Remove autoresearch simulator artifacts

Author: ShubyM

agent.md

@@ -176,36 +176,6 @@ The CLI prints the exact service name for the session it generated.
 http://localhost:8080/
 ```
 
-## Local UI Stress Simulation
-
-The UI can be exercised without a real agent or OpenRL backend by generating a
-local log root with the same files the harness writes:
-
-```bash
-cd examples/autoresearch
-uv run python -m harness.simulate_ui \
-  run_name=ui-stress \
-  researchers=2 \
-  attempts=4 \
-  delay_seconds=0.2
-```
-
-In a second terminal, serve that generated run:
-
-```bash
-cd examples/autoresearch
-uv run python -m harness.serve \
-  run_name=ui-stress \
-  serve=True \
-  interval_seconds=1
-```
-
-Then open `http://localhost:8080/`. The simulator writes researcher manifests,
-attempt manifests, `agent.log`, `attempt.log`, `metrics.jsonl`, `notes.md`, and
-`diff.patch` while the UI is watching. It intentionally includes concurrent
-researchers, growing logs, ANSI output, stream-json agent events, split UTF-8
-writes, metric updates, diffs, and a simulated failure.
-
 Researcher pods use a shared init container to wait for vLLM, the trainer
 worker, and the gateway before the agent starts.