Claire Vo · 2 min read · Gen AI Tools
I left Claude for months. Opus 5.5 is why I'm back
Read the original on lennysnewsletter.com →
Product leader Claire Vo shares her hands-on return to Claude after months away, driven by Anthropic's release of Opus 5.5—a faster, 40% cheaper model with a revamped alignment approach. She ran the model through a week of real-world tasks including agentic workflows, frontend prototyping, and writing tests, delivering an honest assessment of where it excels and where it still falls short. Her verdict reshapes how she splits her AI model stack between Claude and OpenAI's Codex.
Key Takeaways
- Opus 5.5 is 40% cheaper and faster than Opus 5, making the cost math significantly more attractive for long-running agentic tasks where per-token pricing compounds quickly
- Anthropic's updated alignment approach reduces the preachy disclaimers and hedging that drove many power users away, though some frustrating limits remain
- Opus 5.5 emerged as Claire's go-to for frontend prototyping, delivering strong results on a full homepage redesign and SVG benchmark
- Claude demonstrated a firm refusal during testing, illustrating that Anthropic's safety posture still has real boundaries—particularly around cybersecurity use cases
- Codex still wins in specific areas, leading Claire to maintain a split model stack rather than consolidating entirely on one tool
Excerpt
Listen or watch on YouTube , Spotify , or Apple Podcasts I’ve been off Claude for months. Not because it got dumb, but because it got annoying. The rambling, the hedging, the preachy little disclaimers on tasks that didn’t need them. I moved most of my daily work to Codex and I didn’t miss it. Then Anthropic shipped Opus 5.5: 40% cheaper than Opus 5, faster, and with what they’re calling a fundamentally different alignment approach. I ran it for a week across real work, including four long-running agentic tasks, a full ChatPRD homepage redesign, an SVG benchmark, and one…