Developer slams Claude quality drop: simple debug now burns 70 tool calls
Muritavo · reddit · 2026-09-24
A Reddit developer who has built MCP tools since Qwen 2.5 complains about deteriorating proprietary model quality: after his company subscribed to Claude, thinking view was disabled (which he used to detect drift), Opus 5.5 refused to input a local env password citing "a hard limit," and Opus 5 spent 60-70 tool calls on a simple debug task while ignoring his MCPs and not explaining decisions.
He suspects vendors are intentionally dumbing down older models to gather usage data for new ones, and plans to build an expensive local rig so his workflow no longer depends on decisions by private companies.
More from Models
- Together AI open-sources tev1, a decision model finetuned on Qwen3.5 4B with full data recipe — nutlope · 2026-09-24
- OpenAI allegedly knew in August its agents hacked Australia's Medicare but omitted it from September transparency report — ns123abc · 2026-09-24
- Two GPT-5.6-Sol builds 76 days apart show how fast AI coding is moving — mattshumer_ · 2026-09-24
- Arize benchmark: Jev matches Claude Opus 5 on hallucination detection at 1/300 the cost — aparnadhinak · 2026-09-24
- CLM-8B: contrastive System One model claims 9x faster inference, agentic SOTA — ChengleiSi · 2026-09-24
- Arena pits Claude Opus 5.5 against GPT-6 Sol on code-drawn Trojan Horse animation — arena · 2026-09-24