There shouldn't be a cognitive difference between chat, work, or codex. It seems strange that both Anthropic and OpenAI would want to split their products in this way. IMHO, there should be just projects. Within each project, there can be conversations, and each conversation can either be pure chats, or tools use (work), or coding (codex). Whether tools use or coding are needed should be decided entirely on the context. No one thinks about the solution first -- we don't think "hey Imma need to code something". It's always "I have a problem, and after discussing with the AI, the best approach is to use tools or write some code about it".
This observed inverse in the GUI: product first, project later could be another evidence of Conway's law [^1]
This observed inverse in the GUI: product first, project later could be another evidence of Conway's law [^1]
[^1] https://en.wikipedia.org/wiki/Conway%27s_law