{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/0a0d06dcc3d44874b881c3e7ce9e403e\" frameborder=\"0\" width=\"1712\" height=\"1284\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1284,"width":1712,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1284,"thumbnail_width":1712,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/0a0d06dcc3d44874b881c3e7ce9e403e-05044631f3e2e7f7.gif","duration":145.467,"title":"How We Stopped AI Cost Ambushes","description":"This Loom explains how Stackspend was created to prevent surprise AI infrastructure and API costs. Andrew describes getting ambushed by AI bills across tools like Cursor and Claude Code, and a major incident where a Google Cloud account was infiltrated and $12,000 in Gemini API requests were submitted to generate training data. The team built Stackspend to provide a 360 view of AI usage across providers with real-time alerting, forecasting, anomaly detection, and GitHub-linked cost history to identify likely release diffs. The dashboard also includes an explorer and tracked tasks for digging into issues and seeing what changed and when."}