# E114 limitations

1. **Not AGI.** Constructed simulations (binary causal worlds, list-DSL tasks, synthetic classification streams).
2. No model inference and no language-model training in E114. NeMo/RL/distillation BLOCKED (no GPU; CPU shared with E113).
3. Collective intelligence NOT TESTED here (delegated to E113).
4. External benchmarks (AMA-Bench, LongMemEval, LoCoMo, SWE-bench, scientific discovery) NOT TESTED.
5. Governance is local: in-process decision point, one-host file-locked enforcer; not the production MCPGate; ACS conformance not tested.
6. The mission ran for seconds, not 24 hours.
7. Designed hypothesis spaces; the grown grammar contains the shift generator's forms.
8. Same project, same host, self-generated keys; the verifier was written by a different session, not an external reviewer.
9. Skills cover one environment family; the eight professional domains are not covered.
