Skip to main content
AI Strategy

AI Agents — A Ghajini Story

👨‍💻
Shivaling Sannalli
Quality Engineering Lead
March 31, 2026
3 min read
AI AgentsContextMemoryLLM
"What was I doing yesterday?"

Do you wake up like this sometimes? If your memory is strong, you will remember Sanjay Singhania from Ghajini — and the original Tamil film of the same name.

If it is short-term memory we are talking about… well, can't help it. 🙂

🎬 Imagine This

Sanjay Singhania is now an AI agent

Instructions written everywhere
Notes, tattoos, photographs — the context is all externalized.
A clear goal
He knows exactly what he is trying to accomplish.
The ability to execute tasks perfectly
He wakes up, reads the instructions, and starts executing. Everything looks smooth.

⚙️ Until Things Change…

Midway through the run

Someone edits the plan

  • Someone changes the order of the tasks
  • Someone adds a new instruction
  • Someone removes a step

Now what happens? He pauses. He gets confused. He re-reads partial context. He forgets what happened before. And suddenly the whole flow breaks.

🧩 That's Your AI Agent

Exactly the same pattern

The behaviour you are actually seeing

  • Works great with a structured flow
  • Struggles when instructions change mid-run
  • Depends heavily on the context you provide
  • Has hard limits on how much it can "remember"
🧠

It is not thinking continuously

It is reconstructing context every single time. That single sentence explains most of the agent behaviour that surprises teams in production.

😂 The Funny (But Real) Moment

When the notes contradict each other

Conflicting instructions

Sanjay writes: "Find the target. Don't trust anyone." The next note says: "This person will help you." Now he meets that person — and he will either trust the wrong person or fight the right one. That is your AI agent when context conflicts, instructions are inconsistent, and memory is incomplete.

Conclusion

"Before I forget…"

AI agents are powerful. But today they are not continuous thinkers. They are context-driven, memory-limited, and instruction-dependent.

Until memory and reasoning truly evolve, they will keep waking up every time asking: "What was I doing?"

About the Author

👨‍💻
Shivaling Sannalli
Quality Engineering Lead

Expert in quality engineering, test automation, and DevOps practices

View Medium Profile
Working on this?

We do this for a living.

If this article describes a decision you are making right now, bring it to a Blueprint Session — a working session, not a sales call.