The AI Coding Velocity Trap

vnturing.github.io

3 pointselcapithanos3 comments

Velocity won't matter. How fast you code isn't what matters to the end user. We select software based on what works the best for us. Interfaces that are slop can only be successful if they fit how we work. If we assume coding is essentially an AI task now, the winner will be the team that builds the best software. Think of it like Apple under Steve Jobs. It wasn't velocity that was important, but attention and intentionality. I don't think LLMs can get us Steve Jobs quality. I see developers focusing on code review and optimizing interfaces to function well.

What about first mover advantage though? Some people believe it's all about being first to market then let others have to convince people to switch to their products. I also think velocity won't matter. But I don't think the companies pushing their employees to use AI are thinking of anything other than "others are using AI and moving fast therefore we need to keep up".

I don't think first mover is going to matter. I think the race to be first is a waste of money. Deepseek sort of showed that LLMs aren't signicantly protected from duplicating results. AI will drastically change how we deal with 3rd party services. Third party services are going to have a very hard time locking in users. You can't lock me in if I can get AI to scrape my data out and translate it into data another provider uses. Going forward I am of the opinion the winners will be the innovators that can create intuitive interfaces. Innovation alone won't be enough. Someone else can rebuild your concept with a better interface.

Edit: fixed intent.

Reply on news.ycombinator.com