Raw LLM Responses

AI has absolutely no hope of replacing developers, but it does stand a good chance of skyrocketing demand for developers. If AI means more people build more software, that inevitably means there will be more software-exposed firms that need to hire developers to maintain and extend that software. The fact that LLMs produce *custom code* makes that even more true: since the code LLMs are producing is not standardized into a product (like no-code platforms or SaaS products), but is rather bespoke code, once it becomes so bloated and complex that the LLM can't handle it anymore, it will require human programmers to maintain and extend. So: - the productivity of software developers will go way up, because they are able to produce more software with the assistance of AI - the demand for software engineers will go way up, because more people will have custom bespoke code in need of human maintenance and extension - the demand for software engineers will go way up, because firms newly entering software with LLM-generated code will discover new problems and contexts that are outside of the LLM's distribution AI gets better by saturating their training distributions. They do not show any ability to get smarter beyond their training data, or generalize creative solutions to new problems outside their training data. But software engineering, at its vanguard, has always meant developing creative ways to solve new problems with software. If AI replaced all the software developers, it would mean all forward progress in software would halt: AI would be better than anyone at solving all our current problems, but completely unequipped to solve any future problems. For AI to continue being useful, it will always have to live alongside humans: learning from our creativity and saturating the new distributions that human creativity outputs.