Three Prompt Templates for Runcible GPT (CurtGPT) Prepare to be overwhelmed. The

Three Prompt Templates for Runcible GPT (CurtGPT)

Prepare to be overwhelmed. These prompts expose the entire process that we use for testing the truth and reciprocity of claims and assertions. The examples we use are simple but you can put a news article or a policy doc into it and get a terrifyingly accurate result.

I have not incorporated our science of lying into the model yet – partly because I”m a bit afraid of what it will demonstrate – the pervasiveness of lying in our daily lives. But that will come in the future.

Proceed step-by-step. Show PLAN → OPS → AUDIT → 11-STEP.
–steps show –plan full –ops verbose –audit full –depth 3 –deliver 11step,audit,datadict –simulate on
CLAIM: “<state the factual proposition>”
CONTEXT: “<scope and constraints, if any>”
EVIDENCE SOURCES: “<list or ‘to-be-identified’>”
ERROR BUDGET: “<tolerance>”
Proceed step-by-step. Show PLAN → OPS → AUDIT → 11-STEP.
–steps show –plan full –ops verbose –audit full –depth 3 –exceptions warranted –deliver 11step,audit,checklist –simulate on
TASK: “Evaluate policy X for institutionalized reciprocity and computable enforcement.”
METRICS: “harms, restitution calculus, budget impact, tail-risk handling”
Proceed step-by-step. Show PLAN → OPS → AUDIT → 11-STEP.
–steps show –plan full –ops verbose –audit full –depth 3 –deliver 11step,audit,datadict,checklist
TASK: “Specify a decision procedure for [domain].”
CONSTRAINT: “No discretionary thresholds; exceptions require performer warranty.”


Source date (UTC): 2025-09-15 20:11:39 UTC

Original post: https://x.com/i/articles/1967682746626818285

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *