entry hiring."],"sourceRefs":[{"sourceId":"anthropic-workforce-2026","spanIds":["anth-junior"],"role":"context"},{"sourceId":"nyfed-business-survey-2026","spanIds":["nyf-fewer","nyf-more"],"role":"context"},{"sourceId":"stanford-canaries-2026","spanIds":["hiring-first"],"role":"supports"},{"sourceId":"brynjolfsson-ai-work-v2","spanIds":["novice-gains"],"role":"context"}],"positionRefs":["anthropic-workforce-authors","stanford-entry-researchers","customer-support-researchers"],"horizon":{"start":"2027","end":"2027","reason":"The existing hiring divergence and Anthropic’s reported change in junior hiring make continued pressure through the next hiring year more plausible than a rapid reversal."},"weight":"base-case","signposts":["Junior openings stay weak relative to senior openings in exposed occupations.","Employers raise expectations for independent ownership while reducing paid learning roles."]},{"id":"entry-opportunities-renew","questionId":"ai-entry-and-worker-gains","outcome":"AI becomes an apprenticeship multiplier","scope":"Employers use assistance to make newcomers productive sooner and give them progressively harder work. New entry routes grow because teaching and supervision become worth the investment.","conditions":["The employer rewards learning and creates responsibility rather than using every gain to reduce hiring.","Workers can question the tool, receive feedback and build expertise beyond accepting its output.","Employers and educators build paid apprenticeships around real outcomes, with experienced workers given time and credit for teaching."],"sourceRefs":[{"sourceId":"steering-beyond-automation","spanIds":["un-voice"],"role":"context"},{"sourceId":"anthropic-workforce-2026","spanIds":["anth-junior"],"role":"challenges"},{"sourceId":"brynjolfsson-ai-work-v2","spanIds":["novice-gains"],"role":"supports"}],"positionRefs":["unger-narayanan-collaboration","customer-support-researchers"],"horizon":{"start":"2027","end":"2029","reason":"Assisted novice gains already exist in one studied setting; expanding paid entry routes requires employers to turn that capability into hiring and progression over several cohorts."},"weight":"alternative","signposts":["Firms add entry roles and show progression into more responsible work.","Novices retain useful skills when the tool is unavailable and share in improved pay or opportunity."]}],"dimension":{"id":"teams-careers","label":"Teams and careers","question":"Will better tools create room to grow—or close the way in?","verdict":"Managers should redesign work and develop people together. Workers should use AI to take responsibility for a useful result, while continuing to build the judgment needed to recognize when the tool is wrong.","who":"Team leaders deciding how work is assigned, and people entering or changing a career. A license is not a training plan, and producing more is not the same as becoming ready for a larger responsibility.","tasks":[{"title":"Decide who owns the work after AI produces it","how":"Choose a recurring task, the person who needs its result and the person who checks it. Give that person time and authority to reject or improve the output.","sourceIds":["anthropic-workforce-2026"]},{"title":"Use assistance to build responsibility","how":"Give newer colleagues useful paid work with explanation and feedback. Ask them to show how they checked the result and what they would do differently, then expand the responsibility they can carry.","sourceIds":["brynjolfsson-ai-work-v2","stanford-canaries-2026"]}],"limits":[{"title":"A more productive beginner may still find fewer openings","why":"Stanford’s August 2026 payroll study finds young workers in AI-exposed occupations falling behind less-exposed peers mainly through hiring. The support study’s novice gains concern people already doing the job. Both can be true.","sourceIds":["stanford-canaries-2026","brynjolfsson-ai-work-v2"]},{"title":"More output can make attention the constraint","why":"Anthropic’s report identifies demands on human attention as agent output grows. A team can adopt tools faster than it learns to decide what deserves review and which work matters.","sourceIds":["anthropic-workforce-2026"]}],"practice":"Learn where AI improves your actual work, and make the result and your judgment visible. If you lead a team, use gains to strengthen service and learning as well as delivery; decide deliberately where the next generation of experienced people will come from.","sourceIds":["brynjolfsson-ai-work-v2","stanford-canaries-2026","anthropic-workforce-2026","nyfed-business-survey-2026","context-chatgpt-source","context-ai-act-source"],"paths":[{"id":"careers-entry-pressure","title":"Entry gets harder while experienced teams gain capacity","summary":"Through 2027, we expect employers in AI-exposed digital work to keep scrutinizing new hires even as some existing workers accomplish more with assistance.","mechanism":"An employer can leave an opening unfilled while tools help the current team handle tasks previously assigned to beginners. That choice changes access to experience before it eliminates an entire profession.","horizon":{"start":"2027","end":"2027","label":"Through 2027","reason":"The relative hiring pattern persisted into mid-2026, and the October company report adds a changing junior hiring mix. That makes continued pressure through another hiring cycle more plausible than a quick reversal."