Can agentic coding raise the quality bar?
9 points by Jrmurr
9 points by Jrmurr
Could but won't. It's like how performance clusters around human response times regardless of how fast CPUs get, similarly the quality bar will always be "enough for sales in this product category."
I tend to agree with your premise, but if the cost of doing better quality falls enough, someone may try to use it as a competitive differentiator. Back in the day windows was always falling over and required user administration. If you could afford a Mac, you had the option of avoiding all that.
I suspect you are right about the quality of typical software. But I believe we will see the amount of quality software increase too, just by a much smaller amount than increase of junk.
Eg: 2022 sees 10k pieces of software with 100 meeting a high quality bar. 2027 sees 1M pieces of software with 1000 meeting a high quality bar. It's a 10x increase of quality software and a 10x decrease in average software quality.
These examples are illuminating. They focus on two broad categories: decreasing risk by experimentation (derisking) and offloading tedious tasks (toil).
I was recently speaking with a colleague at a large Silicon Valley company about how they’re adopting agentic coding and the biggest area where they’re seeing late adopters engage is pushing backlog work (toil) over to the agents - even for skeptics, they report sentiment is positive for this use case.