Skip to main content
AI Engineering Hiring

Anthropic's junior share of tech hires fell to 9%. Should yours?

Calin Muresan•
#junior engineers#AI and hiring#Anthropic#engineering hiring#senior engineers#AI agents

Anthropic’s junior share of tech hires fell to 9%. Should yours?

By Calin Muresan, co-founder of Wise Step Recruiting. Calin spent about 20 years as a software engineer and cloud architect before he started recruiting engineers.

Not by default. Don’t cut your junior intake reflexively. Anthropic’s own data shows the junior share of its technical hires falling from 27% to 9% since early 2025, but that is one company at the leading edge of this shift, and its own report says none of the employer actions it describes directly address the slowdown in junior hires. Junior engineers used to be the obvious way into a growing tech org: hire a cohort, let the seniors mentor them up, repeat next year. At Anthropic, that front door has narrowed sharply, and a frontier AI lab has now published its own numbers instead of leaving it to speculation.

We are former software engineers who now recruit for Cloud & DevOps and Data & AI teams, and the question in this report is one clients ask us on most searches: if we stop building a junior bench, where do our seniors come from in three years? This piece covers what the report says, what it does not, what the wider market data adds, and a way to decide for your own team.

Key takeaways

  • In a technical report published October 16, 2026, Anthropic says the junior share of its technical hires fell from 27% to 9% since early 2025, and the junior share of technical headcount from 21% to 14%. Postings for junior roles fell “particularly sharply” after December 2025.
  • These are shares, not counts. Anthropic tripled in size in a year and says technical hiring kept growing in absolute terms. The report does not say how many fewer juniors it hired.
  • In an April 2026 survey, around a third of Anthropic’s ML managers and a fifth of its SWE managers said they were hiring less because of AI. Interview scorecards shifted toward autonomy and judgment, away from domain expertise.
  • The report admits none of the employer actions it describes “directly address the slowdown in junior hires.” Wider data points the same direction: Stanford finds employment of US workers aged 22 to 25 in AI-exposed jobs 19% below the path of their less-exposed peers.
  • Our read: the scarce hire is the senior engineer who can direct and verify agent output. Hire junior only where you have that senior capacity to review their work.

In this article: What the report says · Hires vs headcount · What managers hire for · Is this only Anthropic? · The pipeline question · Decision table · Limits

What Anthropic’s report actually says

Anthropic’s junior share of technical hires dropped by two-thirds, from 27% to 9%, while the junior share of technical headcount fell more slowly, from 21% to 14%. The source is How Employment is Changing at Anthropic, a technical report from The Anthropic Institute by Nathan Wilmers, Regina Catipon, Maxim Massenkoff, and Peter McCrory, published October 16, 2026. It draws on internal hiring records, calendar and Slack data, agent telemetry, employee surveys, and 17 interviews.

The figures that matter for a hiring manager:

What was measured Figure Window
Junior share of technical hires 27% → 9% Since early 2025
Junior share of technical headcount 21% → 14% Since early 2025
Junior postings Fell “particularly sharply” Since December 2025
Managers hiring less due to AI ~1/3 of ML, ~1/5 of SWE managers April 2026 survey
Survey answers describing a shift to managing agents 20% April 2026, open-ended question
Meeting time, technical individual contributors +16% Year over year
Agent hours per employee hour (employee-directed sessions) 6 to 1 By August 2026

Two details from the report change how you should read this table. First, the timing: internal discussion of falling demand for junior technical staff began in January 2026, and by April engineering and research teams “had deemphasized hiring for lower band technical positions.” Second, the split by function. In the report’s Figure 1, the junior share of non-technical hires sits at 26% at the end of the series, against 9% in technical departments. This is not a company-wide junior freeze. The report adds, descriptively, that junior employment slowed in jobs that use Claude for the most complex tasks (its Figure 5, January to July 2026). It does not claim Claude caused the split.

