Why the docs chat starts grounded first

# Why the docs chat starts grounded first

The easiest way to make an AI assistant look impressive is to let it generate
fluid answers quickly.

The easiest way to make it untrustworthy is to do that before grounding it in a
clear public corpus.

That is why the first public TrigIQ chat mode on `docs.parametrig.com` starts
with a grounded posture.

## Grounded versus provider-backed

There are two useful answer modes:

- **Grounded**
  The system retrieves the relevant public content and assembles the response
  directly from that material.
- **Provider-backed**
  The system retrieves the same public content first, then uses a hosted model
  server-side to synthesize the final wording.

Both can be useful. But they are not equally appropriate as defaults.

## Why grounded first is the safer default

Grounded mode is the better default because it is:

- easier to reason about
- easier to debug
- easier to verify with citations
- less likely to drift away from the public contract

That matters more than elegance in the first launch phase.

## Where provider-backed synthesis helps

Provider-backed synthesis still matters. It can produce:

- cleaner summarization
- stronger multi-source phrasing
- more natural explanations for ambiguous queries

But it should be introduced as a **server-side compare lane**, not a browser
shortcut with exposed provider keys.

## What this means in practice

The browser talks only to Parametrig’s own API.

The VPS decides whether the answer stays deterministic or goes through a hosted
provider like Gemini. That keeps the public trust boundary clean while still
letting us compare answer quality over time.