New data suggests that software engineering jobs -- long predicted to be among the first casualties of AI -- are in fact some of the most resilient roles to AI-driven disruption, according to TechCrunch. The finding cuts against a narrative that has hardened into conventional wisdom as agentic coding tools have proliferated.
The prevailing story has been stark: tools like GitHub Copilot, Cursor and Cognition's Devin can write, modify and debug code, so the number of human engineers needed should collapse. That assumption has driven both corporate hiring caution and eye-watering valuations for code-automation startups -- crystallized this very week by SpaceX's $60 billion acquisition of Cursor, a deal premised partly on enterprises 'significantly paring back their reliance on human engineers.'
The new data complicates that tidy thesis. One plausible reading is the productivity paradox: when a task gets dramatically cheaper, demand for it often rises rather than falls, because more software becomes economically worth building. AI may be making each engineer more productive while simultaneously expanding the universe of projects worth pursuing, leaving aggregate demand for engineering judgment intact or even growing. What changes is the nature of the work -- more architecture, review, and orchestration of AI tools, less rote implementation.
“What changes is the nature of the work -- more architecture, review, and orchestration of AI tools, less rote implementation.”
The context is a labor market where AI's impact has been deeply uneven, hitting some white-collar functions hard while leaving others surprisingly stable. Engineering's resilience, if it holds, suggests that roles combining judgment, system design and accountability are harder to automate than roles that are primarily about producing standardized output.
For founders, the implication is practical: AI tooling should be treated as leverage that lets a given team build more, not primarily as a way to shrink headcount -- and the engineers who master AI tools become more valuable, not less. For investors in code-automation companies, the data is a yellow flag worth weighing against the dominant 'AI replaces developers' pitch underpinning valuations.
The bear case on the bullish read: it is early, the data is noisy, and a genuine step-change in agent capability could still compress demand later. What to watch: whether engineering employment holds up through the next wave of agentic tools, how enterprise hiring patterns actually shift, and whether the 'AI replaces engineers' valuations survive contact with the data.