OpenCode (DeepSeek V4 Pro via OpenRouter) in a Vercel Sandbox. Starter only (AGENTS.md, serve.py, player/, helpers/dom/view_kit). No hardcoded demo code. Cursor did not edit workspace/.
2026-10-15 · task eHUL07FwbriO · opencode · openrouter/deepseek/deepseek-v4-pro-0813
Build a hair and scalp clinic. Each round is one person. I can look at the whole scalp, zoom into a region at medium or high magnification, and switch between two lights: one that shows vessels better and one that shows scale and fine hairs better. I can part the midline, do a gentle pull test, and grade how advanced the loss looks. Then I name the hair-loss pattern, or on some rounds I mark each hair's root and tip with type, growth state, and how curly it is. Saying I am not sure should be safer than a confident wrong label. A cutaway that shows the inside of the skin should not raise the score. Show the photo, my tools, and the score after I submit. Do not show the true diagnosis before I submit. Several cases. The same case should look the same if I restart it.
Running the prompt-created workspace on this machine — not a mock. Save an episode, then open Replay below to scrub the timeline.
opening · 0 ms
Exact files the agent wrote — not rewritten by hand.
# AI : shared Action contract — the page posts {"move": "..."}; keep this shape.
from pydantic import BaseModel
class Action(BaseModel):
move: str