Skip to content

AI, decoded

Is AI taking software engineering jobs?

The hiring data discussed with Greenhouse CEO Daniel Chait points to a strained matching layer, though application-funnel figures alone cannot show whether there are fewer jobs. Figures cited in the episode put applications up 239% since ChatGPT while 75% fewer reached the hire stage, and software engineers are the heaviest users of the automation driving that. Much of it is a signal problem, and each side makes it worse by responding rationally to the other. Separate research finds employment for young workers in AI-exposed occupations lagging their peers, a pattern its authors do not yet attribute to AI. Chait told Chain of Thought that an individual recruiter may be looking at four or even five times as many applications as just a couple of years ago.

· Updated · Chain of Thought

Level 3: Leading with AI · 3.6 Teams and skills

AI CodingAI Engineering

  • 5 episodes
  • 6 guest clips
  • 6 glossary terms
A haystack with a blue needle half hidden in it, beside the hook: a needle made to look like hay. It's a signal problem.

In detail

1. What the doom loop is

Daniel Chait runs Greenhouse, which more than 7,500 companies hire through, and he named the pattern. Candidates use AI to find and apply to more jobs and to tailor every resume to the posting. Recruiters, who may each be looking at four or five times the applications they saw a couple of years ago, point an off-the-shelf model at the inbox and ask which 20 of the thousand to call. That makes it harder to get seen, so candidates send more. Both responses are individually reasonable and jointly ruinous.

2. Engineers are the heaviest auto-appliers

By Greenhouse’s own data, software engineers automate applications more than any other group and send the most. Chait is careful that this does not make them the villains: they are early adopters doing what the tooling makes easy. It does mean engineering roles feel the signal collapse acutely.

3. The signal collapsed

Chait’s read is that more postings alone would not fix it. When every resume is tailored by a model to the job description, resumes converge, and a thousand near-identical applications carry less information than a hundred varied ones. Layer on the trust problem. A Greenhouse survey cited in the episode found 91% of recruiters have spotted candidate deception and 41% of candidates admit to using prompt injection to get past AI filters, and the episode cites resume hacks like white-fonting and injection up 500%. Screening turns more defensive in response, which attenuates the signal further.

4. The job did change, which is a separate claim

Chait grants the part that is real: software engineering looks different than it did three years ago, and different than it did sixty days ago. That is a change in what the work is, and it shows up in what teams hire for. It is a separate question from whether the roles exist, and evidence about one does not settle the other.

Whether AI is already costing young workers jobs is a separate, open question. A Stanford Digital Economy Lab update in August 2026 found employment for 22 to 25 year olds in the most AI-exposed occupations about 19% below where it would be had it kept pace with same-age workers in less-exposed ones. The authors describe that as a descriptive pattern and say they cannot yet show that AI is the cause.

Some guests see the tools helping new engineers once they are hired. In episode 10 at 7:31, Rizel Scarlett, then a developer advocate at Block, says that “if I had this when I was a junior developer, this would have made me so much more confident.” Her reasoning at 7:52 is that junior engineers often hesitate to ask colleagues questions, and an AI assistant lets them ask why code works the way it does. That is her view of the work itself. It does not address whether entry-level roles are being posted.

Other guests expect the value to move away from writing code. Asked about the second-order effects for builders and creators, Intangible founder Bharat Vasan said in episode 30 at 7:45, published in July 2025, that he believes the “marginal cost of writing code is gonna go to zero,” and the cost of creating content too, so what will matter is taste and distribution: whether you can make something for a particular audience and reach it. That is a forecast about builders in general, not a measure of engineering hiring. The episode aired in season two, which Galileo sponsored while Conor worked there; Conor points listeners to Galileo’s YouTube channel at 0:06.

Check each figure’s denominator

Greenhouse’s March 2026 benchmark report, covering over 6,000 companies and 640 million applications from 2022 to 2025, measures related but different things: applications per job up 111% (116 to 244), annual applications per recruiter up 203.7% (1,610 to 4,890) and monthly hires per recruiter up 122% (2.2 to 4.9). Hires per recruiter measures recruiter throughput, not total hires, and it can rise while the pipeline is overloaded. On average the load per recruiter roughly tripled; Chait’s four or five times describes what an individual recruiter may face.

The survey figures have their own populations. In Greenhouse’s November 2025 survey, the 41% who “admit to using prompt injections” are from 1,200 U.S. job seekers reporting on themselves, and the 91% are recruiters who “have spotted candidate deception,” which is a share of recruiters, not of applications. Neither figure is specific to engineers, and neither can be combined with platform activity data.

Why it matters

Funnel figures cannot settle how much of the problem is weaker demand. To the extent it is signal, the fix is anything that puts information back in the channel: a targeted application instead of a hundred automated ones, and a screening process that rewards evidence over keyword overlap. Both sides are currently optimizing against a system that stopped carrying information.

The AI hiring doom loop: a signal problem A loop of four steps. Candidates use AI to apply to more jobs and tailor every resume to the posting. A recruiter may now handle four or five times the applications they did a couple of years ago. They point an off-the-shelf model at the inbox to pick the 20 of a thousand to call. That makes it harder to get seen, so candidates send more. In the middle, the signal drains away: resumes converge, a needle made to look more like hay. To the right, the way out is to put information back: a targeted application instead of a hundred automated ones, and screening that rewards evidence over keyword overlap. Below: the job itself did change, which is a separate question from whether the roles exist; by Greenhouse's data, software engineers use automation the most and send the most applications. The hiring doom loop Each side responds reasonably to the other. Together they drain the signal out of the channel. Candidates auto-apply,tailoring every resume 4–5× applications a recruitermay handle, vs. a coupleof years ago A model picks 20 of 1,000to call Harder to get seen,so candidatessend more signal drains away BREAK IT: PUT SIGNAL BACK “You’re making your needlelook more like hay.” Daniel Chait A targeted application instead of a hundred automated ones. Screening that rewards evidence over keyword overlap. The job did change. Whether the roles exist is a separate question. Daniel Chait, Greenhouse CEO, ep 58: by Greenhouse’s data, software engineers use automation the most and send the most applications. He calls it a signal problem, a matching problem, and doesn’t cast engineers as the villains.
Both sides respond reasonably and together drain the signal. The recruiter load, the needle-and-hay line and the Greenhouse data on engineers are Daniel Chait’s, from episode 58. Download the image

Hear it from the guest

“Ultimately, what it becomes is kind of a signal problem. It's like a matching problem”
Watch on YouTube
“It's like you're sending in … a needle to an ever growing haystack, and you're making your needle look more like hay.”
Watch on YouTube
“Software engineering is not about writing code. It is about solving business problems with technology, which has historically included a lot of writing code.”
“The job of being an engineer these days is so much more about managing people and delegating tasks. It's just that those people are not people, they're agents”
Watch on YouTube
“If I had this when I was a junior developer, this would have made me so much more confident.”
“So I do believe marginal cost of writing code is gonna go to zero, and the marginal cost of creating content is gonna go to zero as well.”

Quotes lightly edited to remove filler words.

From the conversation

This explainer is drawn from these episodes — each carries its full transcript.

Concepts in this explainer

AI Code GenerationAI EngineerProduct engineerAI AssistantPrompt InjectionAI Benchmark