VetPilot

Tracking since May 26, 2026.

Changelog

• Jun 8, 2026
  description

  Doing SWE tasks on Agentic Vet should be about writing the solution, not battling the dashboard. VetPilot is a tiny shortcut layer that lives in your toolbar — built by someone who got tired of the same three monotonous detours every single task.
  
  You know the ones:
  
  🪟 Scrolling past the giant test-case megabook just to read the prompt.
  📦 Downloading every output zip, unzipping each, hunting through `verifier/reward.json` for one speedup number, repeat ×N.
  🎛️ Clicking through the Run Evaluation form for Oracle, then again for Claude, then Gemini, then GPT.
  
  VetPilot collapses all three into single clicks from one small popup:
  
  → Hide the leaderboard chunk when you want a clean view.
  → Capture every speedup at once. One click fetches all outputs, parses them in memory (no zips on your disk), brands them per model, and dumps the numbers in the popup.
  → Launch a new evaluation for Oracle, Claude, Gemini, or GPT straight from the toolbar, using your existing session.
  
  Look, I can't make your Oracle solution faster (sorry 😅) — but the busywork that piles up around it? That's covered.
  
  Stays an extension on purpose. If this ever wants to be a whole app, it has missed the point.

  v1.1.0 — adapts to AV's new checkbox + bundled-zip download flow.
  • Tick evals on the page, then click Capture Speedups; the bundle is parsed in memory.
  • Wrong-page detection: dims the UI and shows a hint when you're not on a task page.
  • Click any speedup to copy the number to clipboard.
  • Cleaner toasts (auto-dismiss, no replay on popup reopen)
  
  Doing SWE tasks on Agentic Vet should be about writing the solution, not battling the dashboard. VetPilot is a tiny shortcut layer that lives in your toolbar — built by someone who got tired of the same three monotonous detours every single task.
  
  You know the ones:
  
  🪟 Scrolling past the giant test-case megabook just to read the prompt.
  📦 Downloading every output zip, unzipping each, hunting through `verifier/reward.json` for one speedup number, repeat ×N.
  🎛️ Clicking through the Run Evaluation form for Oracle, then again for Claude, then Gemini, then GPT.
  
  VetPilot collapses all three into single clicks from one small popup:
  
  → Hide the leaderboard chunk when you want a clean view.
  → Capture every speedup at once. One click fetches all outputs, parses them in memory (no zips on your disk), brands them per model, and dumps the numbers in the popup.
  → Launch a new evaluation for Oracle, Claude, Gemini, or GPT straight from the toolbar, using your existing session.
  
  Look, I can't make your Oracle solution faster (sorry 😅) — but the busywork that piles up around it? That's covered.
  
  Stays an extension on purpose. If this ever wants to be a whole app, it has missed the point.

• Jun 8, 2026
  host_permissions

  https://api.gdm-agentic-vet.turing.com/*, https://*.turing.com/*

  https://api.gdm-agentic-vet.turing.com/*, https://*.turing.com/*, https://storage.googleapis.com/*

Permissions & access

Permissions

activeTabscriptingdownloadsstorage

Host access

https://api.gdm-agentic-vet.turing.com/*, https://*.turing.com/*, https://storage.googleapis.com/*

Screenshots