The central problem that Atlas targets is thinking with AI. AI clearly has the potential to augment/extend/expand our thinking, yet it risks us completely delegating away/depending on AI to do the thinking for us.
Atlas aims to be a knowledge workspace - a workspace for thinking. Claude and Codex are designed for doing work, not expanding one’s understanding. Eventually, one reaches cognitive bottleneck, or even collapse, as we concede our understanding, willingly or unwillingly, to AI systems.
On the other hand, traditional tools like Obsidian (and various ideas like LLM wikis) focus on purely-human or purely-AI thinking. There is a lack and a need for human-AI thinking, to help us think with AI.
Thus, Atlas aims to be a space for thinking with AI. In Atlas, human thoughts are clearly separated from AI outputs. Yet, they’re integrated and interfaces with each other.