Junior share of Anthropic's technical hires and technical headcount, early 2025 to mid-2026 Two dumbbells on a 0 to 30 percent scale. Junior share of technical hires falls from 27 percent to 9 percent. Junior share of technical headcount falls from 21 percent to 14 percent. The hires line falls further and faster than the headcount line. Share of technical hires Share of technical headcount 0%10%20%30% 27%9% 21%14% Grey dot: early 2025. Coloured dot: latest.
Junior share of Anthropic's technical hires and technical headcount. Source: Anthropic, "How Employment is Changing at Anthropic," October 16, 2026, Figure 1 and text.

Why the gap between hires and headcount is the real story

Headcount is the stock; hiring is the flow. When the flow drops faster than the stock, the people already inside are carrying the share while the entry point narrows. The junior share of headcount fell 7 percentage points; the junior share of hires fell 18. Our reading, not the report’s: the people who joined as juniors in 2024 and 2025 are still there or have been promoted, so headcount dilutes slowly. The open-roles line in Figure 1 ends at 6% junior for technical departments, which suggests where hiring is heading.

Read the caveat with the same care. Anthropic says technical hiring “has continued growing in absolute terms,” and that the company tripled in size over the last year. A smaller share of a much larger intake may not mean fewer junior hires in absolute numbers. The report does not publish counts, and it does not define where “junior” ends. It also frames the junior pattern as descriptive, not causal.

That is why we would not write the headline many outlets ran. “AI is coming for junior jobs” is a claim about the whole market. This is a claim about one company, written by that company, about itself.

Latest junior share at Anthropic, technical versus non-technical departments Grouped horizontal bars. Hires: technical 9 percent, non-technical 26 percent. Headcount: technical 14 percent, non-technical 25 percent. Open roles: technical 6 percent, non-technical 13 percent. Technical departments are lower on every measure. HiresHeadcountOpen roles 9%26% 14%25% 6%13% Technical Non-technical
Junior share at the end of the series in Anthropic's Figure 1 (mid-2026). Source: Anthropic, "How Employment is Changing at Anthropic," October 16, 2026, Figure 1 end-point labels.

What Anthropic’s managers say they are hiring for now

The shift is not only fewer junior seats. It is a different job description: the report says hiring managers increasingly assess “autonomy and judgment,” while discussion of domain expertise shrank. Anthropic coded interview scorecards for thousands of hires since 2024. For technical candidates, the biggest increases were in comments on autonomy and judgment. Mentions of a candidate’s track record with AI tools tripled for technical roles, from around 5% of assessments in 2024.

Our reading of the rest of the report: human attention is becoming the bottleneck.

  • By August 2026, interactive agent hours in employee-directed sessions outnumbered employee hours about 6 to 1.
  • Interactive Claude sessions spend around 80% of their time idle, waiting for a person.
  • “Meta-work” rose from 10% to 22% of the average employee’s interactive sessions between fall 2025 and spring 2026. Much of the rise was agent management: specs for multi-agent runs, one agent reviewing another, and pipeline monitoring.
  • 20% of answers to an open-ended April 2026 survey question described a role shifting “from actually doing work to managing Claudes.”
  • Meeting time for technical individual contributors rose 16% in a year. Our read: the coordination agents cannot do lands on people.

Put plainly: the tasks that used to train a junior engineer, writing code and finding bugs, are tasks the report says could increasingly be offloaded to Claude. The work that grew, in our reading, is directing and verifying agents, which is senior work. We made a similar argument about the new applied AI engineer role, and about what AI-native engineering asks of a team.

Where engineering work moved at Anthropic, according to its report Two columns. Left, work absorbed by agents: writing code, finding bugs, routine tests and checks, logging progress, drafting emails. Right, work that grew for people: writing specs for multi-agent runs, setting up agent review loops, monitoring agent pipelines, coordination and meetings, judgment calls on systemic problems. A label between them reads: the training ground moved. Absorbed by agents Grew for people • Writing code • Finding bugs • Routine tests and checks • Logging progress • Drafting emails • Specs for multi-agent runs • Agent review loops • Monitoring pipelines • Coordination, meetings • Systemic judgment calls the training ground moved
Task categories as described in Anthropic's report; grouping and labels are ours. Source: Anthropic, "How Employment is Changing at Anthropic," October 16, 2026.

Is this only Anthropic?

