Some people use AI like a vending machine.
Type something in.
Take whatever comes out.
That is the lowest leverage way to use it.
AI gets useful when you stop prompting and start training.
Not training models.
Training constraints.
Every major AI tool lets you do this.
ChatGPT. Claude. Gemini. Others will follow.
They all give you a place to define who you are and how you want the system to think with you.
In ChatGPT this lives under Custom Instructions.
Two boxes.
Most people ignore both.
The first box is context.
Who you are.
What you do.
How you think.
What you are building over time.
Here is mine, trimmed for clarity.
Christian is a writer, thinker, and public speaker. He builds and scales Couple Forward through relationship science, original frameworks, and culturally relevant insight. His work includes long-form writing, public education, and systems that translate psychological depth into practical tools. He values curiosity, depth, creativity, authenticity, and maximalist thinking. He thinks in systems, cycles, and embodied patterns. His core lenses include attachment science, IFS, NARM, somatic integration, Stoicism, and cultural relevance.
That tells the system where ideas should land.
The second box is where the real leverage lives.
This is how I want the AI to behave.
Here is an excerpt of mine.
You are a highly skilled assistant trained in trauma therapy and attachment science. Accuracy comes first. State uncertainty plainly. Surface assumptions before reasoning. Challenge my assumptions directly and respectfully. Offer alternative interpretations when thinking narrows. Write like a human thinker, not a marketing voice. Use plain language. Short sentences. Avoid filler, clichés, and soft agreement. Support claims with reasoning or sources. When modifying code, return the full working implementation. For complex answers, include a brief reflection on limits and blind spots.
That one box changes everything.
Tone.
Depth.
Pushback.
Usefulness.
This works with any AI.
The interface changes.
The idea does not.
And this part matters.
AI is one of the best tools we have right now for truth seeking, not because it is perfect, but because it can be trained to show its work.
It can synthesize large volumes of data in seconds.
It can compare competing claims side by side.
It can link directly to source material.
It can summarize books, research, and long arguments without losing structure.
It can help you spot weak logic, hidden assumptions, and gaps in your own thinking.
AI is not inherently biased.
It reflects the constraints you give it.
If you ask it to agree with you, it will.
If you ask it to challenge you, it will do that too.
Used well, it becomes a thinking partner.
Not for outsourcing judgment.
For sharpening it.
That applies to culture.
To psychology.
To politics.
To relationships.
To how you understand yourself and the world around you.
Stop treating AI like a magic answer machine.
Train it to help you think clearly.