Writing code with Bing Chat 'was' a viable business idea but the restriction of 4000 characters has changed my perspective. With a 10k limit I was able to get BC to generate code then run it passed Claude for ideas and then also Bard if it got stuck in a loop of errors. Sometimes BC got to the right answer in hours after repeated 'this doesn't work because of.... can you fix it' etc and doing what it said. It was not entirely a brain-free process and could also be stressy when it kept getting it wrong. With a 4k limit getting it wrong was more likely because it didn't see all the code to fix it.

Claude can do it - but it often introduces code ideas from other platforms and then apologises. So can can waste a lot of time. It also only gives partial code fixes. Bard is similar and looks like it is writing good code but it often amends the working code and because that isn't the issue you may not see it has messed up your code.

You need to be strict with keeping 'good' versions but it is hard when you reiterate in seconds. When you mess up a good version and then spend a week trying to get back to square 1 is a painful experience. You never quite get back to the great solution you had. But other times it will suggest something you never knew about and it is mind expanding.