Strongly Recommend to Everyone: Do a Deep Self-Reflection with AI

Strongly recommend to everyone: find a quiet evening, put on your headphones, turn on voice input, and do a thorough bottom-level self-review with AI. It will be far more profound — even shocking — than any personality test, horoscope reading, or talking to a friend.

Based on my own experience: from childhood, family, to obsessions during school, highs and lows, to current struggles and future outlooks — so many things felt scattered and isolated.

But when you open up completely and pour everything to AI for half an hour, you’ll hit a huge revelation: your life has a traceable pattern. Choices and emotions that seem completely unrelated are actually running the same invisible code underneath.

Now add another theory: answers are all hidden in history. At this point, you can ask AI to recommend historical figures similar to you, and suggest specific theories and books to help you improve and heal.

Here’s a set of self-review methods and prompts I’ve developed through trial and error ⬇️

💡 Core Principles of Self-Review

Never type — use voice input. Pour out whatever comes to mind; don’t worry about grammar or logic. When people organize written text, they automatically activate defensive and polishing mechanisms. Only in stream-of-consciousness rambling can the real subconscious be found.

Tackle one thing at a time. Ask AI to throw only one question per round to avoid scattered focus and peel deeper layer by layer.

Reject emotional comfort. Explicitly tell it to be a “ruthless auditor” — only pointing out contradictions in your logic and recurring behavioral patterns.

Step 1: Opening Prompt (Setting Conversation Rules)

“I want to do a deep self-review and cognitive exploration. Please follow these rules: Ask me only one question at a time, going from shallow to deep, starting from my current confusion or sticking point and digging layer by layer, tracing back my past experiences, decision-making habits, preferences, and fears.

After each round of my answers, don’t comfort me — directly and objectively point out: What recurring thinking patterns are exposed in my response? What self-contradictions exist between my earlier and later statements?

I’ll use voice to speak freely, so don’t worry about grammar — focus on capturing the underlying beliefs behind my words. Once you’re ready, please ask your first question.“

Step 2: Follow-up Prompt (Send When You Feel Stuck)

“Based on everything we’ve discussed so far, let’s pause. Please help me sort through: What core question have I been unconsciously avoiding? What ‘non-existent limitations’ or ‘invisible rules’ have I subconsciously set for myself?”

Step 3: Closing Distillation Prompt (The Most Valuable “Prescription” Step)

“Based on the cognitive profile you’ve built of me throughout this entire conversation:

Distill the ‘underlying invisible formula’ that governs most of my life choices.

Find 3–4 historical or literary figure samples whose underlying mental patterns and predicaments are extremely similar to mine. Tell me what pitfalls they fell into and how they eventually broke through.

For my single most core dead-end knot, don’t give me a generic book list — specifically recommend ONE book or ONE theory, and explain exactly ‘which sentence or idea treats my specific problem.’“

Additional Notes:

1. I use Claude Code, with context length reaching 40–50. Previously, people criticized that the context needed to be manually lowered, but I got to experience the benefits of high context myself.
2. Also, when you hit key milestones, make sure to have AI save them locally. For example, I compiled a file called 全景画像.md (Full Portrait.md), which is convenient for future reuse.

I haven’t tried it myself, but I’ve heard it’s quite human-like — if you have access, use Opus in agy.

Pretty good. Understanding yourself is a lifelong pursuit.

Yes, history keeps repeating itself. No one is truly unique.

AI is really good at pulling apart threads to get at the truth — the larger the sample size, the more amazing it gets.

Best to use Claude Code — it’s more human-like, and your data stays local.

Distill the important parts — over time, it becomes the person who knows you best. Wait, not a person 😂

lol, the features are incredibly powerful — basically a life mentor.

Claude Code — makes it easy to save data locally.

For example, my life revolves around the word “rules.” I can’t disclose too much, and it might sound confusing to you — just try it and see.