I've worked on everything from banking software and complex backend systems to SaaS products and small independent projects. One lesson has stayed remarkably consistent: there is no universally correct stack, and there is no single development process that fits every environment.
I like to use the best tool for the job. That means understanding the problem, the people involved, the operating constraints and the cost of the solution over time. Familiarity is useful, but it shouldn't be confused with suitability.
Efficiency is a design constraint
Efficient engineering isn't simply writing less code or finishing a ticket sooner. It's avoiding complexity that creates more work later. It's recognising when a simple solution is enough, and when security, scale or reliability genuinely require more structure.
I care about architecture because good architectural decisions make a product easier to change. They let a team make progress without continually paying for yesterday's shortcuts.
Process should adapt to the environment
Governance and discipline matter. A system handling payments or sensitive data deserves careful controls. But the same amount of ceremony does not make sense for every product or every team.
I prefer processes that serve delivery and evolve with the work. Clear decisions, reliable testing, shared standards and feedback loops can be more valuable than layers of approval that nobody can explain.
New tools, old responsibilities
AI-directed development has changed how I build. It can speed up the implementation of an idea, but it doesn't remove the need to understand the architecture, review behaviour, test assumptions or exercise judgement. Those responsibilities remain, even when the code is produced differently.
For me, the point of engineering is not to defend a particular tool or process. It's to create software that solves a real problem, works well and remains sensible to maintain.