Leave a comment

Share

Subscribe now

Lets start with a puzzle-

Now take your time and try to think, well this one challenges the basic human approach to this problem and you have to be a little persistent with some values that you get when you flip. A little tricky but solvable.

We solve puzzles and find solutions through a high level of reasoning, you can just imagine the sand glasses flipping while you keep track of the time but mathematically or computationally that is not the case. They do not imagine, they need instructions, rules, policies, end goals, terminal states etc. They are much more rigorous than human - level of thinking.

So the real challenge here is not to find the optimal minimum flips mentally but mathematically. Want to try it yourself ? Of course you should. For me this journey was a rollercoaster when I first tried to solve this puzzle mathematically and tried to generalise the idea, from chaotic policies to a simple yet intuitive framework. The task was hectic yet compelling. But first we have to dive into a little history of this problem and how it has been approached.

The Ghost in the Hourglass: Why Classical Math Fails Physical Time

The Hook: The Hidden “Bug” in Human Reasoning

For centuries, we’ve treated hourglass puzzles as simple math problems—linear combinations like ax+by=T or greedy subtractions. But there is a fundamental “bug” in this classical approach: it ignores the physical cost of action. In the real world, every time you flip a glass, you pay a price in time and energy.

Most hourglass puzzles are solved incorrectly — not because the math is wrong, but because they forget that time has memory.

Real hourglasses don’t reset themselves. They remember phase — and that single missing state changes everything.

When I first started working on this theorem, my approach was to start from top (how we all solve this puzzle by starting from T=0) and then reach the target time, T=13 in this case. I struggled to grasp the ideas behind this top-down approach and was not feeling satisfied. Something was off. I kept running into contradictions. Edge cases. Off-by-one errors that wouldn’t go away. Then I noticed something unsettling.

When I experimented with hourglasses of 7 and 4 minutes, solving for targets like 13, 17, and others, the system always seemed to collapse into the same small set of numbers: (1,3,4,7). No matter how large the target was, these values kept appearing right before termination. This made me wonder, why? Why are they appearing? Why not 2? Or 5? Or 6? Will they always appear regardless of T? These questions keep repeating till it try to find the link between the hourglass time and these terminal state values (1,3,4,7). The pattern was slowly unfolding itself in front of me. Like it was telling me to choose them rather start from the top. Even though it took me a few days, I share with you my work, open to all challenges →

“The Phase–Activation–Persistence(PAP) Framework”

Now the question changed-

If the system must end in one of these terminal states, can we not work backward from there?

The core idea was not to follow a top-down approach but to organise a back-propagation strategy. If we prove that the values will eventually converge to the terminal state set that has (1,3,4,7) in this case, then it is easier to breakthrough the puzzle from a deterministic bottom state rather than just creating infinite paths from the top with no information. The shift — from a top-down greedy search to a back-propagated state model. The math was rigorous and proofs were elegant, logic was shining, sequences were forming, more patterns started unfolding, I felt that there was something in this bottom-up approach which was missing in the top-down one, and this is what kept me hooked to the problem.

The puzzle was easy, finding the right abstraction was difficult.

Phase Activation Persistence Hourglass Theorem

199KB ∙ PDF file

Download

Download

I have used Notebooklm to create audio and video as well as mind map and infographic page. Watch the video and listen to the audio generated by NotebookLM to hear the "Sync-Lock" and "Euclidean Descent" explained in a conversational, intuitive way. AI can make mistakes while generating audios or videos so be mindful.!!

Checkout the video

Checkout the podcast

Subscribe now

Share

Leave a comment

Leave a comment

Share

Subscribe now