No, but Anthropic is the leading edge, not the average. Independent data points the same direction across the US market, on different measures that are not directly comparable. Three sources worth knowing:

  1. Stanford Digital Economy Lab. Canaries in the Coal Mine?, revised August 12, 2026 with ADP payroll data through June 2026, finds employment of workers aged 22 to 25 in AI-exposed occupations 19% below where it would be had it kept pace with less-exposed peers. Experienced workers show no comparable gap, and the effect runs mainly through reduced hiring. The authors call these descriptive indicators, not causal estimates.
  2. Hosseini Maasoum and Lichtinger. Generative AI as Seniority-Biased Technological Change, a working paper on US résumé and posting data (SSRN 5425555, presented at the Stanford Digital Economy Lab), finds junior employment at firms adopting generative AI falling relative to non-adopters from early 2023, while senior employment holds up. Again, the mechanism is slower hiring, not separations.
  3. SignalFire. Its State of Talent Report 2026, published June 22, 2026, puts new-grad hiring about 65% below 2019 at 12 large tech companies and about 76% below at early-stage startups. It also says senior and staff individual-contributor roles are growing as a share of hiring.

None of these is Romanian or CEE data, and we did not find a reliable official count of junior IT vacancies in Romania. For the regional picture, see our take on IT recruitment trends in Eastern Europe. Treat the US numbers as direction, not a forecast for your market.

Where do your senior engineers come from in three years?

This is the question Anthropic’s report raises and does not answer. Its own words: none of the employer actions it describes “directly address the slowdown in junior hires,” and protections for incumbents “will not benefit those not hired to begin with.” The report also notes that some employees are “increasingly removed from the object-level work that originally built their skills.” That sentence is the pipeline problem in one line.

Here is the inversion underneath the headline. If the company building the tools is shrinking the junior share of its technical hiring this fast, and others follow, the market will see more experienced engineers and fewer entry-level openings, and fewer people getting the years of object-level work that made today’s seniors good at reviewing code. The senior engineer who can direct and verify agent output becomes scarcer over time, not cheaper. SignalFire’s note that the ratio of experienced candidates to openings is at a decade high describes today. It does not describe 2029, when today’s missing juniors would have been mid-level.

So we would not tell a client to stop hiring juniors. We would tell them that a junior hire is now only as good as the senior capacity around it. A junior with nobody to review their agent-assisted work is an expensive hire: their output lands on seniors who are already the bottleneck.

Should you still hire junior engineers? A decision table

Hire juniors when you have the senior review capacity and a real training path; hire the judgment-layer senior first when you do not. The table below is how we frame it with clients. It is our framework, built from the report’s findings, not Anthropic’s advice.

Your situation What we would do Why
Strong seniors, agents already in daily use, review capacity to spare Keep a small junior intake, designed around reviewing and directing agent output You can afford to train the people you will need in three years
Seniors already swamped reviewing agent output Hire one judgment-layer senior before any junior Adding juniors adds review load to the bottleneck
No senior who owns architecture or verification Hire the senior first, full stop Nobody can check what juniors and agents produce
Regulated or safety-critical systems Senior-heavy, juniors paired with named reviewers The cost of an unreviewed change is high
Team building its first agent workflows Mid-level engineers with demonstrated agent use Anthropic’s scorecards now mention AI-tool track record and autonomy far more often

If you are hiring the senior, test for the skill the report says managers now look for. Three questions we use in technical assessment for Cloud & DevOps and Data & AI roles:

  1. “Walk me through the last time an agent’s output looked right and was wrong. How did you catch it?” You are listening for a verification habit, not a tool name.
  2. “How would you split this task between yourself and two agents, and what would you check before merging?” Judgment shows in what they refuse to delegate.
  3. “What would you not let an agent touch in this system?” Seniors have a short, specific answer. Weaker candidates say “nothing” or “everything.”

For senior and leadership roles that never reach a job board, this is where headhunting passive candidates earns its fee.

What we see on Cloud & DevOps and Data & AI searches

Our searches run mid-level to executive, so we see the senior side of this market, not the graduate side. The pattern there matches the report: clients ask less for “a good coder” and more for someone who can own a system and check work they did not write. We are a two-founder firm, so we will not dress this up as a dataset. What we can say is that the bench question, where seniors will come from, now comes up early on most searches, usually before salary.

