Dots 健忘和卡顿背后是 OpenAI 的锁定策略

rompasso @0xrompasso原推发布于 2026.10.11

在 X 上打开原推

Dots keeps forgetting what you told it yesterday. OpenAI doesn’t need it to remember - it needs you wired into four thousand apps. This isn’t a reliability story. The stage freeze and the memory complaints aren’t the point. Dots sits on top of ChatGPT’s own user base and quietly plugs each person into an integration layer that’s expensive to leave. What’s actually being built underneath the complaints: → Every dot created ties an account to a sprawling catalog of connected apps, so switching providers later means reconnecting all of them → The same week a dot froze mid-demo, OpenAI made creating one the default inside the ChatGPT app itself, not an opt-in extra → Each dot hands finished work straight to Codex, so a product that looks like a personal assistant is really routing people toward a coding subscription → Complaints about forgotten context sit beside a weekly user base in the billions, and only one of those numbers moves the business → None of the reliability problems block what the product is actually optimizing, which is how many people get wired in before it gets good One number got buried at launch next to four thousand connected apps: 1.2 billion people already open ChatGPT every week. The bugs are real. They’re also beside the point. You’re not watching a product ship. You’re watching a distribution funnel get built in public. How the app layer turns a shaky agent into a lock-in is in the article below 👇

这条是引用着另一条写的：x.com/0xcodio/status/2105965239158104215