During the last year, I have been developing PoCs using AI tools such as Kiro, Codex... and other ADEs.
Like many devs, I started by focusing on the questions.
How to write better prompts, how to get better code, how to make AI generate more code?...
But after months of working this way, I realized something... the biggest challenge was not generating code, the biggest challenge was preserving intent.
As projects grow, architectures evolve, requirements change, and teams change, decisions made months ago begin to fade away!!!
The original vision becomes harder to perceive. Because AI does not know why a specific architecture was chosen, nor does it know why a particular cloud strategy was selected. AI only knows the context available during the current interaction.
Then we see this:
"AI does not follow the architecture, but the available context"
This creates a new challenge for software engineering today. The more capable AI becomes, the more important human intent or human solution vision becomes. Well... this is not less important, it is very important.
Because AI can generate solutions and accelerate delivery. But AI does not own the outcome, we do... architecture remains a human responsibility.
The software architect of the future will not compete with AI. The software architect of the future will govern it.
That is why I increasingly believe that we need to rethink AI-assisted development; prioritize human intent through governed context. etc. The real work is no longer just about writing code.
NOTE!!
The real work is about preserving the knowledge, decisions, architecture, and intent that guide the creation of that code
I am currently exploring these ideas through a methodology that I am calling Context-Driven AI Development (CDAD), focused on treating context as a governed architectural asset, rather than a collection of isolated prompts.
I am interested in knowing how others approach this challenge and whether they see it as a challenge as well...
Thanks for reading!
Griott
Top comments (0)