Boilerplates have long been the go-to for developers starting new projects. They provide a ready-made structure, enforce standards, and save us from writing the same setup code again and again.
But with AI-powered tools like GitHub Copilot, ChatGPT, and other code generators, the question arises: do we still need boilerplates in 2025?
AI can scaffold a project in minutes, generate configs, and even suggest best practices on the fly. It feels like the gap boilerplates once filled is now being covered by smarter, context-aware assistants.
On the other hand, boilerplates still offer:
- Consistency: Everyone in the team starts with the same foundation.
- Standards baked in: Security, logging, CI/CD, and architecture patterns aren’t left to chance.
- Battle-tested reliability: AI may hallucinate, but boilerplates capture lessons from real-world deployments.
So maybe it’s not “AI vs boilerplates” but rather AI + boilerplates—where AI personalizes and extends the foundations boilerplates provide.
What do you think?
- Will boilerplates fade away as AI gets better?
- Or do they still have an irreplaceable role in ensuring standardization and reliability?
I’d love to hear how you and your team approach this—especially if you’ve tried blending AI with boilerplate setups.
Top comments (3)
Everything is getting worse right now in my opinion. Documentation is lazy, development is lazy, forums, wikis, and knowledge sites are already full of outdated code and AI keeps suggesting more outdated code adding to bad examples every day.
Better boilerplates being kep up to date could provide a better alternative to the AI-generated vibe coding roulette.
Outdated docs and AI-suggested snippets just make things messier. An up-to-date boilerplate can save so much hassle by baking in best practices from the start. That’s why Ive been keeping it fresh. it’s been a real game-changer compared to the usual “AI roulette.”
I think the way Drupal is has introduced bootstrapping by recipes, it looks like a way that can live together with AI.
When there is a method behind the composability of the recipes, I think it will be possible to ask an AI to get all the recipes based on the requirements and show how they need to be executed. The next step is automating the fetching of the recipes and executing them.
I think boilerplate code with a strong opinion is on the way out.