},"conditions":["Employers use some additional capacity to reduce recruitment.","New paid learning routes do not expand enough to offset the lost openings."],"implications":["A job seeker needs opportunities to demonstrate judgment, not only familiarity with tools.","Experienced teams may carry more work while the route into those teams narrows."],"signpost":"Entry hiring remains weak relative to comparable less-exposed work as AI adoption deepens.","sourceIds":["stanford-canaries-2026","anthropic-workforce-2026","nyfed-business-survey-2026"],"kind":"favored"},{"id":"careers-apprenticeship","title":"AI becomes part of a stronger apprenticeship","summary":"Employers use AI assistance to bring more people into useful work, with support while they learn and a clear route toward independent responsibility.","mechanism":"Novice performance gains create an opportunity to make early work more valuable. It becomes an apprenticeship only when employers keep hiring, explain decisions and let people practice the judgment the tool cannot supply for them.","horizon":{"start":null,"end":null,"label":"A future people can shape","reason":"Its pace depends on employers preserving paid entry roles, mentoring people and expanding responsibility as they demonstrate competence."},"conditions":["Employers preserve paid entry work and mentoring.","Progression depends on demonstrated competence, including detecting and correcting AI mistakes."],"implications":["More people can reach skilled work rather than watching it from outside.","Organizations build the experienced workforce they will need later."],"signpost":"Paid entry routes expand and participants progress into greater responsibility while quality holds up.","sourceIds":["brynjolfsson-ai-work-v2","stanford-canaries-2026"],"kind":"opportunity"},{"id":"careers-hollow-learning","title":"Teams remove the work that used to teach the job","summary":"Organizations save effort today but leave fewer ways for people to acquire the experience needed to make tomorrow’s difficult decisions.","mechanism":"If routine entry work disappears without replacement, novices lose practice and employers lose a route for developing future expertise. Current hiring pressure makes that mechanism worth acting on, even though the long-run outcome is not yet measured.","horizon":{"start":null,"end":null,"label":"Over successive hiring and training cycles","reason":"The effects accumulate as hiring and training decisions determine who gets enough practice to become experienced."},"conditions":["Employers remove entry tasks without creating another route to learn.","People are rewarded for producing AI output rather than understanding and improving it."],"implications":["Workers can become more dependent on assistance without becoming ready for greater responsibility.","Teams may struggle to replace the experienced people whose judgment currently keeps the work reliable."],"signpost":"Fewer newcomers progress to independent work even as output per current worker rises.","sourceIds":["stanford-canaries-2026","brynjolfsson-ai-work-v2","anthropic-workforce-2026"],"kind":"risk"}],"contextSourceIds":["context-chatgpt-source","context-ai-act-source"]},"foundation":{"editionId":"perspectives-scope-correction-2026-11-14","sources":[{"id":"anthropic-workforce-2026","title":"How employment is changing at Anthropic","url":"https://www.anthropic.com/research/employment-at-anthropic","publishedAt":"2026-10-16","dateBasis":"Publisher publication date, checked on original page.","access":{"status":"public-page","note":"Original publisher page and exact retained source reviewed. Source retention does not establish every downstream reuse or training right.","checkedAt":"2026-10-29"},"provenance":{"kind":"stream-snapshot","workId":"18f3c3ca31491462a8ab52f797ebadd2a7087803317e69a5152e2a0c4cfbed65","snapshotId":"a633ce13b2d661ad9c008ac99d3357f15a80c4bcd91009290d11712c2f565793","textHash":"8ea3533aa08b1f63c143b0a383911da93e88f464edd2d8e61edebf2a97a42232","reason":"Exact retained Stream work/snapshot and SHA256 verified against the 29 October SELECT-only source packet. That lookup found no joined public Stream or bridge reading revision; no reading identity is invented."