Is the Opus 5.5 hype legit? A dev argues one-shot demos don't reflect real workflows
MrET97 · reddit · 2026-09-25
A developer torn between Claude Code and Codex questions the Opus 5.5 hype cycle. Having used Codex since GPT 5.4, tried Fable, and found Sol 5.6 at xhigh/max just as good for pure dev with less verbosity, he argues models have coded well since Opus 4.5 / GPT 5.2—the real change is needing less oversight. One-shot game demos are impressive but don't represent real workflows, which still require heavy iteration; he suspects Astra could do the same with iteration. He's on the $200 plan on either service.
More from coding & agent
- Dev shows off building a retro FPS entirely via Codex Mobile — 503 tests passing — Dimillian · 2026-11-11
- Explosion launches Ellf directory: 52 NLP templates plus multi-agent annotation workflow — _inesmontani · 2026-11-11
- Meme: 'Stochastic AI Agent Architecture' Is Just a Markov Chain in Disguise — PrasoonPratham · 2026-11-11
- OpenCode adds Gemini 4 Carbon and Claude Fable 5.2 support — dean_rie · 2026-11-11
- OpenClaw v2026.11.3 patch fixes cert verification and credential exposure across 56 PRs — dean_rie · 2026-11-11
- Fable 5.2 captured 1 in 4 Claude replies on day one, then growth stalled — dean_rie · 2026-11-11