Analysis
Anthropic disclosed in a report published Friday that junior-level employees have fallen from 21% to 14% of its technical departments since early 2025, and that junior hires have dropped even further -- from 27% to just 9% of new technical hires -- according to TechCrunch. Technical staff overall shrank from 61% to 56% of the company as business functions expanded, which Anthropic attributes partly to coding tools letting non-technical employees build their own software.
The disclosure lands as Anthropic, which Pulse has previously covered extensively, scales at a pace few enterprise software companies have matched, with Pulse reporting this week that the company is targeting a roughly $2 trillion IPO valuation on Q3 revenue that grew 46% to nearly $17 billion. A company adding revenue and headcount that fast would normally staff up broadly across seniority levels; Anthropic's data suggests it is instead concentrating hiring at the senior end and using its own Claude Code tooling to cover work that used to require junior engineers and analysts.
## How much has Anthropic's hiring mix actually shifted? The starkest number is the hiring rate, not the headcount snapshot: nine out of every hundred new technical hires are now junior, down from 27 a year and a half ago. That is a faster shift than the stock figures (21% to 14%) suggest, because it reflects where marginal hiring dollars are going right now rather than the accumulated workforce. Other labs, including OpenAI and Google DeepMind, have not published comparable breakdowns, so there is no clean read on whether Anthropic's shift is company-specific or an early data point for the whole frontier-lab tier.
“The starkest number is the hiring rate, not the headcount snapshot: nine out of every hundred new technical hires are now junior, down from 27 a year and a half ago.”
The meetings finding cuts against the cleanest AI-productivity narrative. Calendar data shows meeting time per employee up 20% year over year, and up 16% for technical contributors specifically -- the opposite of what you'd expect if AI agents were simply absorbing busywork. Anthropic frames this as a feature, crediting the increase to "human trust and accountability" as more decisions get delegated to Claude instances that still need human review.
For founders and GPs evaluating AI-leverage pitches, this is a useful corrective. An April 2026 internal survey found only 27% of Anthropic's own employees said Claude made them more productive, while 20% said their work had shifted from doing tasks to managing multiple Claude instances -- one unnamed engineer described it as "babysitting," since each instance demands separate attention. That is a company that builds the product reporting real coordination costs, not just gains.
However, all of this is self-reported, internal, and unaudited by any outside party, which is the risk in taking it at face value. Anthropic has an incentive to frame rising meeting time as "trust and accountability" rather than friction, and the 27%-said-more-productive figure leaves open what the other 73% experienced. More code shipped per person is not the same as better code, and Anthropic's own report concedes it has not ruled out quality tradeoffs, only that it hasn't yet detected decline.
What to watch: whether other frontier labs publish similar internal breakdowns once investors start asking the same headcount-mix question on IPO roadshows, and whether the junior-hiring pullback shows up in entry-level AI engineering postings across the broader market in the next two quarters.
