Key Takeaways
- Anthropic's chief economist Peter McCrory published two internal reports on October 16, 2026, covering hiring changes inside the company and staff expectations for AI's economic impact.
- Technical entry-level hiring at Anthropic fell from 27 percent to 9 percent of technical hires, and the share of employees working in technical roles also declined.
- 91 percent of Anthropic employees think AI could produce a Nobel-prize-level discovery by 2030, compared with 39 percent of US adults, a gap the report ties to hands-on experience.
- AI agents passed employees in total work hours in March 2026 and reached six times employee hours by August 2026, according to the company's own tracking.
Quick Facts
| Type | Internal research report in two parts |
|---|---|
| Devices | Desktop and phone browsers |
| Price | Free, no account or app required |
| Languages | English; browser translation works |
| Highlight | 91% of staff expect Nobel-level AI by 2030 |
| Learning curve | Easy to read; data needs context |
| Category | Tech Buzz |
| Creator | Peter McCrory |
| Official site | anthropic.com ↗ |
| Heat | 🔥 275.9K Heat |
| Last updated |
What the Anthropic Internal Expectations Report Actually Is
On October 16, 2026, Anthropic's chief economist Peter McCrory published two internal reports that examine how AI is reshaping work inside the company and how far its own staff expect the technology to go. The first tracks internal employment shifts, including the changing mix of roles and seniority. The second compares Anthropic employees' expectations for the economy with those of US adults. Both are readable in the research section of anthropic.com, free of charge and without an account.
The numbers are what pulled readers in. Anthropic says 91 percent of its staff believe AI could produce a Nobel-prize-level discovery by 2030, while only 39 percent of US adults expect the same. Separately, technical entry-level hiring inside the company fell from 27 percent to 9 percent. The report also tracks AI agents, noting that agent work hours passed employee work hours in March 2026 and reached six times employee hours by August 2026.
That combination is unusual. Most public debate about AI and jobs leans on outside forecasts, academic models or anecdotal reporting. Here, a frontier AI lab is publishing its own hiring data next to its employees' beliefs about the future. Readers who look this up are usually trying to answer one question: does the company building the technology actually behave the way its own predictions suggest? The report is a partial, self-reported answer, and it is worth reading with that framing in mind.
- Two reports in one place
The employment findings and the expectations survey sit together under Anthropic's Team Economics work, alongside topics such as labor market, productivity, agents and economic futures.
- First-party hiring data
The 27 percent to 9 percent figure comes from Anthropic's own recruiting and headcount records rather than an outside survey panel, which makes it narrow but unusually concrete.
- A named author
Peter McCrory, Anthropic's chief economist, is credited on the report, and the companion article How employment is changing at Anthropic is linked from the same area.
- No cost, no signup
The page opens in a normal browser on desktop and phone, with no app install, no account and no payment required.
Inside the Two Halves: Hiring Data and the Expectation Gap
The first half of the report looks inward at Anthropic's own workforce. It reports that technical entry-level hiring dropped from 27 percent to 9 percent of technical hires, and that the share of employees working in technical roles also declined over the period studied. Read carefully, this describes one company's recruiting mix, not the whole labor market. Anthropic is a relatively small, fast-growing lab, and its hiring choices reflect its own product priorities as much as any industry-wide trend.
The second half turns outward. Anthropic surveyed its staff and compared the results with data on US adults, finding that 91 percent of employees think AI could deliver a Nobel-prize-level discovery by 2030, against 39 percent of the general adult public. McCrory's analysis frames that 52-point spread as a difference in hands-on experience with frontier models rather than simple optimism versus pessimism. People who use the most capable systems daily see capability growth differently from people who rarely touch them.
The agent data is the third thread. According to the report, AI agents passed employees in total work hours in March 2026, and by August 2026 agents were logging six times the hours of the human workforce, with productivity improving alongside. That is a striking claim, and it depends heavily on how the company defines an agent's work hours. Treat the direction of travel as the signal and check the official site for the exact methodology before quoting the multiple.
- The 27 to 9 shift
Technical entry-level hiring fell from 27 percent to 9 percent of technical hires, a change the report attributes to how the company's work and tooling have evolved.
- 91 percent versus 39 percent
Anthropic staff are far more bullish than US adults about Nobel-prize-level AI discoveries by 2030, a 52-point gap the report links to direct experience.
- Agents overtake employees
Agent work hours passed employee hours in March 2026 and reached six times that level by August 2026, according to the company's own tracking.
- What it does not claim
The report does not say AI caused the hiring shift, and it does not present its staff as a representative sample of workers or the public.
How to Read the Report and Get the Most From the Data
Start in the research section of anthropic.com. The page is a standard web article, so it opens on any modern desktop or phone browser with no app, no account and no payment. The publication date, October 16, 2026, appears near the top, followed by topic labels such as labor market, productivity, how people use AI, agents and economic futures. It is written in English, so if that is not your first language, turn on your browser's built-in translation.
Read the two halves as one argument. The hiring data shows what Anthropic is doing; the survey shows what its people expect. The gap between 91 percent and 39 percent is best understood as a measure of exposure, not a verdict on who is right. Ask yourself how the questions were worded, who counted as a US adult, and what counts as a Nobel-prize-level discovery. Those definitions shape the headline numbers more than most readers assume.
