There appears to be three basic schools of thought to LLM/GPT changes in software, one of which seems to be an illusion. The three schools are: - Replace People: Software engineers will be replaced… | Simon Wardley | 38 comments

There appears to be three basic schools of thought to LLM/GPT changes in software, one of which seems to be an illusion. The three schools are: - Replace People: Software engineers will be replaced by agentic swarms. It's happening fast and you'll be gone next year. This is basically where the extreme of Silicon Valley and Vibe Coding are. Replace Tasks: Software engineering is changing. It's adapting to a world of agentic swarms and practices are slowly co-evolving. It'll take a bit of time. No Change: Software engineering won't change. I have yet to find a single person who actually believes in "no change". The only time I hear about this mythical "no change" crowd is when the high priests of "Replace People" write articles about them. However, given these high priests started claiming that software engineers would be gone in a year (over a year ago) and I see no signs of that, I'm far from convinced of their ability to predict the future or separate fact from fiction. Ignoring the illusionary "no change" school, the remaining two schools which do seem to be real are the "Replace People" and the "Replace Tasks". Both of these categories appeared in the population study of companies that I ran in 2020, finishing in 2021. I've attached the results and you can read more about it in the comments. As far as I could determine in 2021, AI was leaning towards a replacement of tasks in the next generation of companies. Please also remember, I was speaking about vibe coding (what we used to call conversational programming) and intelligent agents over a decade ago. So, none of these changes are new or surprising to me. There are a lot of thorny issues we have yet to tackle, a lot of questions that need to be asked from the loss of human skill in reasoning to the loss in the chain of comprehension and what practices do we need to maintain such networks. But, this is normal. Practices can take 5-8 years to co-evolve and we are only a few years in. Yes, engineering will change. Yes, you should be learning and experimenting with these systems. That's a given. But I have yet to see anything serious to demonstrate that we won't see replacement of tasks and co-evolution of engineering. I've certainly seen companies reducing staff count, only to start rehiring later. Instead, the main thing I've noted over the last year is furious backpedalling on statements of how AI will replace coders, changing them to "10x more productive" (another questionable claim). My only advice is to learn, to experiment and be wary. When you read articles claiming software engineers will be replaced, how you need to adapt immediately to using swarms or claims of 10x … consider the author, how they are funded and why you are being fed FOMO. When you read articles claiming that this or that are the right practices, be mindful that those practices are still emerging, we are still learning. As with the early days of cloud, there are a lot of chancers about. | 38 comments on LinkedIn