Paper: No Major AI Effect Yet On Jobs For 2026 Grads
LLM project finds discrepancies in most economics papers
September 28, 2026
D.A.D. today covers 7 stories — about a 6-minute read. What's New, What's Innovative, What's Controversial, What's in the Lab, and What's in Academe.
The Daily AI Digest is a daily AI briefing automated by Alexander Panetta — a veteran political journalist tracking the field during a Master's in AI Management at Georgetown University.
D.A.D. Joke of the Day: I told ChatGPT it was wrong. It apologized, thanked me, and said I was absolutely right. I've never felt less sure of anything.
What's New
AI developments from the last 24 hours
Trump Hosted the CEO He Blacklisted
President Trump had Dario Amodei to the White House for a private dinner on Sunday evening — their first one-on-one meeting, scooped by Maria Curi and Marc Caputo of Axios. Speaking to reporters at Joint Base Andrews hours earlier, Trump called the Anthropic chief executive "highly respected."
Three weeks ago he called him something else. "The Trump Administration has stopped AI 'people' from doing bad, or potentially bad, 'things,'" Trump posted, "like Dario (Anthropic!), who is now pretending to be a 'perfect little angel'" (D.A.D., September 15).
The blacklist has not moved. In February the administration ordered every federal agency to stop using Anthropic's technology after the company refused Pentagon demands to drop its red lines on autonomous weapons and domestic surveillance, and Defense Secretary Pete Hegseth designated it a supply-chain risk — a label normally reserved for arms of a hostile state. Two days before the dinner, a federal appeals court upheld that designation 2-1 (D.A.D., September 26). A separate federal court in San Francisco had already struck down the parallel one as illegal.
The invitation followed a snub that was not one. Amodei missed last week's state dinner for Xi Jinping over a scheduling conflict; Sam Altman and Sundar Pichai were in the room. Trump invited him personally afterward. Months ago, some advised against a one-on-one because "Dario's a little too weird" for Trump, a senior official told Axios, and last week Trump allies circulated talking points emphasising Amodei's ties to effective altruism. On Saturday, SNL mocked him over his safety warnings.
Neither side has said what was discussed. The White House offered only a statement to CNBC: "President Trump has been clear: America will lead the world in Super Intelligence, while protecting American consumers."
Why it matters: Sunday was the warm-up. On Tuesday, Trump and House Speaker Mike Johnson meet the top AI chief executives at the White House, and that is where anything decided will show. Until then, hold Anthropic's federal position in your head as it actually stands: barred by executive order, blacklisted by the Pentagon, upheld in one court, struck down in another, and dining at the White House. Any institution trying to work out whether it can put Claude near government work is reading the same contradictory signals you are — which is the real state of American AI policy, whatever Tuesday produces.
Sources: Axios — Maria Curi and Marc Caputo · Forbes · CNBC
AI Agents Won't Break the Banks. They'll Make Them Pay You More.
Torsten Slok, chief economist at Apollo Global Management, published a one-page note Sunday headlined "Is an Agentic Bank Run Coming?" Muse and similar agents, he wrote, "could soon sweep household cash automatically into accounts paying 3.3% to 5.0%, instead of the 0.1% national average on checking accounts." If every household did it, banks "could lose a large share of the cheap deposits they rely on to make loans, which would be a problem for the entire financial system."
The pushback came fast, and it is worth more than the warning. Cullen Roche — whose primer on the monetary system is among the most-downloaded papers on SSRN — points out that the money does not go anywhere. Cash moving from checking at Bank A to savings at Bank B lowers one balance and raises the other; system-wide deposits are unchanged. Most of the fintechs on Slok's own chart are banks, or sweep cash into partner banks. "The money never leaves the system. It just changes addresses."
And the premise is backwards, he argues: banks do not lend out deposits, loans create deposits — which is not a contrarian view but the Bank of England's, stated plainly in its 2014 bulletin. Banks need deposits to settle payments and hold reserves. Losing cheap ones means buying pricier ones. "That's a real cost, but it's a cost, not a collapse."
Two things survive the rebuttal, and Roche concedes both. Silicon Valley Bank was a run from one bank to other banks, and it still killed the bank; if agents can move uninsured deposits in hours rather than days, individual banks fail faster. And sticky low-rate deposits are how banks manage interest-rate risk. Take the stickiness away and weak balance sheets get exposed. "That's a legitimate risk for specific banks with bad balance sheets. It is not a system-wide run."
Worth noting the tool can't do it yet, either: Meta's Muse reads across 12,000-plus US financial institutions through Plaid and cancels subscriptions, but Plaid's announcement says nothing about moving money.
Why it matters: Roche's conclusion is the sentence to keep. "Calling it a bank run confuses a repricing of bank funding with a panic driven collapse of the system. Those are very different animals." The likely outcome is that banks pay savers closer to market rates and their margins get squeezed — bad for bank shareholders, good for anyone holding cash. Which is the part that generalizes. High-yield accounts have paid multiples of checking rates for years and most Americans never opened one; that gap is friction, and friction is what agents delete. Any revenue built on customers not bothering — auto-renewals, legacy tariffs, the account that quietly stopped being competitive — is income collected from inattention. Muse already cancels subscriptions. Amazon already locked it out of its store (D.A.D., September 22). The question is not whether your customers could get a better deal elsewhere. It is what happens the first morning something checks for them.
