https://llmvibes.ai/research Independent analysis of AI model quality and incidents from the LLM Vibes data set. en-us Sat, 25 Apr 2026 12:00:00 GMT https://llmvibes.ai/research/claude-april-2026 https://llmvibes.ai/research/claude-april-2026 Independent sentiment data caught Claude Code grumbling on March 26, the day Anthropic shipped the cache bug. 28 days before the postmortem. noreply@llmvibes.ai (David Kelly) Sat, 25 Apr 2026 12:00:00 GMT claude anthropic postmortem incident case-study https://llmvibes.ai/research/how-llm-vibes-classifies-sentiment https://llmvibes.ai/research/how-llm-vibes-classifies-sentiment The full pipeline from scraper to score. Five platforms, 12 complaint categories, a volume-weighted 0–100 score, and the failure modes we've documented but not yet solved. noreply@llmvibes.ai (David Kelly) Sat, 25 Apr 2026 12:00:00 GMT methodology https://llmvibes.ai/research/cross-model-deltas-march-april-2026 https://llmvibes.ai/research/cross-model-deltas-march-april-2026 Comparing absolute scores or even bug-window deltas across LLM Vibes models can mislead you. What actually caught Claude's March 2026 regression was the post-fix recovery shape: ChatGPT recovered, Gemini stayed flat, Claude kept sliding. noreply@llmvibes.ai (David Kelly) Sat, 25 Apr 2026 12:00:00 GMT cross-model case-study claude chatgpt gemini grok