The argument is simple but uncomfortable. As more engineers lean on LLMs to write their code, the ones who still practice by hand will become rare and valuable.
A SiteBloom opinion piece lays out what's pushing engineers toward AI dependency. Social pressure from all those "fever tweets" about hyperspeed shipping is real. The models also genuinely work and keep getting better. Oh, and we're lazy. More code gets shipped while less gets understood by the humans shipping it.
The author paints two scenarios for engineers who go all-in on agentic workflows. Employees risk becoming "prompt monkeys" who compete with product designers and business folks in an oversaturated market.
That's not where you want to end up.
Startup builders face a different problem. Failed projects don't prove you can actually build anything anymore. They just prove you can prompt an idea into existence and try to market it. Venture capitalists love this since they pick up the winners. But it leaves individuals with little to show for their failures.
Major tech companies are already adjusting hiring practices in ways that reward manual skill. Interview platforms like Karat and HackerRank report a shift toward live, interactive coding sessions where candidates can't hide behind AI generation. Some companies now evaluate AI literacy as its own competency. They're asking candidates how they'd validate machine-generated code rather than simply banning the tools. Harvard physicist Avi Loeb has noted students submitting papers with AI-hallucinated references. It's a sign that cognitive atrophy extends well beyond software engineering.
The SiteBloom piece predicts this window lasts at least a decade, until superintelligence changes the game entirely. That's a long time to bet against fundamentals.