# E127 Limitations

- **100 registered identities are not 100 workers.** Only 8 were configured; peak online/active was 5. Online is 0 after the run.
- **Not 24/7.** Measured research uptime is 177.87 s. No 24-hour, 7-day or 30-day continuity is claimed.
- **Deterministic executors, not LLM agents.** The 100 identities have roles, but tasks are rule/tool audits; no model inference occurred.
- **Single-host SQLite queue.** Durable on one host; not distributed, not Byzantine fault tolerant.
- **Events are hash-chained, not externally witnessed.** Integrity is local; no independent notary.
- **API local.** 8/8 TestClient endpoints; the SSE stream was implemented but not stream-tested. Production gateway deployment not verified.
- **Public pages are static snapshots.** They show the last measured run, not a live feed. No animation implies live activity.
- **No new intelligence result.** E127 audits E126.b; it does not improve continual learning, causal selection, novel environments or adaptation.
- **E126.b failures remain:** causal active FAIL, novel-environment H1 missed, adaptation retention FAIL, external benchmarks BLOCKED.
- **No external reviewer**, certification, SOC 2, hardware attestation, customer revenue or patent.
- **Hugging Face / live production not deployed.**
- **No AGI.**
