Simon Smith 退订个人 ChatGPT：个人 agent 太多

Simon Smith @_simonsmith原推发布于 2026.10.05

在 X 上打开原推

I canceled my personal ChatGPT subscription after being a longtime personal subscriber, but it's not what you think, and I continue to use it via my ChatGPT Enterprise account. The breaking point: personal agents. For my personal life, I'm currently Muse-pilled. Perhaps I'll regret this, as Meta doesn't have the best track record when it comes to, say, data privacy. But for now, I'm loving it, it's serving me well in my personal life, I'm curious to see where it goes, I'm hopeful about Meta's Confidential VM, I use Meta Ray-Bans, I've started playing with the Muse Gadget SDK, and I generally like Meta's direction with personal superintelligence. This said, I still primarily use ChatGPT for work via our ChatGPT Enterprise account. I love features of the Work/Codex interface, like appshots, and I find GPT-6.1 Sol and Astra to be solid workhorse models. I also like having image generation natively. For my most complex and design-heavy tasks, I now turn to Claude, as we're also fortunate to have Claude Enterprise accounts. I find Opus 5.5 magnificent for presentations in particular. We can do research together, I can barf out ideas, we can iterate on visual direction, and then it will one-shot write and design a gorgeous presentation on the level of some of the best copywriters and communication designers I've worked with. (With some guidance, like "no eyebrows!") But now that we've got personal agents in ChatGPT, and they're proactive, I definitely don't need both Muse and Dots for my personal life. I don't need to be inundated by messages from multiple proactive agents. I don't need to try and manage multiple agents who might try to work on the same thing. It just gets unnecessarily complicated, like having two personal assistants who are both trying to help you without being able to talk to each other. Simultaneously, I've got Dots in our Enterprise account, where we've started to find their proactivity in particular very useful. I want my Dot to be able to reach me about work stuff on my phone. So, my setup is now: Muse for personal life, ChatGPT and Claude for work. Muse for my personal proactive agent, ChatGPT Dots for my work proactive agent. This keeps things clean and nicely separated for me. This said, I'm absolutely open to changing again at any point, and to this end increasingly keep my personal life system of record cleanly separated from my AI tools. I can connect my personal Google Workspace and Todoist accounts to any AI tool, and bring lots of context to the table right away.