Scraping 30 days of team review comments via Frame MCP to train editing agents on real feedback
alexgoughcooper · x · 2026-10-13
The author shares a feedback-loop workflow: using Frame MCP to scrape every review comment their team left in Frame.io over the past 30 days (3,181 comments across 1,200 video files in 25 projects), then auto-bucketing the most common errors.
The resulting document gets injected into their editing skills/QA agents, so the agents are trained on exact feedback from senior strategists. The doc ranks where review time goes: caption text corrections top the list (800 comments, almost all avoidable), followed by "cut/remove a line" (700).
Core method: feedback loops are everything — turn documented human review history into agent training data. Any team can replicate this.
More from coding & agent
- Cursor now ships real Supabase backend changes via MCP, no SQL needed — XFreeze · 2026-10-14
- Developer: superhuman agents finished the last 20%, so vibe coding is dead — josh_bickett · 2026-10-14
- MIRA-AC trains research agents on proxy feedback, lifting symbolic regression from 50 to 63 solved — anirudhg9119 · 2026-10-14
- MIRA-AC behavior shift: hypothesis-distinguishing tests up 16.8% to 28%, stalled tweaks halved — anirudhg9119 · 2026-10-14
- MIRA-AC RL details: stitching decisions into one logical episode, updating only the meta-reasoner — anirudhg9119 · 2026-10-14
- MCP Event spec debuts: agents that listen to your system and act proactively — irvinebroque · 2026-10-14