2026-09-14
A growing chorus of paying customers is reporting that Opus 5 is no longer delivering the quality that justified their subscriptions. Across recent posts, users describe arriving at the same conclusion from different angles: heavy daily coding work that used to feel effortless now produces slip-ups, regressions, and outputs they no longer trust. The pattern is not a single bad day or a lone grumpy thread — it is a recurring complaint from the people who pay the most and use the product the most.
One-liner: · Status: Payment-backed · Score: 484.71
Evidence: 33 independent content items across 2 source families (33 distinct authors).
"I'm canceling my Claude Max Pro 20x subscription –…" — source
"I'm thinking about canceling my subscription becau…" — source
"I have two Codex Pro plans and a Claude Max subscr…" — source
When frustration surfaces independently on two distinct communities like Hacker News and Reddit, it stops looking like an isolated rant and starts looking like a signal worth taking seriously. The voices captured here are specifically paying subscribers — including Max Pro 20x holders and multi-seat Pro accounts — which means the issue is showing up in the cohort with the highest expectations and the most direct leverage to churn. Cross-source agreement among high-value customers is the kind of evidence that tends to move product, pricing, and roadmap conversations faster than any single complaint.
Source independence: families hackernews, reddit; non-Reddit hackernews.
These are public-discussion signals, not verified purchase data; treat single-source claims cautiously. No explicit counter-evidence was surfaced among this period's qualifying signals.
Next, pressure-test the narrative by interviewing long-tenured power users on each platform to pin down whether the perceived drop is consistent across task types or concentrated in specific workflows like code generation, refactoring, or agentic loops. Independently, check public and internal evals — token-level reasoning benchmarks, code-correctness suites, and regression-diff metrics on identical prompts from earlier model versions — to see whether measurable quality shifts line up with the user-reported timeline. Finally, monitor churn telemetry and refund-request volume on Max and Pro tiers over the same window the complaints cluster in, so the qualitative signal can be weighed against a quantitative one.
→ Full evidence dossier · Methodology
--- Want this market watched for you? Create your Market Watch.