Product1 distinct publisher3 min readPublished
Vendr says postings that drew up to 100 applications now draw more than a thousand. Greenhouse says recruiters who once wanted one-click applying are now asking for it to be harder.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
Two thousand applicants in a day is not a screening problem at any speed a person works at. Take the 30-second skim that Greenhouse's Ophir Samson says recruiters are already budgeting per application [6] and apply it to the 2,000 applicants in 24 hours he describes one role attracting [7]: that is roughly 17 hours of reading for the first day of one job [2]. Put two readers on it and you have two people doing eight and a half hours each. A faster filter does not change the size of the pile.
The multiplier is worth pinning down, because someone is advertising it. Andrew Stockwell's postings at Vendr went from a few dozen applications, up to 100 on a good day [1], to hundreds within 48 hours and sometimes past a thousand [2]. That is about a tenfold increase [3], and 10x is precisely what the auto-apply vendors sell: JobAssist, Sonara and Ladder's Apply4Me promise 10x as many applications with less effort than one manual application [11]. The cost those products drive toward zero is applicant effort. So friction priced in applicant effort is the exact variable they exist to absorb, and "make it harder" only bites if harder means something a script cannot do on the candidate's behalf.
Which makes the provenance of the friction pitch awkward. Samson runs voice AI at Greenhouse, which acquired his startup earlier this year, and that startup used AI to conduct job interviews [5]. The proposed cure for a funnel full of machine-written applications is being assembled by people shipping machine-conducted interviews. It may well work as a gate, since a live conversation is harder to autofill than a form. It is still automation answering automation, with the candidate on the far side of both.
Underneath sits a market that stopped resembling the one the funnel was built for. Openings peaked at a record 12.3 million in March 2022 and have run near 7 million since mid-2024 [10], about 43 percent below that peak [1]. Meanwhile LinkedIn puts applications up 22 percent since ChatGPT's release [c12b], which arrived in late 2022 [15]. Fewer doors, more knocking, and each knock cheaper than the last. Jane Curran of JLL, who remembers a market where people could hold three offers in an afternoon, calls the current one the polar opposite [14].
Stockwell's own description of the fake candidates and AI-padded resumes in his queue [3] is the tell. Nothing in the intake path was ever metered, because for years nobody needed it to be. Friction is not a strategy here. It is a meter being retrofitted to a pipe that was deliberately widened, by the same industry that widened it.
Ranked by verification strength, evidence, and original report placement.
Andrew Stockwell, head of people at the software-buying company Vendr, says that three years ago he would post a listing, wait a few days, and review a few dozen applications, up to 100 if he was lucky.
Stockwell says that sometime in the last year or so, within a day or two of posting he was flooded with hundreds of applications, and sometimes the incoming volume topped a thousand.
Stockwell says a good chunk of the incoming applications were "total bogus" fake candidates, and many more appeared to have been written using AI, stretching the truth and making applicants hard to distinguish from one another.
Stockwell says he found himself paying his talent-acquisition professionals, whom he describes as high-paid, high-quality individuals, just to look through applications all day long.
Ophir Samson is head of voice AI for the recruiting platform Greenhouse, which he joined earlier this year when it acquired his startup, a company that used AI to conduct job interviews.
According to Samson, recruiters spend hours in applicant tracking systems tweaking filters to try to determine whose applications are worth a 30-second skim.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Named practitioners plus one platform statistic, one publisher
Claims rest on on-record executives at Vendr, Greenhouse and JLL, a BLS openings series, and LinkedIn-supplied growth figures — better than anonymous sourcing, but the whole cluster is one article, the volume figures are recollections rather than exported system data, and the LinkedIn percentages come with no methodology.
Platform-level shipping and volume disclosures, thin firm-level breadth
Adoption of the underlying behaviour is documented at scale by LinkedIn's volume disclosures, and the counter-response is already shipping (application limits, an underqualified-applicant nudge rolling out this month) with an ATS incumbent acquiring AI interviewing. Firm-side evidence of actually adding friction is limited to two employers' statements of intent.
Framing outruns the measurement
The 'hiring is fundamentally broken, recruiters want friction' framing and the vendors' '10x applications' marketing are stronger than what is measured: friction demand is two executives' testimony, the quality-degradation claim is unquantified, and the only hard numbers (LinkedIn growth, BLS openings) are consistent with more volume without establishing that screening outcomes have worsened.
Most quantified sources sell into the problem they describe
Greenhouse sells recruiting software and just bought an AI-interview startup; LinkedIn both supplies the growth statistics and monetizes the hiring funnel while shipping friction features; the auto-apply vendors' 10x claim is marketing copy. The employer voices (Vendr, JLL) and a former HR executive with an 800,000-follower audience have reputational or audience incentives of their own, though the BLS series is disinterested.
Directionally credible, single-publisher and largely testimonial
The direction — more applications per opening, fewer openings, platforms adding friction — is supported by a government series and shipped product changes, so it is unlikely to be wrong in direction. Magnitude and consequences are weakly grounded: one publisher, recollected employer counts, vendor-supplied statistics, and no measurement of hire quality.
security
North Korea's hiring funnel: 60 applications a day, 22 personas, ten jobs landed1 distinct publisher
product
Rillet's $100M reads as proof mid-market ERP is rip-and-replace, mostly at the cheap end1 distinct publisher
science
Text watermarks land on 2 December. The detection they imply does not.1 distinct publisher
invest
A Connecticut judge just priced prompt injection: no fine, no e-filing2 distinct publishers
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 25, 2026