Releases: andreiturcea-creator/movement-studio-build-week-final
Release list
Movement Studio — OpenAI Build Week 2026
GitHub release draft — Movement Studio · OpenAI Build Week
Proposed tag: v0.1.0-build-week
Tag target: 70679556b5aec824fafbb17b3f733b5c3b33fc87
Status: private draft; do not publish until repository-publication approval and
green public exact-SHA checks.
Movement Studio
Movement Studio is a source-bound content-planning workflow for practitioners.
GPT-5.6 proposes a structured brief and ordered transcript segment identifiers;
deterministic code verifies exact source provenance, derives transcript evidence,
and keeps model-authored editorial copy labelled as inference. The user approves
the only output action: a browser-local JSON review bundle.
Two-minute judge path
No credentials are required.
- Clone the repository at this tag.
- Run
./demo.sh. - Open the printed loopback URL.
- Replay the source-bound brief and inspect one exact excerpt plus SHA-256.
- Confirm review of the synthetic Answer Card.
- Replay the selection plan and inspect its code-derived excerpts, timestamps,
and full transcript hash. - Approve the browser-local JSON export.
Run ./verify.sh --release to reproduce the automated, offline-integrity,
privacy, eligibility, and exact-package gates after the pinned dependency
bootstrap.
Release evidence
- 137/137 automated tests passed on the exact clean candidate release.
- 18/18 offline integrity cases passed across six categories.
- The default judge path is a wholly synthetic offline replay and makes no model
request after dependency bootstrap.
Assets
movement-studio-build-week-entry-7067955.zip
- Exact Git archive of commit
70679556… - SHA-256:
3a1625f83cb2165063e67aa97f4e50c9a8e0aff08006b2cddf2c3fda2f347109 - Integrity-tested and privacy-scanned
live-capture-70679556.json
- Sanitized release-bound execution receipt—not independent third-party
attestation - SHA-256:
1c2d4b91e8c988aeb6739f144d399ba0f67ce43d6c82a3222b82f65952f5d24b - Records one successful synthetic two-request run: brief, then selection
- Both requests returned
openai/gpt-5.6-luna, selected providerOpenAI, and
passed strict validation with OpenAI-only routing and fallbacks disabled - Retains only output hashes and allowlisted model/provider proof; it contains no
prompt/source text, raw response, credential, usage payload, request/account
identifier, or real-person data
Important limits
- Exact excerpts and hashes verify provenance, not semantic truth.
- Review and export controls are user attestations, not authentication of
professional qualifications. - The two-request statement describes the successful release-bound capture; a
future failed attempt could stop after one request. - This Build Week prototype uses synthetic material and does not claim clinical,
production, or measured productivity validation.