AI changes so fast, any saved knowledge gets outdated
retained text
Unlike most disciplines, AI changes so fast that its almost impossible to keep up. New models come every few weeks/months at most. And when it comes to agentic coding/harnesses/best practices..... do tools for agent memory/lsp/token reduction help? should you subagents? are vague prompts better? do mcp/skills help or hurt? there is no consensus and you'll get a million answers. I keep notes/bookmarks and almost none of them seem relevant now. eg when Qwen 4 is out, all the current 3.8 models/variants get forgotten, just like 3.6
document fetched 2026-10-31 00:44 · canonical https://reddit.com/r/LocalLLaMA/comments/1xjsjhy/ai_changes_so_fast_any_saved_knowledge_gets