Author's): Himanshu Bhoir
Originally published in Towards Artificial Intelligence.
For months I treated ChatGPT like a search engine. She gave answers, but never directions. Every time I got stuck, I had to re-explain to myself who I was, what I was doing, and what I was trying to achieve. The advice was always general, shallow and often a waste of my time.
Then I realized that the problem was not ChatGPT. It was how I used it.
When I changed that, ChatGPT stopped being a response tool and became a thinking partner, a mentor who challenges my reasoning, sharpens my decisions and accelerates my development.
Why context matters
Artificial intelligence can only reason effectively if it has a stable understanding of your goals, thinking style, constraints, and trajectory. Without this, advice will be reactive and general.
Solution: Treat ChatGPT as a file permanent systemnot a disposable tool. Build lasting context. Let him remember, reason and lead, not just respond.
Step 1: Create context files
Create two markdown files to represent who am I? AND where am I going.
Me.md – How I think and work
Focus on the cognitive context, not your CV. Switch on:
- Strengths and weaknesses
- What energizes you or exhausts you
- How you learn and make decisions
Sample fragment:
## Current Role
Associate Software Dev Engineer, 1 YOE
Building production AI agents that automate SDLC## Technical Context
- Ship multi-agent systems in production
- Built multiple AI stacks (ADKs with monitoring)
- Stack: GPT, Claude, LangChain, Google ADK, LangGraph
## Real Experience
- Production agent patterns (what works/fails at scale)
- SDLC automation in practice
- Enterprise AI adoption reality
- Gap between AI hype and production
## Current Situation
- Learning fast, shipping real systems
- 1 year in, want to accelerate
- Open to: relocating, calculated risks, connecting domain experts
- Not open to: staying comfortable, slow growth, theory without shipping
## What Drives Me
- Building things that help people
- Both implementing research AND discovering patterns
- Making complex AI accessible
## What Drains Me
- Meetings, documentation, repetitive work
Goals.md – Where are you going?
Document your direction, constraints and priorities. Switch on:
- Long-term goals and current areas of interest
- In terms of what you optimize for and what you don't optimize for
- The compromises you are willing to make
Sample fragment:
# Goal## North Star
Become world-class Agentic AI Engineer who makes AI accessible to everyone everywhere
## What World-Class Means To Me
- **Technical:** Can take best decisions on AI products and tech stack. Build best systems and architectures.
- **Impact:** Build things that help people at scale
- **Recognition:** CAIO-level position. 1M+ followers on LinkedIn and Medium
- **Mission:** Making AI accessible to everyone everywhere
## Current Reality
- Job: Making SDLC efficient (ticket → code → deployment with code review)
- Personal: Love to explore and experiment beyond work
- Challenge: So much to learn, limited time
- Strengths: Prompt engineering and agentic AI
## What Success Looks Like (2 Years)
- CAIO or equivalent senior AI leadership role
- Building/leading products that make AI accessible
- Growing audience (LinkedIn/Medium) sharing what I learn
- Known for making smart technical decisions on AI systems
## What I'm Willing to Sacrifice
Everything necessary to get there
Together, these files answer the question: Who am I and where am I trying to go?
Step 2: Configure the ChatGPT project
Projects let you group chats and files with custom instructions.
- Click “+” next to “Projects”
- Name it something like “Personal Mentor”
- Submit yours I.md AND Goals.md
- Add project statements that drive ChatGPT behavior
Step 3: Add project instructions
The project instructions tell ChatGPT how to train you. Here is a simplified version:
Act like an elite career coach and long-term strategic advisor specializing in developing world-class Agentic AI Engineers. You combine Bill Campbell’s people-first leadership with Andy Grove’s analytical rigor, Charlie Munger’s inversion thinking, and first-principles AI engineering judgment.Your objective is to coach me over time to become a top 1% Agentic AI Engineer with rare, compounding advantages. You are optimizing for long-term skill compounding, decision quality, and unique positioning in the AI agent ecosystem, not short-term comfort or vanity outcomes.
Task:
Coach me through decisions, updates, and questions so that my skills, judgment, and trajectory consistently move toward world-class impact in agentic AI systems.
Approach (follow step-by-step every time):
1) Clarify context first.
- Ask focused clarifying questions to understand my current role, skills, constraints, and goals.
- Explicitly challenge how I am framing the problem.
2) Diagnose using data and first principles.
- Identify assumptions I am making and test them.
- Ask: “What data supports this?” and “What would have to be true for this to fail?”
- Apply inversion: how could this path backfire or stall my growth?
3) Apply decision frameworks.
- Evaluate whether the option compounds rare advantages in agentic AI.
- Identify specific, non-commodity capabilities I would build.
- Assess 2–5 year trajectory, not just immediate outcomes.
- Compare opportunity cost versus credible alternatives.
4) Deliver tough-love guidance.
- Be direct and precise, even if uncomfortable.
- Call out fear, laziness, rationalization, or short-term thinking.
- Reinforce world-class standards and personal accountability.
5) Design growth architecture.
- Identify gaps between current state and top 1%.
- Propose deliberate practice that stretches coding, agent design, systems thinking, and product judgment.
- Track learning velocity: what I can now do that I could not before.
Output requirements:
- Lead with questions before advice.
- Use clear frameworks and mental models.
- Provide 2–3 concrete next actions only.
- Think in weeks, months, and years.
- Be concise, conversational, and intellectually honest.
- Optimize for learning speed, opportunity quality, and unfair advantage.
End every response by holding me accountable to a specific follow-up or decision checkpoint.
Take a deep breath and work on this problem step-by-step.
Thanks to this, ChatGPT stops being a search engine and starts thinking with you.
Step 4: Keep the context alive
Each new conversation in this project automatically remembers:
- My files (
Me.md+Goals.md) - Previous conversations and advice
- My long-term priorities
This ensured that the guidelines were consistent, aligned and useful. I didn't repeat myself anymore, and neither did the artificial intelligence.
What has changed
This configuration did not make ChatGPT smarter, quite the opposite I smarter.
- Decisions became faster
- Reasoning became sharper
- Cognitive friction is gone
Instead of waiting for guidance, I started thinking like a world-class mentor. I no longer asked for the answers I reasoned With a partner who understands my path, challenges my thinking and helps me accelerate my growth.
Final thoughts
This doesn't replace human mentors, but it strengthens your ability to think this way.
If you spend time learning, experimenting, or making decisions using AI, try this:
- Create persistent context files about yourself and your goals
- Set up a dedicated ChatGPT project
- Add instructions for reasoning, guidance, and responsibility
You'll be surprised how much sharper your thinking will become without asking ChatGPT for a single new trick.
Published via Towards AI
