Skip to the end

The article says developers shaped by pre-2021 IT thinking tend to treat software as a tool that helps users reach outcomes. It argues that AI changes that premise because tasks that used to be hard, like recognizing images or understanding natural-language requests, have become easy enough to produce results directly.

Reading notes

#

• Developers from the pre-2021 era are described as having an IT mindset that frames problems through relational databases, form submissions, table schemas, and imperative logic.
• The article contrasts older computing limits with AI capabilities, noting that computer vision and similar tasks were long considered nearly impossible but are now trivial in some cases.
• It argues that the basic assumptions about how computers work have flipped.
• Software should be designed less as a means to an end and more as the end itself.
• Instead of giving users folders, tags, and naming conventions, software can understand and organize content for them.
• Instead of asking users to build state machines and write DSLs, software can execute natural-language rules.
• Instead of requiring template systems and string replacements, software can generate custom messages at scale.