Sources: Apollo — Torsten Slok · posts by Cullen Roche on X · Bank of England · CNBC
Court Filings Allege OpenAI Knew of Piracy Risks, Proceeded Anyway
Newly released court filings in the Authors Guild's copyright lawsuit against OpenAI and Microsoft allege internal documents show employees knew the risks and proceeded anyway. Plaintiffs claim the records reveal a 'mass piracy scheme,' including alleged use of pirated books from what one internal message called a 'sketchy Russian website,' plus a reference to releasing products 'that will make people unemployed.' A hearing is expected in early 2027.
Why it matters: This is one of the highest-stakes AI copyright cases. If courts accept plaintiffs' claim that executives knowingly risked harm, it could shape how courts and lawmakers treat AI training data.
Discuss on Hacker News · Source: authorsguild.org
What's in Academe
New papers on AI and its effects from researchers
AI Re-Ran 4,452 Economics Studies. Three-Quarters Had Discrepancies.
Economists at Harvard and MIT built an AI workflow that automatically re-runs published research using the same data and code the original authors released, then checks the results, speeds up the analysis, and tries new extensions of the study. Tested on 4,452 replication packages from five economics journals, it flagged discrepancies in roughly three-quarters of them, cut computation time by 10x or more in 496 cases without losing accuracy, and generated 923 novel extensions that stayed faithful to the original paper's goals.
Not every discrepancy is an error, however. About a third of the calculation mismatches were within rounding of the last printed digit, though roughly a quarter reached a reported number's first significant digit. When two Claude models graded a sample of 100 of the reproductions, about 83% of gradings found the results reproduced with only minor differences. One of the authors worked on the project as a contractor for Anthropic.
Why it matters: Academic economics has long relied on slow, manual replication to catch errors—this suggests AI could make auditing published research routine, though the high discrepancy rate also raises uncomfortable questions about how much existing literature would hold up to the same scrutiny.
AI Study Finds Kids' Books Show Happy Pictures, Sadder Words
Researchers used machine learning and generative AI to scan emotional content in public-school textbooks and children's literature, comparing what the text says versus what the illustrations show. The mismatch was stark: nearly half of analyzed pages had zero emotional overlap between text and images, with pictures skewing overwhelmingly happy and calm regardless of the emotions in the text. The pattern held across time periods, contexts, and identity groups. Purchase and library data link it to consumer demand for upbeat cover imagery, which may be narrowing kids' exposure to emotional nuance.
Why it matters: The study shows AI's growing use as a research tool for auditing bias in everyday materials—not just algorithms—and raises questions about whether market-driven publishing choices are shaping children's emotional literacy.
Stock Prices Imply AI Coding Tools Lift GDP by Up to 6.5%
A new working paper puts a number on how much AI coding tools are worth to the economy. The authors measured how sensitive firms' stock returns are to an AI stock index, relative to their software-engineering payroll share. They estimate that from November 2022 to December 2025, AI raised markets' expected software engineering productivity by the equivalent of a permanent 32.6% gain, or a 3.6%-6.5% lift in GDP. By mid-2026, the authors say, the effect had more than doubled as coding agents improved.
Why it matters: Markets, not surveys, are pricing in AI's economic impact. By this measure, investors expect coding productivity gains to add a measurable slice of GDP, not just a story tech companies tell about themselves.
Summer 2026 Brought No AI Jobless Spike for New Grads, Economists Find
UCLA economists used government labor survey data to test whether workplace AI adoption drove up unemployment among recent college graduates in summer 2026. They found no such effect. Unemployment among new grads didn't spike relative to prior summers, older graduates, or young workers without degrees. That held even using an expanded definition of joblessness. In one wrinkle, unemployment showed a modest positive link to jobs with high remote-work availability. A potential effect of AI was not ruled out, however. On the expanded measure, which also counts graduates who want a job but aren't actively searching, the paper found summer 2026 unemployment of 10.4%, the highest of the five summers studied — though only 0.3 points above the previous high, a rise the authors did not find statistically significant.
Why it matters: The study complicates the narrative that AI is already gutting entry-level white-collar hiring, suggesting fears about a graduate jobs crisis may be running ahead of the data.
What's Happening on Capitol Hill
Upcoming AI-related committee hearings
Wednesday, September 30 — Hearings to examine rogue AI, focusing on securing the homeland against AI agents. Senate · Senate Homeland Security and Governmental Affairs Subcommittee on Disaster Management, District of Columbia, and Census (Open Hearing) 342, Dirksen Senate Office Building
What's On The Pod
Some new podcast episodes
The Cognitive Revolution — AI:AM: What If It Works Too Well? Colluding Agents, $200M Safety Orgs, Virtual Cells Saturate at 2%
The Cognitive Revolution — What is Utopia? Presenting The Receipt Horizon, by Joel Borgen – Chapters 1–4