@csnyder i also like the next sentence of his speech, coming after this often quoted part:

„The first method is far more difficult.“

with this sentence, he makes clear that there is no shortcut to a simple solution for a given task. the only shortcuts that exist either lead to complex solutions or to wrong solutions.

but, indeed, we cannot repeat sir tony hoares timeless insights often enough. thanks for sharing!

delivering features nobody wants is the biggest problem we currently have in software development. delivering more features without clearly measuring their value will just aggravate the problem.

very important: users/customers define value, not business departments. this misconception is where the problems start.

2/2

@rubenrubiob i think the bubble and the technology are mostly unrelated.

this is what makes is so hard. while the technology per se is interesting and can be very useful, the bubble overlays it with lots of detrimental effects.

however, those effects are due to the wants and demands of the bubble. the technology is just a tool that is abused to feed those wants and demands.

therefore, i think it is hard to predict what happens after the bubble bursts - so much bubble-driven nonsense around :/

@rubenrubiob i think, it is an important issue you raise and all that stuff worries me too. it was just not the focus of this post (it is impossible to put all considerations regarding a topic like AI into a single post and still get something halfway understandable – at least, i can't do that).

personally, i try to understand how far i can get with open weight models as this gives me access to the technology without the issues you mentioned. these models are a bit weaker but often good enough.