How I AI
How I AI

Sep 22, 2026 · 25 min

Claire Vo tests whether Opus 5.5 fixes Claude’s worst habits

I left Claude for months. Opus 5.5 is why I'm back

The episode examines whether a newer model can overcome the verbosity, hedging, and preachiness that drove an experienced user away from Claude.

3 key takeaways
  1. 1Opus 5.5 shows meaningful gains across cost, speed, writing, prototyping, SVG creation, and video workflows.
  2. 2Long-running agentic work improves, but persistent frustrations keep Claude from becoming a clear all-purpose replacement.
  3. 3Codex remains superior in several engineering workflows, making model choice depend on the task rather than brand loyalty.

Don't miss

Claire Vo’s real-world testing finds Opus 5.5 substantially improved, yet still unable to displace Codex across every engineering workflow.

The brief

Claire Vo explains why she abandoned Claude: intelligent answers had become rambling, hedged, preachy, and frustrating enough to earn her label “Claude slop.”

She returns to test Opus 5.5 against real-world work, weighing cost and speed alongside the less measurable question of whether its output is actually less annoying.

The evaluation spans long-running agentic tasks, frontend prototyping, writing, SVG creation, and video workflows, giving the model room to succeed or fail beyond a single prompt.

Opus 5.5 delivers major improvements, but the episode does not turn into a clean comeback story: persistent frustrations remain, and Codex is still better in clear areas.

The broader takeaway is practical rather than tribal: Claude’s return depends on the task, because stronger general performance does not erase meaningful workflow-specific tradeoffs.

Weekly email

State of the Pod, every Monday

The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.

Listen to the full episode and explore every guest, topic, and moment on PodLume.