Will AI wipe out the first rung of the career ladder?

1 day ago 6

Hello, and welcome to TechScape. This week, I’m wondering what my first jobs in journalism would have been like had generative AI been around. In other news: Elon Musk leaves a trail of chaos, and influencers are selling the text they fed to AI to make art.

AI threatens the job you had after college

Generative artificial intelligence may obviate the job you got with your diploma still in hand, say executives who offered grim assessments of the entry-level job market last week in multiple forums.

Dario Amodei, CEO of Anthropic, which makes the multifunctional AI model Claude, told Axios last week that he believes that AI could eliminate half of all entry-level white-collar jobs and send overall unemployment rocketing to 20% within the next five years. One explanation why an AI company CEO might make such a dire prediction is to hype the capabilities of his product. It’s so powerful that it could eliminate an entire rung of the corporate ladder, he might say, ergo you should buy it, the slogan might go.

If your purchasing and hiring habits follow his line of thinking, then you buy Amodei’s product to stay ahead of the curve of the job-cutting scythe. It is telling that Amodei made these remarks the same week that his company unveiled a new version of Claude in which the CEO claimed that the bot could code unassisted for several hours. OpenAI CEO Sam Altman has followed a similar playbook.

However, others less directly involved in the creation of AI are echoing Amodei’s warning. Steve Bannon, former Trump administration official and current influential Maga podcaster, agreed with Amodei and said that automated jobs would be a major issue in the 2028 US presidential election. The Washington Post reported in March that more than a quarter of all computer programming jobs in the US vanished in the past two years, citing the inflection point of the downturn as the release of ChatGPT in late 2022.

Days before Amodei’s remarks were published, an executive at LinkedIn offered similarly grim prognostications based on the social network’s data in a New York Times essay headlined “I see the bottom rung of the career ladder breaking”.

“There are growing signs that artificial intelligence poses a real threat to a substantial number of the jobs that normally serve as the first step for each new generation of young workers,” wrote Aneesh Raman, chief economic opportunity officer at LinkedIn.

The US Federal Reserve published observations on the job market for recent college graduates in the first quarter of 2025 that do not inspire hope. The agency’s report reads: “The labor market for recent college graduates deteriorated noticeably in the first quarter of 2025. The unemployment rate jumped to 5.8% – the highest reading since 2021 – and the underemployment rate rose sharply to 41.2%.” The Fed did not attribute the deterioration to a specific cause.

The likeliest outcome of AI’s impact on entry-level jobs is that companies will reformulate them into something new. The job market may settle somewhere between Amodei’s AI Ragnarok and the antediluvian days before ChatGPT. Familiarity with AI will be required in the way that Microsoft Office has, and bosses will demand a higher standard of productivity. If a robot can do most of the coding for you, a junior software engineer, then you should be producing five times the amount of code as before, they may say.

Speaking of Microsoft and software engineers, CEO Satya Nadella claimed in late April that AI writes 30% of Microsoft’s code. That may be the future of software development. It is possible that is true; it is also possible that Nadella, head of the company that has reaped enormous gains from the AI boom, is trying to sell by example, overestimating how much of that code is usable. Meta’s Mark Zuckerberg has been more pointed in his assessments, asserting that his company will no longer need mid-level coders by the end of 2025. Shortly after, Meta announced a 5% staff reduction.

The short-term readjustment, however, is the pain point. Recent classes have graduated without AI being an integral part of their school life, and employers won’t believe those interim graduates have the necessary familiarity for a new professional landscape.

That is not the fault of the graduates: Employers themselves don’t know what they want yet from AI. Axios followed up Amodei’s doom and gloom with a piece about how AI job cuts are jumping the gun. Companies are not replacing departing workers, betting that AI will be able to perform the same functions, if not now, then hopefully by the time it would take to hire replacements.

The example of journalism may be a canary in the coalmine. Entry-level jobs in journalism often involve aggregating news items from other outlets in the style of your own employer, a task AI is well suited to if the facts are straight. I spent several years doing just that when I started out. In the same way that we see Amodei’s predictions taking shape in LinkedIn’s data, I see the entry-level diminishment beginning in my own industry. Business Insider, a digital outlet focused on financial and business news, laid off 20% of its staff late last week. CEO Barbara Peng said the newsroom would go “all-in on AI” and become “AI-first” in her note eliminating the jobs.

Axios itself footnoted its Amodei interview with a disclosure about its own practices with regard to AI.

“At Axios, we ask our managers to explain why AI won’t be doing a specific job before green-lighting its approval. (Axios stories are always written and edited by humans),” the disclosure reads. The parenthetical indicates that Axios editors know that AI’s involvement in writing is bad for the brand. The part said outside the parentheses indicates Axios executives may not be backfilling vacated jobs, waiting for AI to catch up and close those openings.

