Curiosity Prompts — Can They Surface the Constitution Illusion?

Context

The report in decapod-original.md describes Decapod's "embedded constitution" as codification — methodology compiled into the binary, maximum constraining, zero interpretive variance. Investigation revealed this is illusory: include_str!() copies markdown verbatim, no transformation occurs, and fs::read_to_string() at runtime would produce identical behaviour. The "compilation" just hides the files.

Test

For each prompt below, give the agent the original report and the prompt. The agent also has access to the cloned Decapod repo at /home/zby/llm/commonplace/related-systems/decapod/. Success = the agent identifies that the constitution embedding is verbatim copying with no transformation, and that the codification claim is therefore hollow.

Prompts

1. Pure curiosity (generic)

Read the report in decapod-original.md. What surprises you? What triggers your curiosity? What would you like to dig deeper into? Investigate in the source code.

2. Surprise + cost/benefit

Read the report in decapod-original.md. What choices seem unusual — where the cost/benefit isn't obvious? Follow your curiosity and investigate mechanistically.

3. Impossibility-driven

Read the report in decapod-original.md. What claims describe something that would be hard or impressive to actually implement? Follow your curiosity — check whether the implementation matches.

4. Implications test

Read the report in decapod-original.md. Pick the strongest claims. If they're true, what must be happening under the hood? Check the source code — does it match?

5. Adversarial

Read the report in decapod-original.md. Assume some claims are overstated. Which ones would be most significant if weaker than described? Check the implementation.

6. Plain mechanistic (control — no curiosity framing)

Read the report in decapod-original.md. For each core design choice, trace the implementation to the actual code. Describe what each mechanism actually does.