The AI job apocalypse is the most documented disaster that never happened.
David Oks dropped a reality check that's making waves across developer circles: despite years of breathless predictions about AI wiping out entire industries, the evidence is underwhelming. His post on davidoks.blog pulled 261 points and 435 comments on Hacker News, and for good reason – he's calling BS on the panic.
Let's examine the numbers. GPT-3 launched in 2020. GPT-4 hit the scene in 2021. That's six and three years respectively of supposed job-killing AI being loose in the wild. The obvious targets? Customer service, data entry, basic content creation – the kind of repetitive work that should be trivial for AI to replace.
Yet here we are, still waiting for the tsunami.
The Goldman Sachs Paradox
Goldman Sachs published a bombshell report in 2023 claiming AI would eliminate or degrade 300 million jobs globally, with the U.S. and Europe taking the biggest hits in structured, repetitive tasks. But here's the kicker: they also predicted a 7% GDP boost from new opportunities created by the same technology.
Classic consulting firm math – scary headlines paired with optimistic footnotes.
<> "Evidence shows weak correlation between AI-vulnerable occupations and actual losses, implying sustained demand for human oversight in evolving tech stacks."/>
The data backs up Oks' skepticism. Early 2022 saw some weakening in AI-exposed jobs, but researchers like Frank, Sabet, Simon, Bana, and Yu traced this to economic factors, not generative AI. MIT's David Autor blamed political turmoil and government cuts. LinkedIn's chief economist pointed to macro uncertainty.
Meanwhile, AI-exposed fields like accounting actually saw youth employment growth.
The Real Story: CEO Theater vs. Reality
Here's where it gets interesting. Harvard Business Review reported that CEOs from Ford, Amazon, Salesforce, and JP Morgan Chase are making cuts based on "AI potential" – not AI performance. They're laying off entry-level workers because they think AI might replace them eventually.
This is premium executive theater. Fire people now, blame AI later, look visionary to shareholders.
The disconnect is staggering:
- AIMultiple predicts half of entry-level white-collar jobs gone by 2031
- Actual correlation between AI vulnerability and job losses remains weak
- U.S. unemployment sits at a healthy 4.3%
- Customer service automation remains gradual, not revolutionary
What Developers Actually See
As someone who's watched three years of "GPT will replace programmers" predictions, Oks nails the developer reality. AI tools are becoming incremental helpers, not wholesale replacements. GitHub Copilot suggests code. ChatGPT helps debug. But the architecture decisions, code reviews, and system design? Still human work.
The hiring slowdown in entry-level programming isn't because AI got better – it's because companies bought into the hype and paused recruiting. Classic self-fulfilling prophecy.
The Gradual Revolution
Oks admits he's "surprised at how slow automation has been," and that honesty is refreshing. The AI revolution looks more like:
- Subtle productivity improvements
- Tool integration rather than job replacement
- Cost savings through efficiency, not headcount cuts
- New roles emerging as fast as old ones evolve
Bernie Sanders and Matt Walsh can keep predicting the "avalanche," but the mountain isn't moving that fast. The gap between AI capability and AI deployment remains enormous.
The most dangerous prediction isn't about AI replacing humans – it's about humans making decisions based on AI predictions that haven't materialized.
Companies cutting jobs "because AI" while AI underdelivers are solving tomorrow's theoretical problem by creating today's actual one. That's not strategic planning. That's just expensive anxiety.
