Use agent instead of generic AI, and clarify what simulated data means

[mboss37]

Summary

Follow-up to #1, #2, and #3 (all merged): two terminology fixes across all three READMEs.

• Replaced generic "AI" references with "agent," since the loop's roles are literally Claude Code agents (evaluator, proposer, implementer, gate, meta-improver), not a vague generic AI. Also dropped a couple of overly casual phrases ("vibe check") introduced in the previous readability pass.
• Replaced "fake data" with a precise explanation of what each example actually simulates:
  • Tournament: synthetic support messages plus a deterministic stand-in scorer that reads a hidden tag on each test case instead of a real model call.
  • Latency tuner: computer-generated traffic patterns from a fixed random seed, not recorded production traffic.
• Same content and structure otherwise; wording only.

Test plan

• Grep for leftover generic "AI" or "fake" references: zero matches
• Grep for em/en dashes: zero matches
• examples/prompt-technique-tournament: python -m pytest eval/ -q (129 passed)
• examples/latency-tuner: python -m pytest sim/ -q (90 passed)
• python3 template/tools/verdict.py self-test (all 6 checks pass)

https://claude.ai/code/session_011t5gmHpHh68Jcigy5uPPEx

---

Generated by Claude Code