Then widen the lens. The same area of the site links to the companion article How employment is changing at Anthropic and to other Team Economics research on labor markets, productivity and how people use AI. Reading two or three of those together gives you a much better sense of the company's internal picture than any single figure. If you plan to cite the report in your own writing, quote the specific metric and its date rather than a general claim.
- Translate, do not skim
Browser translation handles the English text well, but keep the original open for numbers and labels so percentages and dates stay unambiguous.
- Anchor on dates
The report is dated October 16, 2026, and the agent figures are tied to March and August 2026, so always attach the month when you repeat them.
- Separate fact from interpretation
The hiring percentages are internal records; the 91 versus 39 comparison is survey-based; the agent hours are company tracking. Each deserves a different level of confidence.
- Follow the thread
The companion article on employment at Anthropic and the wider Team Economics topics let you see whether the pattern holds across more than one report.
What It Means for Job Seekers, Developers and the Wider AI Debate
For anyone entering the tech job market, the 27 percent to 9 percent shift is the number to sit with. It suggests that at least one leading AI lab is hiring fewer people into junior technical roles and leaning more on experienced staff and automated systems. That does not mean entry-level technical work is disappearing everywhere, and the report covers a single employer. It does mean that candidates should expect to demonstrate judgment, evaluation and tool fluency earlier in their careers.
For developers and product teams, the agent metric is the practical takeaway. If agents can accumulate several times the working hours of the humans who supervise them, then the scarce skills become orchestration, review and measurement: deciding what to delegate, checking output quality and knowing when an agent is quietly failing. The report claims productivity improved as agent hours grew, but it does not publish a full breakdown, so treat that as a claim to verify rather than a settled result.
The open questions are the interesting part. This is self-reported data from one company, with no independent audit, and the phrase Nobel-prize-level discovery is broad enough to mean very different things to different readers. The expectation gap could narrow as more people use capable AI tools, or it could persist if access and trust stay uneven. What to watch next: whether Anthropic repeats the survey, whether the entry-level hiring share keeps falling, and whether other labs publish comparable internal numbers.
- If you are job hunting
Assume interviewers care about how you work with AI tools, not just what you know; evidence of directing and checking AI output is becoming a differentiator.
- If you build with AI
Track agent hours and cost the way you track human hours, and define what a completed task looks like before you scale up delegation.
- If you follow AI policy
Remember the sample: Anthropic employees are not a proxy for the workforce, and US adults are not a proxy for the world.
- Where the report stops
It offers no forecast of total employment, no salary data and no external verification, so pair it with labor market statistics before drawing broad conclusions.
Step-by-Step Guide
Open the research section of anthropic.com
Go to anthropic.com in any modern browser and look for the research section, then find the expectations report dated October 16, 2026. No account, app or payment is needed, and the page works on both desktop and phone.
Read the employment findings first
Start with the half that covers hiring inside Anthropic, including the drop in technical entry-level hiring from 27 percent to 9 percent. Note the period covered and remember that these are one company's internal records, not national statistics.
Then read the expectations survey
Move to the comparison between Anthropic staff and US adults, where 91 percent of employees expect a Nobel-prize-level AI discovery by 2030 against 39 percent of adults. Read how the questions were framed before you decide what the gap means.
Check the agent numbers and their definitions
Find the section on AI agents, which reports that agent work hours passed employee hours in March 2026 and reached six times that level by August 2026. Look for how the company defines an agent's working hours before repeating the multiple.
Follow the companion research
Open the related article How employment is changing at Anthropic and browse the other Team Economics topics such as labor market, productivity and economic futures. Reading them together shows whether the patterns in this report repeat across the company's other work.
Frequently Asked Questions
Is the Anthropic expectations report free to read, and do I need an account?
Yes, it is free. The report lives in the research section of anthropic.com and opens in any modern browser on a desktop or phone. There is no app to install, no Anthropic account to create and no payment involved. If English is not your first language, your browser's translation feature can render the page in another language.
Who published it, and who is Peter McCrory?
Peter McCrory is Anthropic's chief economist, and he published the two internal reports on October 16, 2026. The work comes from Anthropic's Economics team, which also produces the companion article How employment is changing at Anthropic and other research on labor markets, productivity and how people use AI. The company is the publisher, so the findings are self-reported.
What does the 91 percent versus 39 percent gap actually mean?
It compares two very different groups. Anthropic employees work with frontier AI systems daily, and 91 percent of them think AI could produce a Nobel-prize-level discovery by 2030. Among US adults, 39 percent expect the same. The report interprets the 52-point spread as a difference in direct experience with AI capability rather than a simple split between optimists and pessimists.
Does the report prove that AI is replacing junior technical jobs?
No. It shows that technical entry-level hiring at Anthropic fell from 27 percent to 9 percent of technical hires, which is a real shift at one company. The report does not claim AI caused it, and it does not measure the wider job market. Use it as a signal about how one leading lab staffs itself, not as proof of a global trend.
Can I read it on my phone, and is there anything to download?
The report is a standard web page, so it reads fine on a phone browser, though you may want to adjust text size for the tables and percentages. There is no app and no required download. If you prefer a translated version, use your browser's built-in translation rather than assuming an official localized edition exists; check the official site for that.
