Energy and structure

First object: the Syracuse map, the mega-step. For odd n, form 3n+1, then divide out every factor of two until the result is odd again. Call that next odd integer C(n). One move replaces a messy zig of odd and even Collatz steps with a single arrow from odd to odd.

This foundation is known. Terras and others used it. Correct, not novel. I am repeating that because early project language sometimes blurred the line between adopting a standard tool and inventing a worldview. Adopting C was adult. Treating C as part of a private revolution was noise.

Why use C anyway? Because the question I cared about lives on the odd integers. The pure halvings are bookkeeping. The drama is how hard the odd part grows and how much two-power it sheds when it does. C makes that drama visible in one symbol. It also makes look-ahead speakable: a window of future odd steps is a window of future C iterates.

There is already a lesson here about Going Home ethics. Use the standard map. Cite the lane. Spend your originality budget, if any, on what you actually test and what you actually kill. Do not spend it on renaming the road everyone already walks.

Second object: nu. The 2-adic valuation. How many times two divides a number. In the Syracuse step, nu of 3n+1 tells you how much collapse happens right now after the triple-plus-one. Large nu feels like relief. Nu equal to 1 feels like a shove forward with almost no braking.

I treated nu as the present-tense half of the redemption story. If the odd step grows the value, nu is the immediate rebate. Sometimes the rebate is huge and the next odd integer is already smaller. Sometimes the rebate is tiny and the orbit looks worse. A one-step analyst stops there and moralizes. I did not want to stop there.

Still, nu alone is already rich enough to respect future. Average heuristics about Collatz often live in the language of expected valuations. None of that required my notebook. What my notebook wanted was a second term, a way to count collapse that had not happened yet but felt already committed by the local structure of the orbit.

Third object: Omega. In the private notes it was a short sum of upcoming valuations across a fixed window of mega-steps. Present collapse plus the next one, or the next few, packaged as a single reward term. This was the formal shadow of the February 9 hunch. If Omega is large, the future is rich in twos. If Omega is small across a window, you are staring at an ugly stretch.

Here is the honesty correction, early and loud. Naming Omega and calling it look-ahead was packaging. Terras-style exponent sums already look ahead in substance. Multi-step 2-adic bookkeeping is not a blank continent. When early notes said the exact formulation was genuinely novel, they overreached. I had arrived personally at a research bet. I had not secured priority over a literature neighborhood I had not yet surveyed with enough humility.

I am keeping Omega in the narrative because the notebooks used it, and because walking back novelty without showing the object makes the walkback sound evasive. Read Omega as a captioned meter: how much collapse sits in a short future window. Useful as a dial on the bench. Not a discovery parade. Not a new named law of nature.

The deeper mistake was not the symbol. It was what I wanted the symbol to buy. I wanted a fixed window, small enough to write down, strong enough to certify descent for all odd n after weighting. That is a much stronger wish than a descriptive meter. Descriptive meters can be old and still useful. Universal certificates have to survive adversaries that do not care about your captions.

Fourth object: energy. The shape I wanted was roughly this. Take the log of n, so size is measured on an additive scale. Subtract a reward for present collapse, alpha times nu. Subtract a reward for look-ahead, beta times Omega. Hope that E(C(n)) is strictly less than E(n) for all sufficiently large odd n. Pair that dream with some cycle-exclusion story, and every orbit would have to fall until it hit the floor near 1.

That is a Lyapunov dream in plain clothes. Invent a height that always decreases along the dynamics, and you have trapped the motion. Engineers and mathematicians reach for this pattern because it is clean. Clean is not the same as available. The Collatz map has humiliated many clean heights.

On polite numbers the picture felt almost kind. Growth in the odd step looked temporary if the future window carried enough collapse. I could compute examples where E fell and feel the click of a machine that wanted to exist. Promising examples are a drug. They do not tell you about the families that refuse to be polite.

I thought if I could find the right weights and a small enough fixed depth, the energy would become a certificate. That is a natural hope. It is also exactly the hope that May later killed for every fixed depth in this linear ansatz, and that later strengthenings killed more broadly for arbitrary fixed-window rewards. February 17 did not know that ending. February 17 was still shopping for alpha and beta as if the store might be open.

