I don't really see the appeal for #agentic #AI, nor #LLM autocorrect, once you interrogate what they are past the market...

I don't really see the appeal for #agentic #AI, nor #LLM autocorrect, once you interrogate what they are past the marketing hype.These systems either:1. Non-deterministically generate code, then require you to check it. 2. Non-deterministically propose a sequence of steps. This is usually meant for quick project scaffolding. You are then meant to approve or reject. Uncritically adopting these technologies shows lack of #systemsthinking skills and foresight: you must already have the skills and experience needed to judge #slop outputs. Not to mention, you risk skill atrophy, not passing on knowledge to juniors, along with other issues that have been repeated over and over on Mastodon.1/2

Read Original

Related