XNews

Software development bottleneck shifts to problem definition

Software development bottleneck shifts to problem definition

Thorsten Ball argued that as AI code quality surpasses human review capability, traditional software engineering practices like code review, unit tests, and agile will become obsolete. The bottleneck shifts from writing code to defining the right problem. Token becomes the new compute paradigm.

Key insights

Code review and unit tests become obsolete when AI writes better code

Ball argued that humans cannot find issues in AI-generated code within reasonable time, so code review dies. Unit tests become 'training wheels' when AI produces zero-compile-error code. The concept of 'good code' (readable, formatted) is for human readers; AI doesn't care about column width. Open source's 'many eyeballs' still applies but with AI eyeballs.

Human value shifts from coding craftsman to architect

The craft of writing code will diminish, but the craft of building software (understanding business problems, leveraging existing solutions, timing releases) becomes more important. Most bugs will be 'you asked for the wrong thing' bugs rather than coding bugs. PM/design/engineering tripartite becomes obsolete.

Token is the new compute paradigm; terminal and UI will die

Ball predicted the terminal will become as archaic as Perl one-liners. UI will be generated on the fly for 'dumb machines' and will shrink as machines get smarter. Token is the new production resource: those without enough tokens will be at a disadvantage. He argued we are entering a 'post-binary era' where everything is rebuilt on top of tokens.

Source posts