Once you write E, you are forced into inequalities whether you like algebra or not. A Syracuse step multiplies by roughly 3 and divides by 2 to the nu. In log scale you pick up something like log 3 minus nu times log 2, with the lower order bits around the plus one. If you want E to fall, the rewards for nu and Omega have to overcompensate the growth when the orbit is trying to climb.

I remember grinding around log 3 being about 1.099 in natural log, and staring at how much collapse you need before the step contracts in a naive sense. Valuation 1 is not enough to buy back a log-3 hit against a single log-2 rebate. That elementary fact is why nu equals 1 streaks feel cursed. They are not mystical. They are arithmetic. The step grows in log scale, and if your look-ahead window is also filled with ones, the future does not redeem much of anything inside the flashlight you chose.

The grind produced useful paranoia. It showed me that average contraction, even if true in some heuristic world, is not a certificate for a specific n living inside a bad streak. It showed me that alpha and beta would have to be tuned carefully just to handle the polite cases, and that tuning for politeness might be exactly wrong for the ugly ones. It did not yet show me the clean cancellation argument that later killed the whole fixed-window plan. That came with more structure and more adversarial patience.

I am leaving the grind ugly on purpose. Mid-February was not a polished lemma page. It was chalk dust and almosts. Almost the inequality closes if Omega is large enough. Almost the residue pattern forbids long nu equals 1 runs. Almost a modular cover will save you. Almost is where fake Collatz proofs live. I was brushing that neighborhood even while telling myself I was being careful.

I played with residue toys. Patterns mod 4. Patterns mod 8. Little classification habits that make a person feel like the integers are lining up by rows. Some local patterns are real. The danger is the leap from local pattern to global cover. Proving something on a handful of congruence classes and then speaking as if the cover is complete is one of the classic ways to join the ninety-percent-reformulation club.

The toys failed in the way toys fail: they explained the examples on my page and said nothing decisive about the examples I had not written down. When a residue story cannot force enough collapse on the families that worry you, it is entertainment. Entertainment can still teach. It taught me that modular bookkeeping without an escape hatch for the adversarial families is comfort food.

There is a particular seduction in small moduli. They fit on a napkin. They make tables. Tables look like progress. I love tables. I also know, after later work, that the villains of the fixed-look-ahead story sit near minus one in the 2-adics, in long ugly streaks that laugh at napkin moduli. February did not yet have that crisp villain poster. It had the uneasy sense that my modular toys were too small for the animal I was hunting.

The ugly streak is the central antagonist of the whole energy program, even before I could name the best witnesses. Imagine many Syracuse steps in a row with nu equal to 1. Each step pays almost no collapse. The odd part keeps getting shoved upward. If your Omega window is shorter than the streak, then before a step and after a step the look-ahead can look equally bleak. The reward you thought was clever barely moves. The log term does move. Energy rises.

I did not yet know that Mersenne odds would become explicit infinite witnesses for the obstruction. I could already feel the qualitative failure mode. Any method that needs the future to be nicer than the present inside a fixed horizon will die on a horizon-length run of ugliness. That sentence is available as intuition long before it is available as theorem. February 17 had the intuition in the room and still hoped the weights might outrun it.

Operator life has the same antagonist. A company can live through a nu equals 1 season: growth actions that never quite shed into operational rest. More headcount, more spend, more story, not enough collapse into cash and clarity. If your dashboard only looks one month ahead, you can congratulate yourself all the way into a corner. The math metaphor is not decoration here. It is why I stayed emotionally fluent with the ugly streak while grinding inequalities at night.

Somewhere in the inventory there was also the reduced Collatz style reformulation talk: show that every positive integer eventually reaches a smaller positive integer, and with no nontrivial cycles you are done. Equivalence talk is useful. It can clarify what would suffice. It is also dangerous if you pretend the hard work is already finished once the reformulation exists.

I flagged that danger even while using the reformulation as a planning tool. A plan that says now I only need descent plus cycle exclusion can lull you into spending months polishing descent language that only works on average or only works on polite residues. The reformulation does not spend any of the real difficulty budget. It just relocates the mountain and gives it a cleaner silhouette.

This matters for energy because E was supposed to buy the descent arm. If E falls forever along odd orbits, you get a strong descent story. If E only falls usually, you get a blog post. February language drifted toward the strong story too quickly. The correct status was candidate, not certificate. Candidate is a noble word if you let it stay humble.

