Ask HN: Is Anthropic doing too much vibe coding?
The Claude app and Claude Code have been an unusable and buggy mess for me lately, has anyone else been experiencing this? Most of my messages get swallowed after sending them or the responses get interrupted or dropped. Sometimes entire conversations disappear from the sidebar only to reappear later. I’ve learnt that when a message appears not to have gone through or have errored midway, it often comes back with a valid response if I wait a bit and then restart the app.
I wonder if it has anything to do with Anthropic eagerly embracing vibe coding. Have you seen Claude Code's TUI React codebase? The amount of pointless over-engineering for what is effectively just a terminal text application is pretty astounding. I guess it is related to the model as well? Is it Sonnet or Opus? Sometimes the old Opus 4.6 gives more appropriate results. The ones I've personally noticed is the repeating permissions for the same command, rereading the same files over and over. Not for me. I use Claude Code daily and rarely see a bug. What is your subscription plan? Subscription plan has really nothing to do with it. Are you serious? I have an error response at least one time per day We ask ourselves this every day. I suspect Google is equally guilty the industry as a whole is doing too much vibe coding. someone is paying the piper eventually. i am using glm its give same feel like opus Same, glm 5.1 has been so good. I used it with OpenCode, no noticeable differences Bad vibes? Yeah. They promised us vibe coding but no one has promised that vibe will be good