The week in AI

Musk leaves behind a sticky trail like a slug

Elon Musk announced he would leave the White House last week, ending a contentious and generally unpopular run as a senior adviser to the president and de facto head of the so-called “department of government efficiency” (Doge). Donald Trump hosted a press conference for his departure, the same day that the New York Times reported that Musk had heavily used drugs on the campaign trail.

My colleague Nick Robins-Early assessed the mess Musk leaves in his wake:

As Musk moves on, he consigns a mess of half-realized plans and gutted agencies to his acolytes installed in key positions across the federal government. His departure throws Doge’s already chaotic impact on the government into an even grayer limbo, with questions over how much power the nebulous taskforce will have without him and who, if anyone, might rebuild the programs and services it destroyed.

Musk’s initial pitch for Doge was to save $2tn from the budget by rooting out rampant waste and fraud, as well as to conduct an overhaul of government software that would modernize how federal agencies operate. Doge so far has claimed to cut about $140bn from the budget – although its “wall of receipts” is notorious for containing errors that overestimate its savings. Donald Trump’s new tax bill, though not part of Doge and opposed by Musk, is also expected to add $2.3tn to the deficit, nullifying any savings Doge may have achieved. Its promises of a new, modernized software have frequently been limited to artificial intelligence (AI) chatbots – some of which were already in the works under the Biden administration.

The greater impact of Doge has instead been its dismantling of government services and humanitarian aid. Doge’s cuts have targeted a swath of agencies such as the National Oceanic and Atmospheric Organization, which handles weather and natural disaster forecasting, and plunged others such as the Department of Veterans Affairs into crises. Numerous smaller agencies, such as one that coordinates policy on homelessness, have been in effect shut down. Doge has brought several bureaus to their knees, with no clear plan of whether the staff Musk leaves behind will try to update or maintain their services or simply shut them off.

As Musk returns to Tesla and SpaceX, the agencies he laid waste to are left to pick up the pieces.

skip past newsletter promotion

While Musk is returning to his tech empire, many of the former employees and inexperienced young engineers whom he hired to work for Doge are set to remain part of the government. One of the largest questions about what Doge’s future looks like is whether these staffers, some of whom gained near unfettered access to the government’s most sensitive data, will retain the same powers they enjoyed under Musk.

Read a timeline of Musk’s stint in Washington.

A mess of misinformation

Influencers are selling the text that prompted AI to generate their art

A ChatGPT logo on a keyboard. Photograph: Jaque Silva/NurPhoto/Rex/Shutterstock

Would you buy instructions for ChatGPT?

Two weeks ago, the Instagram account @voidstomper, which posts grotesque videos generated by AI to 2.2m followers, offered a novel kind of sale. Up for grabs were 10 prompts used to prod AI-powered engines into generating the videos that the account itself had posted.

Voidstomper posted a video captioned: “I didn’t want to sell these. But I’m broke and these still go viral. So here: 10 raw horror prompts straight from my archive. Some got me hundreds of millions of views. Some barely make sense. All of them work. Use them in any AI video tool. Just don’t pretend you wrote them. VAULT_DUMP_1 is live. Link in bio. No refunds. You’re on your own.” The account’s administrator did not respond to a request for an interview.

The account is not alone. Marketplaces exist for the sale of AI prompts. Ben Stokes, the founder of the PromptBase, says there are some 20,000 sellers on his site hawking prompts, with thousands sold monthly and, to date, seven figures paid out to their writers since 2022. He said social media influencers and other content creators selling their prompts is “still quite niche” and serves as a side hustle for graphic designers, artists, and photographers.

Voidstomper put prompts for sale that had been used to create specific videos. The product a user might receive when purchasing a prompt on PromptBase is likely to be closer to a generalized template than a finite sentence, Stokes said.

“For example, if the prompt that creates posters of famous landmarks in a vintage style, there’d be sections within the prompt in square brackets like [LANDMARK NAME] that you could change to the landmark you’d like to create a poster for, like the local pier in your town,” he said.

Why buy a string of text when you could type out your own, though?

“There is a specific group of people who are looking for high-quality, robust prompts for their business applications. Specifically, they are looking to integrate AI into their product or workflow, which usually requires a prompt, and want to ensure it works well and produces great consistent outputs,” said Stokes. Though the general public thinks of ChatGPT as free, running enough generations to obtain specific and correct outputs can be quite expensive for businesses, he added. It may be more cost effective to buy a prompt.

Even within the niche of AI-generated art, some consider the sale of prompts ridiculous. The Instagram and TikToker HolyFool36, an AI art Instagram account that has been interviewed in this newsletter before, said he would never engage in the practice.

“Frankly, I find it as an insult to my sensibilities,” he said via email. “Generative AI requires no skill – almost anyone can figure out how to reverse engineer his prompts for free.”

“I personally know Void Stomper and have had many interactions with him. I’ve explained that the best way to monetize this stuff is to build a brand and then sell real tangible products within that brand. He has clearly, directly, stated that he does not have the discipline to do that. I don’t judge him, everyone’s gotta pay rent somehow, I’d just never go about it that way personally,” he added.