Decart MirageLSD

A dated reading of what is claimed, reported, and independently verified in the current evidence.

Current reading

The readiness gap, in one scan

AI-assisted assembly · derived results

Evidence strength Strong

Decision

Human editorial judgment · 2026-06-28

Proceed with caution

Why: Genuinely novel real-time capability with credible numbers, but Hopper-bound hosting, output-quality gap vs offline models, and long-session drift make it unsuitable for a load-bearing production path today — and the vendor has already moved the frontier to Lucy 2.0, so build against the current model, not MirageLSD.
Next: Run a bounded pilot via Decart's hosted endpoint (Crusoe Cloud exclusive) on a single live-transform use case, and evaluate the newer Lucy 2.0 (30fps/1080p) in the same test; measure latency + drift on real footage before any product dependency.

Constraints

No named blocker is present in the current public projection.

Evidence summary

AI-assisted assembly

Counts and dates only. Raw signals, private excerpts, trust records, and internal corpus material are not published here.

Publication record

Initial public reading

This is the initial public reading. No earlier readiness change is recorded.