},"spans":[{"id":"anth-junior","text":"The proportion of junior technical hires fell even faster, from 27% to 9%.","locator":"Retained extracted text, character offsets [4420, 4494). Context (EM paraphrase): Anthropic’s October 2026 report describes its own workforce; junior employees’ share fell from early 2025, and technical hiring still grew in absolute terms. This is a changing hiring mix, not a count of jobs lost."},{"id":"anth-bottleneck","text":"This acceleration makes human attention the bottleneck for agent output.","locator":"Retained extracted text, character offsets [767, 839). Context (EM paraphrase): The authors report rising internal output alongside agent use and interpret human attention as a constraint. Adoption is not randomized; the report does not establish a general causal productivity effect."}],"citation":{"label":"Anthropic workforce report · October 2026","aliases":["Anthropic’s workforce report","Anthropic workforce report","Anthropic’s hiring mix"],"context":"An internal study of employment and agent use at Anthropic. It describes a changing hiring mix and demands on human attention at one growing AI company.","spanIds":["anth-junior","anth-bottleneck"]},"readingRefs":[]},{"id":"nyfed-business-survey-2026","title":"Businesses Are Using AI to Transform Work, Not Cut Jobs","url":"https://libertystreeteconomics.newyorkfed.org/2026/09/businesses-are-using-ai-to-transform-work-not-cut-jobs/","publishedAt":"2026-09-01","dateBasis":"Publisher publication date, checked on original page.","access":{"status":"public-page","note":"Original publisher page and exact retained source reviewed. Source retention does not establish every downstream reuse or training right.","checkedAt":"2026-10-29"},"provenance":{"kind":"stream-snapshot","workId":"c16ff26252860625027590ed41b03b3773be9d5b324475687ef0304e51a73af8","snapshotId":"d59dc274ed0f66564545055ec05e4a9abf5a5314f05623d021bbe3ac1b2274ab","textHash":"7ec36434625838d3aa01b2819ecb8d90e622fd2f2e8a664dd2c656116c4829dd","reason":"Exact retained Stream work/snapshot and SHA256 verified against the 29 October SELECT-only source packet. That lookup found no joined public Stream or bridge reading revision; no reading identity is invented."},"spans":[{"id":"nyf-fewer","text":"hired fewer workers","locator":"Retained extracted text, character offsets [4475, 4494). Context (EM paraphrase): August 2026 regional survey: about 15% of AI-using service firms reported hiring fewer workers than they would otherwise have because of AI. This is employers’ stated counterfactual, not independently identified job loss. The paired figure for hiring more is about 13%; firm percentages cannot be subtracted into a net job count."},{"id":"nyf-more","text":"About 13 percent of service firms said they had hired more workers due to AI","locator":"Retained extracted text, character offsets [4776, 4852). Context (EM paraphrase): August 2026 regional business survey, reporting workforce adjustments over the preceding six months. The workforce charts concern AI-using firms in the survey; they are not national employment totals. Authors also report retraining as a common response."},{"id":"worker-adoption","text":"17 percent","locator":"Retained extracted text, character offsets [2997, 3007). Context (EM paraphrase): Among AI-using regional service firms, median worker usage in the August 2026 survey was 17%; manufacturers were at 7%. Firm adoption and workforce penetration are different measures."},{"id":"reported-layoffs","text":"4 percent","locator":"Retained extracted text, character offsets [4214, 4223). Context (EM paraphrase): AI-using service firms reporting AI-related layoffs over the preceding six months, versus about 15% reporting reduced hiring. Regional employer reports, not national worker shares."}],"citation":{"label":"New York Fed regional business survey · September 2026","aliases":["New York Fed regional business survey","New York Fed’s regional survey","regional employer survey"],"context":"A report on the August 2026 survey of businesses in New York and northern New Jersey, distinguishing firm adoption, worker usage, hiring and layoffs.","spanIds":["nyf-fewer","nyf-more","worker-adoption","reported-layoffs"]},"readingRefs":[]},{"id":"steering-beyond-automation","title":"Steering AI Beyond Automation","url":"https://www.normaltech.ai/p/steering-ai-beyond-automation","publishedAt":"2026-10-14","dateBasis":"Publisher publication date, checked on original page.","access":{"status":"public-page","note":"Original publisher page and exact retained source reviewed. Source retention does not establish every downstream reuse or training right.","checkedAt":"2026-10-29"},"provenance":{"kind":"stream-snapshot","workId":"5a4e1d4b2fb65d2a40b74bca91665b9dc8321e961ea8e51dea6a078d98139bff","snapshotId":"11ae3a79455a2c294b795984f741f4bb1a3df2cbe7f39d2e0867564a93f6a8a2","textHash":"f2b6cc0e1c8b6bbfe0e9efb6e94a5614b87630de3934d9902943551c31faf389","reason":"Exact retained Stream work/snapshot and SHA256 verified against the 29 October SELECT-only source packet. That lookup found no joined public Stream or bridge reading revision; no reading identity is invented."