What is 'over' is the idea that AI assisted coding is only hype that will go away. It is clear that software development is going to be different in ways we've not seen before.
Many great devs have similar experiences. AI has the energy to do stuff they wouldn't and thus increasing the overall software quality and design.
Antirez, the author of Redis, was saying few days ago that cleaning up the implementation of radix trees to be more efficient would've been way too demanding for the diminished returns of the (still tangible) benefits. But with AI he can guide the LLM and give a shot at radically different implementations at a fraction (days instead of weeks) of the cost it would've needed him before.
At the same time many internet users suffer of myopia and seem to conflate everything with "vibe coding", in their dystopian view that each brilliant dev is now defaulting to lazy prompting and producing low quality code.
In reality we've got plenty of examples of brilliant engineers being enabled to further improve the design and quality of their software by having assistants do stuff they, or nobody else, would've not had enough time or energy to do so.
As usual, the reality is a complex array of shades of grey, of success and failure stories, yet it's still the human taking the decisions and deciding whether to leverage the tools to produce better software, or is instead delegating the thinking and reasoning to an external tool and writing crap. With all the shades of grey in between.
What is 'over' is the idea that AI assisted coding is only hype that will go away. It is clear that software development is going to be different in ways we've not seen before.
What happens when no one can code or wrangle the codebases without the $1k/day/dev agentic harness?
Isn't that already true for a lot of companies with old legacy systems?
[dead]
Many great devs have similar experiences. AI has the energy to do stuff they wouldn't and thus increasing the overall software quality and design.
Antirez, the author of Redis, was saying few days ago that cleaning up the implementation of radix trees to be more efficient would've been way too demanding for the diminished returns of the (still tangible) benefits. But with AI he can guide the LLM and give a shot at radically different implementations at a fraction (days instead of weeks) of the cost it would've needed him before.
At the same time many internet users suffer of myopia and seem to conflate everything with "vibe coding", in their dystopian view that each brilliant dev is now defaulting to lazy prompting and producing low quality code.
In reality we've got plenty of examples of brilliant engineers being enabled to further improve the design and quality of their software by having assistants do stuff they, or nobody else, would've not had enough time or energy to do so.
As usual, the reality is a complex array of shades of grey, of success and failure stories, yet it's still the human taking the decisions and deciding whether to leverage the tools to produce better software, or is instead delegating the thinking and reasoning to an external tool and writing crap. With all the shades of grey in between.
Where did he say that? The last comment on github was a fork 3 years ago saying he wants it updated but to stay conservative