You need to become a full stack person (den.dev)

The author argues that recent LLM advances are already commoditizing many narrow technical tasks—turning Figma mocks into working web prototypes in hours, auto-generating code, PRs and even basic A/B analyses using tools like Claude Code, GitHub Copilot agents and internal helpers—so role boundaries are blurring and a “product engineer” archetype is re-emerging. This isn’t doom-saying: LLMs accelerate implementation and lower friction, but they also precipitate a major role expectation shift. Rather than replacing skilled people, AI changes which combinations of skills create durable career moats. Technically, these systems are powerful at CRUD, pattern recognition and scaffolding, but they still fail at first principles, security, system design and context-dependent judgment (e.g., unsafe auth patterns or fragile architecture). The practical implication for AI/ML practitioners is to become “full-stack” in a broader sense: pair deep domain knowledge with product sense, creativity/taste, systems thinking, critical thinking, execution speed and AI-augmentation fluency. Use AI tools to scale yourself, but retain ownership of architecture, security, UX and trade-offs. The actionable takeaway: lean into cross-domain combos that LLMs can’t replicate—judgment, creativity and technical fundamentals—while adopting AI tooling to ship faster and iterate smarter.