Why 'It Depends' Is Software's Most Valuable Answer
Architectural judgment becomes the premium skill in the AI era.
1Architects' go-to answer
Software architects famously respond 'it depends' to most questions. This phrase marks the boundary of AI capabilities. While AI excels at answering 'how' questions, 'should' questions require human judgment. Architectural decisions depend on company-specific factors like team morale or budget constraints, making standardized answers ineffective in real-world scenarios.
2The first law stands firm
The First Law of Software Architecture states: Everything is a trade-off. AI doesn't invalidate this principle - it reinforces it. AI coding tools can answer 'how' questions by drawing from documentation and open-source code. But strategic decisions depend on internal factors like cloud spending freezes or organizational changes that AI can't access or weigh appropriately.
3Second law: Why beats how
AI are 'how' machines; architects are 'why' people. While AI can list generic pros and cons of distributed architectures, it can't determine their relevance for your specific context. Factors like your ops team size or payment team's morale influence decisions. This makes the ability to explain and justify architectural choices more valuable than ever.
4Cheap code, expensive decisions
Does easier software development make architecture less important? Paradoxically, faster implementation makes bad decisions more costly. Previously, questionable designs took months to build, allowing time for course correction. Now teams can deploy services in weeks. AI acts as an amplifier - it accelerates both good structures and cements flaws before anyone can intervene.
5Half-life of bad decisions
Previously, the impact of poor architectural choices was limited by implementation time. With instant implementation, you may live with consequences for years. As code becomes abundant, judgment becomes the scarce resource. Determining service boundaries, defining 'good enough' availability, and prioritizing architectural characteristics all require human discernment AI can't provide.
6How judgment develops
Trade-off analysis can't be learned through content consumption alone. Books provide vocabulary, not judgment. This skill develops through making decisions and experiencing consequences, or observing experienced architects debate alternatives. The pattern catalog is easy; knowing which pattern not to use (and why) takes years of practice - exactly what AI lacks.
7What AI can't learn
AI models see millions of decisions but almost none of their long-term consequences. Post-mortems tracing outages to architectural decisions rarely appear in training data. Even when available, they're disconnected from the original pull requests. Architectural feedback loops span years - precisely the kind of learning that can't be scraped from existing data.
8Engineers' strategic move
As AI absorbs implementation work, engineers must move up the value chain toward decision-making. The territory of trade-offs, constraints, and 'it depends' is expanding, not shrinking. This skill develops through practicing architectural thinking on real problems with expert feedback. The defining career skill of the next decade will be answering 'should' questions that begin with 'It depends.'
Original article: O'Reilly Radar — Yazılımda 'Duruma Göre Değişir' Cevabının Önemi