Fine-tuning sounds like the answer until you factor in the cost, the data pipeline, and the six months before a bigger model makes yours obsolete. Most of the time, prompt engineering or better context gets you there. But sometimes it doesn't — and that's where things get interesting.

In this free night session, Aaron Gallant covers the real tradeoffs behind fine-tuning LLMs, from synthesizing training data with frontier models to running PEFT and QLoRA on constrained hardware. You'll learn when smaller, specialized models actually beat throwing money at a bigger one — and why data curation is the work nobody wants to talk about. Built for engineers who want to make the right call, not just the cool one.

Live and remote. Wednesday, June 3 at 5 PM CT. Register now.

Register Free

You have probably made at least one job search decision this year based on a headline. Maybe you stopped applying to software roles because "AI is replacing developers." Maybe you panic-pivoted toward something that felt safer. Maybe you read a layoff announcement that blamed AI and quietly crossed an entire industry off your list.

Here is the problem: the person who ran the Bureau of Labor Statistics just looked at the actual numbers, and they do not say what the headlines say. Erika McEntarfer, former BLS Commissioner and now a labor economist at Stanford, published her read this week, and her conclusion is blunt: AI's impact on the current labor market is likely small right now. Her exact words: "I find facts are a lovely reality check on vibes."

The supporting data is stark. The Census Bureau's own business survey shows only about 1 in 5 American companies are using AI in any business function at all. Not "replacing workers with AI." Using it. At all. You cannot have a jobs apocalypse driven by a technology that 80 percent of employers have not deployed.

That does not mean the market is fine. It is genuinely hard out there, especially if you are early career. But the thing making it hard is not what you think it is, and if you misdiagnose the enemy, you fight the wrong war. Today: what the data actually shows, and how to search accordingly.

Here is the most counterintuitive finding in the data: unemployment is not rising fastest in AI-exposed jobs. It is rising fastest among workers who are the least exposed to AI. Software developers, the poster children of "AI will take this job," have seen continued employment growth. The live debate among researchers is whether developer jobs are growing as fast as before AI, not whether they are vanishing.

A Federal Reserve study did find that programmer employment growth dropped roughly 50 percent after ChatGPT launched, from about 5 percent annual growth to near flat in IT services. That is a real slowdown. It is not a collapse. Flat employment in a field still means hundreds of thousands of open seats turning over every year.

The tactical takeaway: if you have been avoiding roles, industries, or job titles because the narrative says they are dying, go check the actual BLS occupational data before you write them off. Self-rejection is the only rejection you fully control, and right now a lot of candidates are doing the recruiters' filtering for them.

McEntarfer called this out directly: there is a certain amount of AI washing happening in how companies explain layoffs. Blaming AI sounds visionary to investors. "We over-hired and our margins are bad" does not.

Her test for whether a layoff is genuinely AI-driven: look at which roles are actually being cut. If a company thinned out a management layer, that is standard cost cutting wearing an AI costume. A real AI-driven reduction shows up in specific roles on specific teams whose work was actually automated, and the pattern looks nothing like a typical restructuring.

Why this matters for your search: candidates treat layoff headlines as signal about where not to apply. If the layoff was AI-washed cost cutting, the company may be hiring again in two quarters, and the "doomed" function may be perfectly healthy at every competitor. Read the role list, not the press release.

So if AI is not the main story, what is? A slow-hiring market with a lot of stacked pressures: rate hikes, tariffs, immigration policy shifts, and general economic uncertainty. Companies are not firing much, but they are not hiring much either, and when hiring slows, it always hits people trying to get on the ladder hardest. McEntarfer is explicit that the young-worker squeeze is "more of a story about the aggregate decline in hiring across all kinds of occupations than an AI story."

One more timing detail that should change how you read every "AI killed entry-level jobs" take: the decline in young-worker hiring in AI-exposed roles started in mid-2022. That is months before ChatGPT even launched publicly, and over a year before businesses could use it with their own data. Companies do not stop hiring a year before the technology exists.

Tactically, a frozen funnel rewards different behavior than a layoff wave. The May jobs report still showed 172,000 jobs added and unemployment steady at 4.3 percent. Hiring exists; it is just concentrated. Your job is to find the pockets where it is actually happening: companies with fresh funding, teams with public growth targets, and roles reposted within the last week. Ten applications into active funnels beat a hundred into frozen ones.

AI changed what senior engineering means. Gauntlet is built for engineers who want to lead that change — not watch it. The most effective way to become AI-first. No tuition. No tutorials. Just the work. Apply now.

Cohort is filling up now

_{Must be a US citizen to qualify.}

Here is the part of the article that should actually change your behavior this week. McEntarfer says hiring is one of the most disrupted things from AI right now. Not the jobs. The process of getting them. AI resume screening, AI-conducted interviews, and AI-written applications are now the norm, and the research is still catching up.

Buried in her comments is a genuinely useful detail: AI interviewers appear to prefer AI-generated responses from candidates, because those answers are more predictable and easier for the model to classify and label. Read that again. The machine round of your interview process is not scoring you on charisma. It is scoring you on classifiability.

That does not mean you should have ChatGPT answer for you. It means you should structure your answers the way a model can parse them: direct answer first, one concrete example, one measurable result, clean stop. Save the personality, the tangents, and the texture for the human rounds, where they actually score points.

Before you make a fear-based move, spend 15 minutes doing what McEntarfer does professionally. Here is the checklist, copy and paste it into your notes:

The candidates losing right now are not losing to AI. They are losing to bad information, and that one is fixable today.

You don't have to scroll every AI thread, track every new tool, or watch every demo. 

The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning. 

Trusted by 2M+ professionals at Apple, Google, and NASA.

Sign up to start learning.