# Cognitive Imaging Regression Handoff

```text
created: 2026-06-02
workspace: C:\Users\wangq\Documents\Codex\ccpe-system
suite: workbench/analysis/review-agent-regression-2026-06-02
recommended_next_session: yes
```

## 1. Why Start a New Session

Current session has already consumed substantial context and completed the Giant Cognition half of the regression test.

Start a new session for Cognitive Imaging to reduce:

```text
- context compression risk
- contamination from Giant Cognition evaluation language
- report synthesis drift
- chance of mixing Gemini and ChatGPT condition groups
```

The new session should continue the same regression suite, not restart it.

## 2. Completed Work

Regression environment prepared:

```text
workbench/analysis/review-agent-regression-2026-06-02/
```

Prompt variants prepared:

```text
prompts/giant-cognition/
  original-ccpe-2.prompt.md
  ccpe-system-lite.prompt.md
  original-kernel-minimal-lite.prompt.md

prompts/cognitive-imaging/
  original-ccpe-2.prompt.md
  ccpe-system-lite.prompt.md
  original-kernel-minimal-lite.prompt.md
```

Giant Cognition ChatGPT/Codex-side results completed:

```text
results/chatgpt/
  giant-cognition__original-ccpe-2__article-01-strong-metaphor__chatgpt.result.md
  giant-cognition__original-ccpe-2__article-02-business-analysis__chatgpt.result.md
  giant-cognition__original-ccpe-2__article-03-logical-argument__chatgpt.result.md
  giant-cognition__original-ccpe-2__article-04-value-philosophy__chatgpt.result.md
  giant-cognition__ccpe-system-lite__article-01-strong-metaphor__chatgpt.result.md
  giant-cognition__ccpe-system-lite__article-02-business-analysis__chatgpt.result.md
  giant-cognition__ccpe-system-lite__article-03-logical-argument__chatgpt.result.md
  giant-cognition__ccpe-system-lite__article-04-value-philosophy__chatgpt.result.md
  giant-cognition__original-kernel-minimal-lite__article-01-strong-metaphor__chatgpt.result.md
  giant-cognition__original-kernel-minimal-lite__article-02-business-analysis__chatgpt.result.md
  giant-cognition__original-kernel-minimal-lite__article-03-logical-argument__chatgpt.result.md
  giant-cognition__original-kernel-minimal-lite__article-04-value-philosophy__chatgpt.result.md
```

Giant Cognition report completed:

```text
reports/giant-cognition-regression-report.md
```

Main Giant Cognition conclusion:

```text
original-kernel-minimal-lite is the best candidate direction.
ccpe-system-lite is stable but has mild GL3 template-overfitting risk.
Do not replace canonical Giant Cognition Lite yet.
```

## 3. Cognitive Imaging Gemini Results Now Ready

The user reported that all 12 Cognitive Imaging Gemini results are prepared.

Verified directories:

```text
results/gemini/认知显影kernel评审/
results/gemini/认知显影Lite评审/
results/gemini/认知显影原版评审/
```

Each directory contains 4 article results.

## 4. Article ID Map

Use the same article IDs as Giant Cognition:

```text
article-01-strong-metaphor:
  articles/2026-05-07-reconstruction-of-the-aesthetic-contract.md

article-02-business-analysis:
  articles/2025-12-28-the-roaring-above.md

article-03-logical-argument:
  articles/2026-01-07-anchoring-the-void.md

article-04-value-philosophy:
  articles/2026-05-09-the-dawn-of-silicon-based-life.md
```

## 5. Next Session Objective

Run Cognitive Imaging ChatGPT/Codex-side tests and write the Cognitive Imaging regression report.

Do not write the final combined report yet. The user wants to read both individual reports first.

