The staffing agency is 5,000 years old. The AI revolution didn't kill it. It just rebranded it.
Anthropic raised $1.5B to send humans to fix what their AI broke. OpenAI countered with $14B. Welcome to Humans as a Service.
A YouTube creator named Mo Bitar posted a video this week that is equal parts comedy and confession. It's called "Anthropic just admitted AI is bullsh*t," and I'd strongly recommend watching it before reading this, because the way he tells it is funnier than anything I'm going to write here.
But the thing underneath the jokes is real, and it's worth sitting with.
Here's what actually happened.
Anthropic announced a $1.5 billion joint venture with Blackstone and Goldman Sachs. The product: engineers. Specifically, "Forward Deployed Engineers," a title borrowed from Palantir, a company famous for sending highly trained humans into client organizations to make enterprise software actually work.
OpenAI saw the announcement and responded with Deployco, their own human deployment venture, valued at $14 billion.
Two of the most well-funded AI companies in history, both racing to build the best staffing agency they can.
Mo Bitar's description of what these engineers actually do is worth quoting directly. They helicopter in, all wearing black short-sleeved t-shirts. They build you an impressive demo in the first week. The suits are floored. And then, about six months later when the thing starts falling apart, the engineers are already gone. "Backward deployed," as he puts it.
You want them back? Sure. Anthropic has more.
Now, the title that Bitar suggests you immediately put on your LinkedIn profile is "Forward Deployed Engineer," and his logic for why is genuinely sharp: the AI that screens your resume is probably using the Anthropic job posting as its matching template. So if your resume contains the same language as that posting, you go to the top of the pile. You don't even have to know what the job is yet. Nobody does. It's new enough that the ambiguity works in your favor.
This is funny and also slightly terrifying as career advice, which makes it the best kind.
But the part of the video that actually landed for me is the argument underneath all the jokes.
These companies were built on a specific promise. Plug in the AI, reduce your dependency on human labor, let the autonomous systems handle it. That was the pitch. That was the $700 billion valuation. That was the reason every enterprise CTO spent 2024 and 2025 canceling engineering contracts and telling their boards the AI would pick up the slack.
What happened instead is that the AI made the remaining engineers worse at their jobs.
This isn't just Bitar's read on the situation. Anthropic published the research themselves. A randomized controlled trial with software developers found that participants who used AI assistance scored 17% lower on comprehension tests than those who coded by hand. Nearly two letter grades. The biggest gap showed up in debugging, which is precisely the skill you need to catch errors in AI-generated code. And buried in the footnotes: the effects are probably more pronounced for tools like Claude Code than what the study even measured.
The desire to actually code, to sit with a hard problem and work through it, is eroding. Bitar calls it plainly: "AI came and it made your engineers stupid. They stopped having the desire to code anymore. That desire is gone."
So now you have a workforce that has been quietly deskilled, a codebase full of AI-generated slop that nobody fully understands, and a product that demos beautifully but cracks under real load. And Anthropic's answer to this problem, which their own product helped create, is to sell you a team of engineers who actually know what they're doing.
Bitar calls this "the greatest con in the history of capitalism." He's being hyperbolic, but only slightly.
The endgame he sketches out is darker, and worth taking seriously even if you think his timeline is off.
If the best software gets built by people who have access to frontier models, and frontier models are increasingly gated, then whoever controls access to those models controls who can build good software. Anthropic is already doing this with Mythos. The model that can find thousands of zero-day vulnerabilities, autonomously, across every major operating system, is not something you can access. It's something Anthropic deploys through Project Glasswing, through vetted partners, through controlled pipelines.
Give the public the model that writes poetry and helps with email. Hold back the model that builds serious software. Then sell access to that capability through your own humans, on your terms, at your prices.
In that world, Bitar says, "the only people doing software in ten years are OpenAI and Anthropic." The forward deployed engineers aren't a product line. They're the preview of a monopoly.
Whether you buy the dystopian version or not, the near-term picture is already strange enough.
The companies that sold enterprises on reducing human labor dependency are now billing enterprises for human labor, at a premium, with a brand new job title that makes it sound like a military operation. The engineers who got laid off to make room for AI are being replaced by AI company engineers who cost more and come with worse job security. And the organizations caught in the middle are paying for both: the AI subscriptions that were supposed to solve everything, and the humans who show up to fix what the AI broke.
Mo Bitar's practical advice, delivered completely straight-faced, is to study agentic paradigms the way a previous generation crammed algorithms they'd never use, just to pass the interview. Learn the theory. Understand the vocabulary. Then open Claude in the technical interview and let it do the work while you narrate.
"That's called AI engineering," he says. "They're going to see that."
He's joking. Mostly.
One thing to think about
The staffing agency is 5,000 years old. It survived the industrial revolution, the computer revolution, the internet. It turns out the AI revolution didn't kill it either.
It just rebranded it.
Watch Mo Bitar's original video here: https://www.youtube.com/watch?v=juHv_Vi4giU