Martin Fowler breaks down how AI is transforming software architecture and development, from refactoring and deterministic techniques to the timeless principles that still anchor great engineering.
I love seeing that we as industry are finally discussion DevAI reasonably, where balanced middle ground is getting a voice!
Nondeterminism is particularly interesting topic. Past best practices, like testing and refactoring, are becoming essential to stabilize AI. Gergely's point that "you need to care about the craft" cuts to the heart of it. But our industry unfortunately might need a critical failure or two to acknowledge this reality and overcome marketing hype.
Many of the experiences discussed are only too relatable. From greenfield vs legacy discussions, where AI is shrinking the timeline it takes to get to unmanageable legacy. Overcomplicated code without learning and building mental models about the codebase. Seeing AI as dodgy collaborator / scatterbrained senior that needs to be managed carefully.
I love seeing that we as industry are finally discussion DevAI reasonably, where balanced middle ground is getting a voice!
Nondeterminism is particularly interesting topic. Past best practices, like testing and refactoring, are becoming essential to stabilize AI. Gergely's point that "you need to care about the craft" cuts to the heart of it. But our industry unfortunately might need a critical failure or two to acknowledge this reality and overcome marketing hype.
Many of the experiences discussed are only too relatable. From greenfield vs legacy discussions, where AI is shrinking the timeline it takes to get to unmanageable legacy. Overcomplicated code without learning and building mental models about the codebase. Seeing AI as dodgy collaborator / scatterbrained senior that needs to be managed carefully.
I've shared my experiences, and painful lessons, in "DevAI: Beyond Hype and Denial", https://www.ivankusalic.com/realistic-DevAI/