■ HIGH RISK ■ Technology
Yes for a large share of the work, with an ironic twist: breadth was the full-stack developer's selling point, and breadth is exactly what a general-purpose model has in abundance.
“Jack of all trades, master of none — and AI is becoming a jack of all trades too.”
Our AI replacement risk score — how we score jobs
Full stack means owning a feature end to end: schema change, API endpoint, business logic, interface, state handling, deployment and the bug report that follows. In practice these developers live in startups and small product teams, moving between a database migration in the morning and a payment flow in the afternoon, plus a fair amount of infrastructure and third-party integration nobody else has time for. The value proposition has always been that one person can ship a whole feature without three handoffs.
That proposition is now available from tooling. An agent can take a feature description and touch every layer, generating the migration, the endpoint, the component and the tests in one pass, and integration platforms handle authentication, payments, email and analytics with minimal code. Application-generation tools have become genuinely usable for the first version of a product, which is precisely the moment when a startup used to hire its first full-stack developer. The paradox is real: an engineer whose advantage was covering many mediocre-depth areas competes directly with something that covers all of them instantly, and our score of 67 reflects that squeeze.
What survives is ownership rather than coverage. Deciding the sequence of work, choosing which third-party service to bet the product on, catching that the generated payment flow does not handle a partial refund, and being accountable for a live system with paying customers are all human roles. Small teams are also where product and engineering blur, so full-stack developers who can talk to customers, scope realistically and make trade-offs about scope and speed become the person directing the tools rather than competing with them. The developers most at risk are those whose portfolio is competent breadth with no depth anywhere and no product judgement to fall back on.
Automatability: our editorial assessment of current and near-term AI capability
Already biting, and hardest in the exact market that employs full-stack developers. Startup application generation is good enough today that founders build a first version before hiring, which pushes the first engineering hire later and more senior. Through 2030 expect small teams to consist of a few product-minded engineers supervising generation across the stack. Generic full-stack contract work faces the steepest decline; product ownership roles hold up.
As a way of working, yes; as your entire identity, less so. Owning a feature end to end remains valuable in small teams, and it pairs well with directing agents across layers. What no longer differentiates you is simply being able to touch both ends of the stack, because that is now the baseline capability of the tools everyone uses.
They are hiring later and more selectively. Founders can now generate a functional first version themselves, so the first engineering hire tends to arrive when the product needs reliability, scale or security rather than when it needs to exist. That hire is expected to be senior and to bring judgement about architecture and operations, not just implementation speed.
Usually yes. Depth in one area, whether distributed backend, accessibility and performance on the frontend, security, data or infrastructure, gives you something that generation handles poorly and keeps your breadth as an advantage rather than a description. Alternatively, specialise in product: the person deciding what to build is not competing with a code generator.
Fewer keystrokes, more decisions. Expect to spend your time specifying features precisely, reviewing generated changes across layers for security and correctness, integrating and operating services, and talking to users about what is actually wrong. Individual output rises considerably, and teams stay small, which concentrates value in the people making the calls.