Some of my ideas on AI coding devmethodology

in response to a post by Robert Erricsson

I found your blog post constructive and pragmatic, which is great to hear from leaders like you. I also support optimism (intermediate-term), similarly. The 100X narrative is inspiring, and it is possible IMO. For 100x, new software engineering methodologies need to be developed, both on human and AI side. I believe AI need to follow certain methodologies, in too of them ( like harness). and those who develop this for AI will stand out in productivity. The challenge, as you put it , is that it is changing “realtime”. I also agree with supporting the narrative that Computer Science as industry is being restructured and transformed (rather than focusing on negative aspects). After using AI heavily, I seen the realities that need to be addressed before attaining the 100X productivity. Usually, AI it works well at 80% initially, but pushing to fix the 20%, the codebases become too complex and twisted in certain ways. progress stagnates, in an irreversible way. But there are technical techniques inspired by concepts from CS itself, that can be used for controling and fixed those. So, IMO, it is not too hard to develop new “workflows” (human involvement , a PO/PM-style involvement ) plus the methodology-like ( augmentimg AI with specific non-AI classic CS tools: but more “dev”, coding, dev-process, rather than purely PM/PO ). So, for a developer, it will feel that a new discipline emerged, not just PO/PM-level workflow, but at coding level. (If my suggested approach is used). I haven’t seen this aspect of AI-assisted coding and development being discussed much. I have more details, if needed.