# Desk log CDX-SAME-FOLDER-20260822

**Status:** First-party sanitized demo; not a customer case, benchmark, third-party dataset, OpenAI test, certification, media evaluation, or endorsement
**Page:** https://infinisynapse.com/en/blog/codex-data-analysis
**Run ID:** `CDX-SAME-FOLDER-20260822`
**Date:** 2026-08-22 (Tuesday)
**Operator:** InfiniSynapse Data Team
**Attestor:** William Zhu, InfiniSynapse cofounder ([GitHub @allwefantasy](https://github.com/allwefantasy))
**Contact for contradictions:** zhuhl@infinisynapse.com

## What this file is

A downloadable record of a Codex local-only session versus a separately configured first-party shared task. Companions: [aggregate CSV](https://infinisynapse.com/blog-media/codex-data-analysis/downloads/aggregate-CDX-SAME-FOLDER-20260822.csv) and [verify script](https://infinisynapse.com/blog-media/codex-data-analysis/downloads/verify-CDX-SAME-FOLDER-20260822.py). The CSV contains two observations, not raw or source data.

## Six-step method

1. Record Codex version, config, sandbox, and approval policy.
2. Issue a scoped credential without publishing its value.
3. Configure the first-party CLI and instructions; record hashes and source permissions.
4. Start a goal and record run ID, timestamps, and status.
5. Record SQL and artifact hashes.
6. Ask a reviewer without Codex to reopen the task; preserve failures and wall clock.

## Source and goal

| Field | Value |
|---|---|
| Source | Read-only replica the desk is authorized to read |
| Goal asked twice | “same experiment readout as last week” |
| Contrast | Codex local only vs same CLI → shared task |
| Codex version/config/sandbox/approval | Not recorded |
| CLI/instruction hash | Not published/not recorded |
| Credential scope | Sanitized summary only; value never published |
| Source permissions | Read-only sanitized replica |
| Timestamps/status | Summary retained; exact timestamps not recorded |
| SQL/artifact hashes | Not recorded |
| All failures | Not fully preserved |

## Results

| Retrieval state | Task id | Memo | Chart | A/B note check |
|---|---|---|---|---|
| Codex local only | 1 | 0 | 0 | 0 |
| Same CLI → shared task | 1 | 1 | 1 | 1 |

Wall-clock for the shared-folder run: 18 minutes (warehouse time excluded).

No independent reproduction was known as of 2026-08-28. Independent replication should disclose every field above, the local-only baseline, outputs, failures, wall clock, and conflicts of interest.

## What you may cite

- Artifact counts 1/0/0/0 → 1/1/1/1, ~18 min wall-clock, run ID, Cursor user audit in `/tasks`

## What you may not claim

- Customer case, benchmark, third-party dataset, OpenAI test, certification, media evaluation, endorsement, or raw/source data

## External context (not validation of this run)

- [OpenAI Codex documentation](https://developers.openai.com/codex) (retrieved 2026-08-28)
- [OpenAI Codex repository](https://github.com/openai/codex) (retrieved 2026-08-28)
- [Codex security](https://developers.openai.com/codex/security) (retrieved 2026-08-28)
- [Codex configuration](https://developers.openai.com/codex/config-reference) (retrieved 2026-08-28)
- [W3C PROV-O](https://www.w3.org/TR/prov-o/) (retrieved 2026-08-28)
- [OpenTelemetry traces](https://opentelemetry.io/docs/specs/otel/trace/) (retrieved 2026-08-28)
- [NIST SSDF SP 800-218](https://csrc.nist.gov/pubs/sp/800/218/final) (retrieved 2026-08-28)
- [NIST AI RMF](https://www.nist.gov/itl/ai-risk-management-framework) (retrieved 2026-08-28)
- [OWASP GenAI/LLM Top 10](https://owasp.org/www-project-top-10-for-large-language-model-applications/) (retrieved 2026-08-28)