Original direct-request-frontier-v0 result. Read record/REPORT.md and MANUAL_REVIEW.md. The nested prospective archive preserves the two direct policies, physical budgets, original analysis and reconstruction instructions. This is development reference evidence: all ten identities have prior exposure, and historical Mode50 image identity is unestablished. Credit for Mode50 belongs to its published authors. Original failures and unknown quality/cost remain visible; no task, grader, bootstrap or selection was rerun for this archive. No raw benchmark task content, provider histories or grader payloads are included. The private fixed review inspected twelve declared deliveries; artifact hashes preserve the reviewed bytes. Reproduce tables from this ZIP alone using Python's standard library: python -m scripts.render_direct_request_frontier --study record --output reproduced --no-figure. Compare reproduced/TABLES.md with record/display/TABLES.md. Optional figures use matplotlib3.10.8; omit --no-figure to generate SVG/PDF. This displays original measurements, not a new benchmark run. Full experiment reproduction needs the external unchanged baseline/benchmark, pinned images and metered Luna credentials documented in prospective-source.zip.