"What is the 'model-lock tradeoff' (Gemini CLI)?"	Gemini-ONLY: deep integration with Gemini features (long context, multimodal, Google ecosystem) at the cost of ZERO model portability. Cannot swap to Claude/GPT/open. Future-proof test fails by design.	harness-engineering::dd05::recall
"Gemini CLI hits a rate limit. How does it differ from Aider in the same situation?"	Aider (100+ models via LiteLLM) FALLS BACK to another provider. Gemini CLI CANNOT — it's locked. It waits or fails. The rate-limiting controversy illustrates the lock-in risk concretely.	harness-engineering::dd05::analysis
"Name 3 GAINS from model-lock (Gemini CLI)."	(1) Deep Gemini integration (long context, multimodal). (2) Native Gemini tool-calling format. (3) Google ecosystem integration (Search, Workspace, Cloud). These are impossible in a multi-model harness.	harness-engineering::dd05::recall
"Name 3 COSTS of model-lock."	(1) Zero portability (cannot swap models). (2) Single point of failure (provider degrades → harness degrades). (3) No leverage on terms/pricing. Supply-chain risk (ASI08 variant).	harness-engineering::dd05::analysis
"Does Gemini CLI pass the future-proof test (Module 1)?"	NO — by design. The test asks 'does performance improve when the model upgrades?' Gemini CLI cannot upgrade independently; it's locked to Google's release cycle and model availability.	harness-engineering::dd05::application
"Why is model-lock described as a 'supply-chain risk'?"	Single-supplier dependency. Every provider risk (rate limits, degradation, terms change, discontinuation) becomes a harness risk with no mitigation. Same risk structure as ASI08 (supply chain).	harness-engineering::dd05::analysis