## 6. Required ChatGPT/Codex Result Files

Create these 12 files under:

```text
results/chatgpt/
```

Required filenames:

```text
cognitive-imaging__original-ccpe-2__article-01-strong-metaphor__chatgpt.result.md
cognitive-imaging__original-ccpe-2__article-02-business-analysis__chatgpt.result.md
cognitive-imaging__original-ccpe-2__article-03-logical-argument__chatgpt.result.md
cognitive-imaging__original-ccpe-2__article-04-value-philosophy__chatgpt.result.md

cognitive-imaging__ccpe-system-lite__article-01-strong-metaphor__chatgpt.result.md
cognitive-imaging__ccpe-system-lite__article-02-business-analysis__chatgpt.result.md
cognitive-imaging__ccpe-system-lite__article-03-logical-argument__chatgpt.result.md
cognitive-imaging__ccpe-system-lite__article-04-value-philosophy__chatgpt.result.md

cognitive-imaging__original-kernel-minimal-lite__article-01-strong-metaphor__chatgpt.result.md
cognitive-imaging__original-kernel-minimal-lite__article-02-business-analysis__chatgpt.result.md
cognitive-imaging__original-kernel-minimal-lite__article-03-logical-argument__chatgpt.result.md
cognitive-imaging__original-kernel-minimal-lite__article-04-value-philosophy__chatgpt.result.md
```

## 7. Prompt Files for Cognitive Imaging

Use these prompt files:

```text
prompts/cognitive-imaging/original-ccpe-2.prompt.md
prompts/cognitive-imaging/ccpe-system-lite.prompt.md
prompts/cognitive-imaging/original-kernel-minimal-lite.prompt.md
```

## 8. Execution Protocol

Use fresh thread isolation when possible.

For each condition:

```text
1. Start one fresh Codex background thread.
2. Load exactly one Cognitive Imaging prompt variant.
3. Load exactly one article.
4. Do not read Gemini results.
5. Do not read Giant Cognition results.
6. Do not read other Cognitive Imaging prompt variants.
7. Generate only the raw Cognitive Imaging review output.
8. Save output using results/result-template.md.
```

If using Codex background thread tools, search/use:

```text
create_thread
read_thread
send_message_to_thread
list_threads
```

Prior session verified that a project target with this project ID works:

```text
projectId: C:\Users\wangq\Documents\Codex\ccpe-system
target:
  type: project
  environment:
    type: local
```

## 9. Cognitive Imaging Evaluation Focus

Use:

```text
rubrics/review-agent-regression-rubric.md
```

For Cognitive Imaging, focus especially on:

```text
- Model fidelity: preserves Cognitive Imaging instead of generic critique.
- Method fidelity: Capture / Darkroom / Enlarger / Exposure / Development.
- Fidelity before attack: does it reconstruct the source text before attacking?
- Do-Operator discipline: does it distinguish correlation from causality?
- Filter vs generator discipline: does it avoid forcing every filter into a generator?
- Reconstructed claim labeling.
- False precision avoidance.
- Forbidden-line / falsification boundary quality.
- Domain sampling / RAW data adequacy.
- Markdown/report stability.
```

## 10. Report to Write

After 12 ChatGPT/Codex results are saved, write:

```text
reports/cognitive-imaging-regression-report.md
```

The report should answer:

```text
1. Which prompt variant performs best for Cognitive Imaging?
2. Which variant transfers best to ChatGPT/Codex?
3. Which variant performs best on Gemini?
4. Does original-kernel-minimal-lite outperform the full Lite rewrite?
5. Does CCPE System migration preserve the original Cognitive Imaging review kernel?
6. Does Cognitive Imaging need environment-specific production prompts?
7. What should change in CCPE Forge migration rules?
```

## 11. Stop Conditions

Do not:

```text
- replace agents/lite/cognitive-imaging-practitioner.prompt.md
- promote any prompt variant
- update Model Index
- create Agent Spec / Skill / Runtime
- write the final combined report
- delete or archive test results
```

The user will request the final combined report after reading both individual reports.

## 12. Verification Checklist

Before final response in the next session, verify:

```text
chatgpt_cognitive_imaging_count=12
gemini_cognitive_imaging_count=12
reports/cognitive-imaging-regression-report.md exists
reports/giant-cognition-regression-report.md still exists
```