},"spans":[{"id":"un-voice","text":"Giving workers more of a voice in bargaining over AI deployment is another way to steer toward augmentation without mandating it.","locator":"Retained extracted text, character offsets [32549, 32678). Context (EM paraphrase): Retained extracted text, character offsets [32549, 32678); verify against the cited snapshot, not current HTML offsets."}],"citation":{"label":"Unger and Narayanan · Steering AI Beyond Automation","aliases":["Steering AI Beyond Automation","Unger and Narayanan’s argument"],"context":"An argument for steering AI toward collaboration and giving workers more influence over deployment. Its proposed direction is distinct from measured effects.","spanIds":["un-voice"]},"readingRefs":[]},{"id":"brynjolfsson-ai-work-v2","title":"Generative AI at Work — arXiv v2","url":"https://arxiv.org/abs/2304.11771v2","publishedAt":"2024-11-06","dateBasis":"Version 2 submission date displayed by arXiv; first submitted 23 April 2023. The versioned abstract is the cited object.","access":{"status":"public-page","note":"Versioned author abstract checked directly. The HTML rendering displayed a conflicting 2026 manuscript date, so this edition cites only the clearly versioned abstract and does not assign that later date to the study.","checkedAt":"2026-10-30"},"provenance":{"kind":"external","workId":null,"snapshotId":null,"textHash":null,"reason":"Original public source read on 30 October 2026. An exact retained Stream identity has not been resolved for this edition."},"spans":[{"id":"resolved-work","text":"by 15\\% on average","locator":"Version 2 abstract, second sentence. EM context: staggered introduction among 5,172 customer-support agents; the outcome is issues resolved per hour, not company profit."},{"id":"novice-gains","text":"Less experienced and lower-skilled workers improve both the speed and quality of their output","locator":"Version 2 abstract, third sentence. EM context: benefits differ by experience; the most experienced workers show small speed gains and small quality declines."}],"citation":{"label":"Brynjolfsson, Li and Raymond · Generative AI at Work","aliases":["the customer-support study","customer-support research","customer-support evidence","customer-support results","customer-support comparison","Generative AI at Work"],"context":"A study of AI assistance for customer-support agents, reporting issues resolved per hour and differences by worker experience. This edition cites the arXiv v2 abstract.","spanIds":["resolved-work","novice-gains"]},"readingRefs":[]},{"id":"stanford-canaries-2026","title":"Canaries in the Coal Mine? — August 2026 revision","url":"https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/","publishedAt":"2026-08-12","dateBasis":"Publisher page identifies the revision as 12 August 2026; the linked PDF is dated August 2026.","access":{"status":"public-page","note":"Original research summary and linked August PDF abstract checked; the findings are attributed to the authors and their stated population.","checkedAt":"2026-10-30"},"provenance":{"kind":"external","workId":null,"snapshotId":null,"textHash":null,"reason":"Original public source read on 30 October 2026. An exact retained Stream identity has not been resolved for this edition."},"spans":[{"id":"hiring-first","text":"reduced hiring of young workers rather than increased separations","locator":"Publisher summary, fourth finding; linked August PDF, abstract. EM context: ADP payroll sample through June 2026. The authors find no widespread displacement, but a relative employment shortfall among ages 22–25 in AI-exposed occupations."},{"id":"descriptive","text":"early, descriptive indicators","locator":"Publisher summary, final paragraph; linked PDF, abstract. EM context: education controls attenuate the pattern; some divergent trends predate generative AI, and the ADP pattern is stronger than national survey benchmarks. These are not causal job-loss estimates."}],"citation":{"label":"Stanford · Canaries in the Coal Mine? · August 2026","aliases":["Stanford’s August 2026 payroll study","Stanford’s payroll evidence","Canaries in the Coal Mine?"],"context":"A descriptive analysis of ADP payroll data through June 2026, examining employment by age and AI exposure. Education, earlier trends and sample differences matter to its interpretation.","spanIds":["hiring-first","descriptive"]},"readingRefs":[]},{"id":"mollick-against-brain-damage-2025","title":"Against “Brain Damage”","url":"https://www.oneusefulthing.org/p/against-brain-damage","publishedAt":"2025-07-07","dateBasis":"JSON-LD datePublished on the original essay page.","access":{"status":"public-page","note":"Original essay page read directly.","checkedAt":"2026-11-13"},"provenance":{"kind":"external","workId":null,"snapshotId":null,"textHash":null,"reason":"Original public source read on 13 November 2026 to verify this position against its words. An exact retained Stream identity has not been resolved for this edition."