mosshot🥚

obsidian + claude code is the evolution of knowledge management. i beg to differ. a few months ago, people were talking about using obsidian as a memory layer for claude code. now, i'm seeing more conversations questioning whether that setup actually solves the memory problem. but i think we're asking the wrong question. a folder full of markdown files, scripts and instructions telling an ai what to remember can be useful. i understand the appeal. you want an ai agent that remembers your preferences, knows what you've researched, picks up where you stopped, and doesn't make you explain the same thing every time you open a new session. that's a problem worth solving. but there's a distinction missing from a lot of these conversations. having somewhere to save information is not the same as having a reliable memory system. a markdown file can preserve a decision. but can your agent find that decision when it needs it? ▹ can it tell the difference between an old conclusion and a newer correction? ▹ can it recover that information if the original machine is unavailable? ▹ can another compatible agent access the same information without you manually moving files between applications? these are different problems. and a folder of notes doesn't automatically solve all of them. to be clear, obsidian isn't the problem. neither is claude code. local files can be an excellent way to store knowledge, inspect what an agent has saved, and maintain control over your information. the problem is treating a convenient setup as though it has solved the entire memory problem. when you start building agents that work across sessions, machines and applications, you have to think about persistence, retrieval, recovery, portability and deletion. this is where projects like @LighthouseWeb3 comes in