Hypothesis on Post-GPT-4 Future Demand for Software Developers

Question

I think GPT-4.5 or GPT-5 with certain Plugins could increase developer productivity by 3x to 10x. The ability to run code and get feedback to iterate will allow it to get much better at solving unseen problems.Consequences: * Harder problems will be solved by top 10-20% of developers, much more quickly than ever before. * More common apps and components can be custom made-to-order by near-future-GPT-plus-Plugins as specified by smart end users or system analysts. * If the amount of software work does not increase proportionately, the demand for software devs will decrease. There will likely still be limitations for problems far from the distribution of the model's training data. Only full-on human-level AGI can solve that sort of problems, and no one can predict with certainty what the world will look like once it emerges.What do you think of the above hypothesis?

crop_rotation · Accepted Answer

3x or 10x are very different things. A 3x increase is a massive productivity bump which will cause more features with maybe less developers.
A 10x increase will just outright restructure companies. It will reduce so many overheads that it will self multiply. For example in an org of 110 people there are several teams, which means meetings, fight over who owns what, fight over where features make sense, who get how many engineers, which manager is next in line for promotion.
10x productivity => 11 people can do the same job with 1 manager, no politics, no arguments. (There will still be interaction outside the org, but that existed before too and was at a much broader level).

lordkrandel · Answer

I highly doubt any of that. There's no Copilot making my job easier. 90% of my job is understanding what needs to be done, the remaining 10% is writing code. So no "many"fold improvement for me. Or for any developer I know.