Got Tech? Will Hack.
Software engineering is changing — again. Not in a loud, overnight way, but in a quiet structural way that’s already reshaping how good teams build software.
We aren’t replacing engineers. We are upgrading the way engineers think, work, and build.
AI isn’t a shortcut to avoid the hard parts — it simply shifts where the hard parts are.
Over 2 years applying AI in prototyping, experimentation, internal tools, production systems, and team workflows — one thing has become very clear:
AI doesn’t make engineering easier. It makes disciplined engineering more valuable.
Great teams are not the ones who “use AI everywhere.” They’re the ones who use AI well — with clarity, responsibility, and intent.
Below is a working set of principles we’ve found useful for building in this new environment. They aren’t commandments, they aren’t finished, and they will evolve — just like our tools will.
But they help keep us grounded in the parts of engineering that matter.
Over the past year of building AI-augmented engineering systems and working closely with teams, a set of values has consistently emerged. They aren’t commandments — they’re reflections from practice. Much like the Agile Manifesto shaped a generation of builders, this is my early articulation of what effective, responsible, AI-native engineering feels like in practice.
AI is changing knowledge work fundamentally. We have learnt through our experience that it is critical to use this powerful capability responsibly to achieve successful outcomes.
Through our work, we have come to value:
That is, while there may be value in the items on the right, we value the items on the left more.
These principles fall into two buckets — what is new, and what remains timeless but more important than ever.
Principles AI use creates or transforms — they wouldn’t exist without it.
AI can extend your reach, accelerate your ideas, and surface possibilities you may not see, but it cannot own the outcome. Engineering judgment, ethical responsibility, and decision-making stay with us. Tools assist; humans remain answerable.
You can use AI to help build systems but we are still accountable for the correctness of the outcome. You check in the code under your name and you are still responsible.
AI outputs only reflect the clarity, completeness, and structure of the input. If we want meaningful results, we must bring meaningful context — not vague requests. Better thinking in produces better thinking out.
Learn how to manage context well. The larger the system, the more important it becomes to build discipline and practices around how context should be managed. Good engineering practices can help ensure new teammates get AI systems primed with up to date and correct context for every project. These practices also help ensure the system stays up to date. If the context is too large for your model to hold, teams should engineer solutions around it like optionally loaded markdowns or, at larger scales, RAG.
As generation gets faster, review must become sharper. Code, ideas, and architectures produced by AI still demand rigorous validation for quality, safety, and alignment with intent. The faster we move, the stronger our checks must be.
Don’t let generic tooling decide how your team works. Choose where AI fits, what it should influence, and how it should be used to support — not reshape — your engineering culture. Intentional adoption prevents accidental dependencies being created.
AI practices evolved weekly; now they evolve monthly. This is still faster pace than many, if not most are used to. Teams that keep experimenting, reflecting, and adapting stay ahead. Treat AI as a moving system — one that rewards curiosity, continuous improvement, and lightweight experimentation. What didn’t work a few months ago might be possible now and the only way you will know is if you experiment.
Good development sense that now matters even more with AI in the loop.
Start small, validate often, and tighten feedback loops to ensure AI continues to deliver real value.
Unmaintainable, insecure, and rigid solutions waste time and money. Always prioritise building the right value over building the wrong one fast.
This isn’t theory. Here’s what it means day-to-day on an engineering team:
This is not “old engineering vs new engineering.” It’s the next chapter of the same story: build well, stay curious, stay accountable.
AI doesn’t remove the craft of engineering. It multiplies the importance of the engineer.
Principles shape how we think. Skills shape what we can do with that thinking.
To build effectively with AI, engineers need to understand not just how to prompt, but how these systems work underneath.
Mastering these skills turns AI from a black box into a design partner. That’s the real craft of intelligent engineering.
Prompt Engineering and Context Engineering are the new craftsmanship of AI-era software building.
It’s no longer about “writing the right prompt” — it’s about structuring intent, constraints, and information so that the model understands your problem the way you do.
To use AI tools responsibly and creatively, engineers should understand the mechanics: how tokens, embeddings, and vector spaces shape what the model “remembers,” “understands,” and “forgets.”
This isn’t about becoming an ML engineer — it’s about having the literacy to reason about your tools.
Modern AI systems go beyond single prompts.
Concepts like vector search, retrieval-augmented generation (RAG), and agents define how context flows and how reasoning chains form.
Engineers should learn to design with prompt libraries, multi-agent orchestration, and feedback loops that adapt over time.
Teams that adopt AI without principles create:
And then they pay for it later — painfully.
Teams that adopt AI with principles:
The future is not “AI builds everything.”
The future is AI-raised engineers who build better than before.
Agile reshaped how we deliver. AI is reshaping how we think while we deliver.
Who are we?
Not Prompt Writers.
Not Tool Operators.
intelligent Engineers
We’re just at the beginning. These principles will evolve. If you’d like to build this thinking together — I’d love to hear your take. What principle would you add or challenge?