Found this article and it articulated something I've felt for a long time: that the software industry took a significant wrong turn when it replaced the culture of detailed specifications (like the German 'Pflichtheft') with the ambiguity of user stories and backlogs.
The author makes a compelling case that this shift, done in the name of "agility," actually introduced chaos and massive costs, especially in critical systems. They make it clear this wasn't a return to waterfall; the article describes how these specs were *living documents, developed iteratively over years.* Serious institutions like Deutsche Bundesbank (the German central bank) used this rigorous approach and built rock-solid, durable systems such as TARGET/SEPA as a result.
The most interesting point, though, is how GenAI completely changes the calculus. It transforms the detailed spec from a mere document for humans into the *ultimate prompt*: an executable, high-level source code for AI. The article links to a brilliant talk by Sean Grove on this exact concept. This makes the old, proven method of spec-driven development hyper-efficient and viable again.
It feels like a "back to the future" moment. Curious what HN thinks. Was the widespread move away from specs a mistake fueled by agile consultants, and is AI the key to bringing rigorous spec engineering back?
Found this article and it articulated something I've felt for a long time: that the software industry took a significant wrong turn when it replaced the culture of detailed specifications (like the German 'Pflichtheft') with the ambiguity of user stories and backlogs.
The author makes a compelling case that this shift, done in the name of "agility," actually introduced chaos and massive costs, especially in critical systems. They make it clear this wasn't a return to waterfall; the article describes how these specs were *living documents, developed iteratively over years.* Serious institutions like Deutsche Bundesbank (the German central bank) used this rigorous approach and built rock-solid, durable systems such as TARGET/SEPA as a result.
The most interesting point, though, is how GenAI completely changes the calculus. It transforms the detailed spec from a mere document for humans into the *ultimate prompt*: an executable, high-level source code for AI. The article links to a brilliant talk by Sean Grove on this exact concept. This makes the old, proven method of spec-driven development hyper-efficient and viable again.
It feels like a "back to the future" moment. Curious what HN thinks. Was the widespread move away from specs a mistake fueled by agile consultants, and is AI the key to bringing rigorous spec engineering back?