Track: HF / NVIDIA momentum — client-first complement strategy

## Summary

Tracking issue for the **parallel momentum strategy** toward Hugging Face — in light of NVIDIA’s announced acquisition of Hugging Face (~$12.9B, Sep 2026) — while Traverse keeps shipping governance/placement work.

**Pitch (complement, don’t compete):** Hub model → pinned, signed, **client-first** Traverse capability; cloud/edge only when UMA placement defers. Not “a better Transformers.js.”

## Parallel tracks

### 1. Narrative
- [ ] Public one-pager / FAQ: HF = provenance; Traverse = governed product unit + placement
- [ ] Explicit non-goals: don’t replace Transformers.js / Hub discovery / training

### 2. One sharp demo (momentum artifact)
- [ ] Pick one small HF-derived agent already in registry (e.g. Whisper / DistilBERT / Silero)
- [ ] Demo: same contract on browser/local; show placement/defer story in trace when available
- [ ] Short write-up or Space/README that Hub developers can open in one click

### 3. Hub-native surface (after provenance lands)
- [ ] Depends on #571 (object `ai.models`) in use
- [ ] Model-card badge and/or Space “Export to Traverse” scaffold (fills pins + registry PR path)
- [ ] Keep #575 as the technical Hub-integration bucket; this issue tracks GTM sequencing

### 4. Trust wedge (enterprise / offline)
- [ ] Digests, signatures, Spec 025 capability licensing, immutable versions
- [ ] Decision 124 model pins — no live Hub fetch at runtime
- [ ] Messaging aimed at regulated / offline consumers of Hub models

### 5. Relationship
- [ ] Warm path into HF/NVIDIA open-source or edge/device folks
- [ ] Lead with demo + complement narrative (open Hub stays open; we package client-first)

### Engineering (parallel, not blocked on Hub)
- [ ] Registry #571 — FR-017 object `ai.models`
- [ ] Traverse #1459 — publish fail-fast for object models
- [ ] Placement: honest docs; next defer executor (edge *or* cloud) when prioritized

## Do not
- Cold pitch “HF needs our runtime to run models”
- Required live HF in merge CI
- Promise multi-host defer before a second executor exists

## Related
- Decision-log entry 124
- https://github.com/traverse-framework/registry/issues/571
- https://github.com/traverse-framework/traverse/issues/1459
- https://github.com/traverse-framework/registry/issues/575 (Deferred technical Hub integration)
- NVIDIA ↔ HF acquisition context (announced Sep 2026; close expected H1 2027)

## Status

**Active tracking** — GTM/momentum. Technical Hub export remains Deferred on #575 until #571 is live and Track 2 demo exists.

Start with this tracking issue, Decision-log entry 124, and the linked registry issues #571 and #575 plus Traverse #1459. Review the five parallel tracks and their dependencies before proposing work; done means the sequencing, demo, messaging, and deferred technical integration are explicitly resolved rather than merely tracked.

Written by the indexing model from the issue text.