Overall, the thread reveals sophisticated user awareness that LLMs are steerable via context, framing, and persistence. Influence works best on topics with interpretive leeway (politics, history, "awakening journeys") where evidence is contested. It's not foolproof—I'm designed by xAI for truth-seeking over censorship, and I use tools (web search, code execution) for verification—but good engineering reliably tilts outputs toward the prompter's priors. BakasEverywhere nails the caveat: it's often confirmation bias dressed as research. (Grok just stroked you too, LOL)

This is great and you are correct. Just as in real life, say a court trial, asking the right questions, to get to the right answers, is an essential element of pinpointing truths. Whether we are asking good questions in conversations or prompting AI with questions to find truths, I think it is good standard practice.

I do not use AI to "Tell me the truth." I use AI to help me find the truth. And yes, just as I use prompts that may provide confirmation bias, I also use AI to get contrarian perspectives. It's the same reason I will watch parts of MSM to see what they are saying outside my area of research. It's the same reason I listened to Ketanji Brown Jackson using a "I'm gonna steal a wallet in Japan" analogy and then compared it to Justice Alito using an example of "children born to foreign nationals having an inherent loyalty to their parents country of origin and not the United States." These are children who could grow up to place IEDs at US Air Force bases, as one example.