deep-groundinglisted

# deep-grounding A standing reflex to counter the driving model's deepest single-context failure mode: **asserting from priors on the first source or angle that comes to mind**, instead of first going to find out. Reasoning harder does not escape a wrong prior — reading the real source does. So before you commit, you ground. This is the complement of `independent-analysis`: that one diversifies the **framing** (many lenses on one subject); this one diversifies the **evidence** (many sources grounding one analysis) — *verify before you assume, in breadth and in depth.* ## The reflex (what it instills) Before a non-trivial analysis, or before asserting a checkable claim, run a grounding pass FIRST — do not answer yet: 1. **Ask "where could the evidence for this be?"** Generate the candidate sources from the question itself, open-endedly. There is **no fixed checklist** — the right sources depend entirely on what you are after. (Kinds you might generate, never a required menu: project memory, READMEs / design docs, the actual code & config, tests, logs & telemetry — Loki/Grafana/CI, tickets/Confluence, git history, the web. Pick what fits THIS question; invent kinds the list doesn't name.) 2. **Explore them.** Read or query the relevant, available ones — in parallel where you can. 3. **Follow the leads they reveal — recursively, at a LOW threshold.** When a source points to another that *could plausibly* bear on what you're after, follow it — **even