In a candid LinkedIn put up stirring debate throughout tech circles, a Machine Studying engineer has issued a stark warning: AI instruments usually are not simply reshaping software program growth — they may be wrecking it.
“AI is creating the worst era of builders in historical past,” the engineer declared, predicting that by 2026, the business will see the primary wave of AI-native engineers getting fired.
New breed of ‘engineers’
On the coronary heart of the critique is a rising dependence on instruments like ChatGPT, which, the engineer argues, has birthed a era of builders who can paste code — however can’t clarify it, debug it, or construct something end-to-end.
Many junior devs right now, the put up claimed, observe a worrying sample:
- Paste code from ChatGPT
- Don’t perceive the way it works
- Can’t repair it when it breaks
- Showcase damaged, unfinished tasks
“When their AI-generated code breaks in manufacturing (and it’ll), they will rapidly understand:
- They can not repair it
- ChatGPT cannot repair it
- Stack Overflow cannot save them
They’re functionally illiterate,” the put up learn.
Hiring crimson flags already seen
Drawing from current technical interviews, the engineer shared what number of candidates now lean solely on AI output — usually with out understanding a single line.
“‘Stroll me by way of this code.’
‘Nicely, ChatGPT stated…’
‘However WHY does it work?’
[Silence].”
New premium: the ‘pre-AI developer’
Trying forward, the put up predicts a pointy divide. By 2027, builders who constructed foundational abilities earlier than the AI increase will probably be in demand — likened to artisans in an age of automation.
“The uncommon people who can debug with out a chatbot will command a premium. We’re speedrunning from ‘everybody can code’ to ‘nobody is aware of how something works.'”
The priority stretches past people. With AI instruments liable to hallucinations, outages, and price limits, over-reliant groups danger grinding to a halt.
“When the AI fashions go down, or simply hallucinate fallacious, your total engineering crew turns into ineffective,” the engineer warned.
“Controversial? Perhaps. True? Let’s see in 24 months. What number of builders in your crew may debug with out AI? Be sincere. Mine went from 8/10 to three/10 in two years.”
The put up ends with a blunt problem for tech leaders: Are we coaching drawback solvers — or immediate engineers?
Builders weigh in
The put up struck a chord on-line, with many echoing the priority.
“I discover that copying 1-2 traces from Copilot or an LLM works effectively if we retain the data by understanding it,” one person shared. “However once we blindly copy massive blocks, particularly unfamiliar code, it turns into a lot more durable to debug and really study.”
One other added: “AI will not kill engineering, however over-reliance simply would possibly. You may’t construct skyscrapers on sand. The long run belongs to groups pairing AI with actual understanding.”