},"spans":[{"id":"brain-outsource","text":"If you outsource your thinking to the AI instead of doing the work yourself, then you will miss the opportunity to learn.","locator":"Section “The Learning Brain”, first paragraph."},{"id":"brain-how-you-use","text":"it is how you use AI, rather than use of AI at all, that determines whether it helps or hurts your brain when learning.","locator":"Same section, after the classroom studies he cites."}],"citation":{"label":"Mollick · Against “Brain Damage”","aliases":["Against “Brain Damage”","Mollick’s learning essay"],"context":"An essay on when AI use helps or hurts learning, drawing on classroom studies.","spanIds":["brain-outsource","brain-how-you-use"]},"readingRefs":[]}],"positions":[{"id":"anthropic-workforce-authors","speaker":"Nathan Wilmers, Regina Catipon, Maxim Massenkoff and Peter McCrory","role":"Anthropic Institute researchers reporting on their employer","statement":"The authors report a falling junior share of technical hires and propose that intensive AI adoption at Anthropic may offer an early indication of wider workplace change.","conditions":["One rapidly growing AI company; technical hiring still grew in absolute terms.","Reported internal patterns and the authors’ interpretation do not identify an economy-wide causal effect."],"sourceRefs":[{"sourceId":"anthropic-workforce-2026","spanIds":["anth-junior"],"role":"supports"}],"statedAt":"2026-10-16","dateBasis":"Dated authored publication; this is not the date of every underlying observation or interview.","attribution":"direct"},{"id":"unger-narayanan-collaboration","speaker":"Gabriel Unger and Arvind Narayanan","role":"Authors of a research agenda and argument about AI development","statement":"They argue for deliberately developing collaborative AI and for workers having a stronger voice in deployment choices.","conditions":["This is a proposed direction and mechanism, not an observed estimate of its effects.","Agreement with their preferred outcome does not establish feasibility or labor-market results."],"sourceRefs":[{"sourceId":"steering-beyond-automation","spanIds":["un-voice"],"role":"supports"}],"statedAt":"2026-10-14","dateBasis":"Dated authored publication; this is not the date of every underlying observation or interview.","attribution":"direct"},{"id":"customer-support-researchers","speaker":"Erik Brynjolfsson, Danielle Li and Lindsey Raymond","role":"Researchers studying AI assistance in customer support","statement":"The researchers find higher resolved issues per hour after AI assistance, with the largest benefits for less experienced workers.","conditions":["One employer and an assisted workflow; the study does not estimate aggregate employment or full deployment returns."],"sourceRefs":[{"sourceId":"brynjolfsson-ai-work-v2","spanIds":["resolved-work","novice-gains"],"role":"supports"}],"statedAt":"2024-11-06","dateBasis":"Date of the cited version or source update.","attribution":"direct"},{"id":"stanford-entry-researchers","speaker":"Erik Brynjolfsson, Bharat Chandar and Ruyu Chen","role":"Researchers analyzing ADP payroll records","statement":"The authors find that young workers in AI-exposed occupations have fallen behind less-exposed peers, primarily through reduced hiring, while finding no widespread economy-wide displacement.","conditions":["Descriptive pattern through June 2026; education, prior trends and sample differences matter for attribution."],"sourceRefs":[{"sourceId":"stanford-canaries-2026","spanIds":["hiring-first","descriptive"],"role":"supports"}],"statedAt":"2026-08-12","dateBasis":"Date of the cited version or source update.","attribution":"direct"},{"id":"mollick-retain-the-learning","speaker":"Ethan Mollick","role":"Wharton professor and author of One Useful Thing","statement":"Mollick argues that people who outsource their thinking to AI miss the opportunity to learn, even with good intentions, because assistants tend to supply the answer. How AI is used, as a guided tutor rather than an answer engine, decides whether it helps or hurts learning.","conditions":["The studies he cites concern students in school settings; applying them to people learning a job is EM’s connection.","He describes the evidence for AI tutoring as still early."],"sourceRefs":[{"sourceId":"mollick-against-brain-damage-2025","spanIds":["brain-outsource","brain-how-you-use"],"role":"supports"}],"statedAt":"2025-07-07","dateBasis":"Dated authored publication.","attribution":"direct"}]}}}