We also run our own work on agents, which is why we read this report with recognition. In our searches, AI handles the scale: sourcing, matching, the first pass. We handle the judgment: the brief, the technical assessment, the offer conversation. That is a similar split to the one Anthropic describes in its engineering teams, and it is the split we screen candidates for. For how we apply it to recruiting itself, see whether a recruiting team should use coding agents.

Where this argument is weakest

Anthropic is an atypical employer for this question, and its data comes with limits it states itself. Four to keep in mind:

  1. Shares, not counts. The company tripled in size. A falling junior share of a growing intake is not proof of fewer junior hires.
  2. Self-reported. Every figure is Anthropic’s, about Anthropic. The report does not describe any independent audit of the hiring data.
  3. Correlation, not cause. The report calls the junior pattern descriptive. Adoption of AI tools was not random, and the company’s rapid growth is a confounder.
  4. No definition of “junior.” The report never says where the junior band ends, so the 27% and 9% may not map onto your own levels.

We could also be wrong about the three-year shortage. If agents get good enough at verification, the senior review bottleneck shrinks, and so does the case for hiring seniors to remove it. We do not see that in 2026 data. Anthropic’s own data points the other way: meta-work rose from 10% to 22% of interactive sessions, and meeting time for technical individual contributors rose 16%.

Frequently asked questions

Did Anthropic stop hiring junior engineers?

No. The junior share of its technical hires fell from 27% to 9% since early 2025, and the report says technical hiring kept growing in absolute terms while the company tripled in size. Junior hiring slowed sharply as a share; the report does not publish absolute counts.

Why did Anthropic hire fewer junior technical staff?

The report links the shift to Claude taking on code writing and bug finding, which it says were increasingly offloaded in the teams that relied on Claude most. By April 2026, engineering and research teams had deemphasized lower-band technical hiring. The report presents this as descriptive, not proven cause.

Should my company still hire junior developers in 2026?

Yes, if you have senior engineers with time to review agent-assisted work and a real training path. If your seniors are already the bottleneck, hire a senior who can direct and verify agent output first, then add juniors.

What skills are AI-heavy companies hiring for now?

At Anthropic, interview scorecards shifted toward autonomy and judgment, while discussion of domain expertise shrank. Mentions of a candidate’s track record with AI tools tripled for technical roles compared with 2024.

Is junior tech hiring falling across the whole market?

The direction is consistent in US data. Stanford finds employment of workers aged 22 to 25 in AI-exposed occupations 19% below the path of their less-exposed peers, mainly through reduced hiring. SignalFire puts new-grad hiring at large tech companies about 65% below 2019. Anthropic’s figures measure something different, a share at one company, so they are not directly comparable and do not describe the market.

The bottom line

Junior engineers used to be the default entry point. At Anthropic, the junior share of technical hires fell from 27% to 9%, and the company’s own report says none of the employer actions it describes directly address the slowdown.

Three things to take from it:

  1. Read the numbers as shares at a company at the leading edge, not proof the market already looks like this.
  2. The scarce hire is the judgment layer: senior engineers who direct and verify agent output.
  3. Hire juniors where seniors can review them. Without that capacity, a junior is a cost, not a bench.

If you are hiring Cloud & DevOps or Data & AI engineers in Romania or the wider CEE region, mid-level to executive, we are former software engineers who assess candidates on exactly that judgment. You get a curated shortlist of three to five, with first candidates typically in 10 to 15 days. Brief us on the senior you need.


Last updated: October 20, 2026. Sources verified October 20, 2026: Anthropic, “How Employment is Changing at Anthropic” (Wilmers, Catipon, Massenkoff, McCrory; The Anthropic Institute technical report, Oct 16, 2026); Stanford Digital Economy Lab, “Canaries in the Coal Mine?” (Brynjolfsson, Chandar, Chen; revised Aug 12, 2026); Hosseini Maasoum and Lichtinger, “Generative AI as Seniority-Biased Technological Change” (SSRN working paper 5425555; Stanford DEL seminar page); SignalFire, State of Talent Report 2026 (Jun 22, 2026). Wise Step process figures are our own operating averages.