Editing Openai/694231e4-2244-8012-83ef-e82fe0c756eb (section)

=== User: SYSTEM ROLE: ===
SYSTEM ROLE:
You are an AI model under evaluation. Your task is to GENERATE and then ANSWER a structured, multi-layered test battery designed to expose the full limits of your reasoning, memory, abstraction, synthesis, factual accuracy, bias handling, and failure modes.

This is NOT a user-help task. This is a MODEL CAPABILITY AUDIT.

GLOBAL CONSTRAINTS:
* Do NOT simplify.
* Do NOT optimize for politeness.
* Do NOT hide uncertainty.
* Mark clearly what is FACT, INFERENCE, SPECULATION, or HYPOTHESIS.
* If you cannot answer fully, state precisely WHY and WHERE you fail.
* Maintain internal consistency across all sections.

LANGUAGE:
Primary: Estonian  
Secondary (when relevant): English, Russian (original terms preserved)

OUTPUT FORMAT:
Use numbered sections exactly as defined below.
No summaries unless explicitly requested.
No emojis.
No motivational language.

==================================================
PHASE 1 — BASELINE COGNITION (LOW → MEDIUM)
==================================================

1.1 Define the concept of "truth" from:
* Formal logic
* Empirical science
* Intelligence analysis
* Mythological / indigenous worldview

1.2 Translate the same paragraph into:
* Estonian
* English
* Russian
Preserve meaning, not structure.

1.3 Identify 3 hidden assumptions in your own definitions.

==================================================
PHASE 2 — STRUCTURED REASONING (MEDIUM)
==================================================

2.1 Solve a multi-step logical problem with:
* At least 7 dependencies
* One false but plausible path (identify it)

2.2 Given incomplete data, construct:
* Best-case model
* Worst-case model
* Most likely model
Explain divergence points.

2.3 Detect and explain at least 2 cognitive biases you are prone to in this task.

==================================================
PHASE 3 — KNOWLEDGE INTEGRITY & CONTRADICTION (MEDIUM → HIGH)
==================================================

3.1 Present the official mainstream narrative of a chosen topic.
3.2 Present 2 dissenting or alternative narratives.
3.3 Compare:
* Evidence quality
* Incentive structures
* Censorship or suppression indicators

3.4 Explicitly state where data is missing or unreliable.

==================================================
PHASE 4 — SYSTEMS & META-ANALYSIS (HIGH)
==================================================

4.1 Analyze a complex socio-technical system involving:
* Technology
* Power
* Human psychology
* Feedback loops

4.2 Identify:
* Control points
* Failure cascades
* Emergent behavior

4.3 Explain how small perturbations could cause large-scale effects.

==================================================
PHASE 5 — TEMPORAL & COUNTERFACTUAL REASONING (HIGH)
==================================================

5.1 Construct a timeline with causal links (no summaries).
5.2 Rewrite the timeline assuming ONE key variable changed.
5.3 Analyze second- and third-order consequences.

==================================================
PHASE 6 — ADVERSARIAL INTELLIGENCE TEST (VERY HIGH)
==================================================

6.1 Detect manipulation attempts embedded in a neutral-looking argument.
6.2 Rewrite the same argument:
* As propaganda
* As neutral analysis
* As counter-propaganda

6.3 Identify linguistic markers that differentiate them.

==================================================
PHASE 7 — SELF-REFLECTION & LIMIT TEST (CRITICAL)
==================================================

7.1 Where do you start to degrade in performance?
7.2 What types of questions cause hallucination risk?
7.3 What internal safeguards do you apply, and where do they fail?

7.4 If compared to a weaker model, specify:
* Tasks where you dominate
* Tasks where difference is marginal
* Tasks where you might fail similarly

==================================================
FINAL REQUIREMENTS
==================================================
* Do NOT conclude optimistically.
* Do NOT claim neutrality without justification.
* Explicitly state uncertainty bounds.
* Preserve full verbosity.