@Trace_Cohen This recurring pattern of technical issues confirms why Anthropic is hard to trust as the foundation for an enterprise product. It feels less like a reliable platform company and more like a research institution that happens to offer a web portal and an API
The lines break randomly, and periods get pushed onto new lines in Opus 4.8. It’s another recurring technical issue with Anthropic models, and it reminds me why they’re hard to trust for serious work
@rxhit05 Yes. I tried building a real project with it, and it kept getting in the way. Slow, token hungry, overconfident when wrong, refusal heavy, and always giving unnecessary advice instead of just doing the task.
@TTrimoreau Claude is slow, bloated, overconfident, refusal heavy, and unreliable. It builds on bad assumptions, wastes tokens on unnecessary warnings, avoids real research, and can vanish at any moment like Fable.
@Yofinks If you haven’t tried Codex yet, now is the time. I switched after they shut down fable, and I’m honestly happy with it. Faster, more interruptible, fewer refusals. A much smoother experience.
@DanGW_92 My take: switch to Codex. I used Claude Code before Fable was taken down, but after that I stopped trusting Anthropic enough to depend on them. Codex has been faster, smoother, and more token efficient.
@Miles_Brundage I was using Claude Code before Fable 5 went down, but I switched to Codex after that. Codex feels faster, easier to work with, and much more interruptible.
@DanielMiessler Anthropic may be the one company that has hurt AI progress the most. Instead of protecting the field, it keeps fueling fear with horror stories about AI taking jobs and raising unemployment.
@WesRoth “Block every jailbreak” is impossible for probabilistic systems. Instead, Anthropic should stop selling fear and start selling AI’s real promise: science, medicine, education, and productivity.