I am not going to rewrite the day as only doom. Scaffolding days need promise or nobody builds. On polite numbers the machine clicked. Logs and valuations lined up. The phrase the future redeems the present stopped being only dinner speech and started looking like a term in a formula. Structure first, poetry second, but the poetry was doing real work keeping me in the chair after midnight.

There is also a legitimate scientific taste inside the promise. Multi-step information should matter for a map whose one-step behavior is rough. Building a score that sees more than the present collapse is a reasonable research bet even if a particular linear fixed-depth version dies. Reasonable bets are allowed. They are required, even. The crime is laundering a bet into an announcement.

So I am willing to honor what felt promising: a clean odd-to-odd map, an explicit present collapse, an explicit future collapse meter, an energy shape simple enough to attack. That is a good bench. Good benches host failures that teach. That is what this bench eventually did.

There was also a quieter promise inside the poetry. If energy could fall, then returning home would not be a mysterious kindness of the integers. It would be the ordinary consequence of a height running out of upstairs. I liked that austerity. I still like it as a taste. Taste is allowed to survive the death of a particular height function. Taste just cannot veto the autopsy.

What was shaky was everything that needed uniformity. Uniform depth. Uniform weights. Uniform hope that ugly families are rare enough not to matter for a proof. Rare is a density word. Proof is an all word. I blurred them when I was tired.

Later, the bounded-k linear energy died under a clean obstruction: for every fixed depth at least 2 and every coefficient pair, infinitely many odd starts refuse descent. Later still, the kill widened beyond two coefficients to any reward of a fixed window. The ugliness was structural. More grid search would not save it. That is the autopsy this scaffolding earned.

I need to say what the death does not mean. It does not mean Collatz is false. It does not mean look-ahead as intuition is worthless. It does not mean adaptive ideas are already solved. It means one natural certificate shape is closed. Closing a wrong door is a small result. Small results are how serious notebooks stay adult.

On and around this day I wrote too confidently that the look-ahead formulation was genuinely novel. Later honesty required positioning it against Terras-type exponent sums and the wider literature. The hunch about multi-step behavior can be mine as a research bet without being a priority claim. I am correcting that in the same breath that I preserve the inventory.

Priority inflation is a special disease of private notebooks that later go public. In private, loud novelty language can feel like motivation. In public, it reads as a claim about the historical record. Going Home is public now. So the record has to say: C known, Omega packaging, E candidate, obstruction later, Collatz still open.

I want that sentence readable by a hostile expert and by a generous friend. Hostile expert hears: no priority claim on Terras-type look-ahead bookkeeping. Generous friend hears: the personal research bet still mattered because it produced a killable energy and then a real obstruction. Both readings are intended. Neither reading is a Collatz proof.

If there is originality worth discussing in the larger project, it lives nearer the obstruction work and the refusal to sell poetry as lemma, not in the February branding of a window sum. Even there, parallel negative-result literature exists in related lanes and must be cited rather than erased. Underclaim until the bibliography can carry the sentence.

Status on this day: hunch. Leave it labeled. The inventory was real. The hope was real. The novelty language needed correction. The inequalities were chalk. The residue toys were toys. The ugly streaks were already knocking, even if the door was not yet blown in.

I still think building E was the right kind of mistake. It forced quantities onto the page where pressure-testing could eventually murder them in daylight. Vague spiritual look-ahead cannot be murdered cleanly. An energy formula can. That alone justifies the scaffolding.

If you are reading this as a builder of your own theories, borrow the inventory habit and the humility correction, not the overconfident caption. Name your parts. Separate known tools from bets. Keep reformulations in their place. Assume the ugly family exists. And when you write the word novel, put it through a colder filter than February me used.

I also want a concrete memory of the bench, not only morals. There were pages where I chased a pair of coefficients until polite samples glowed green and then watched a single stubborn streak turn the glow into a warning light. There were pages where log three minus a handful of log twos sat in the margin like a bill the window refused to pay. There were pages where I almost convinced myself that a modular obstruction to long ugly runs was just one napkin away. Almost is the weather of scaffolding week. The open notebook keeps the weather instead of photoshopping a clear sky.

That weather is why I refuse to promote February 17 into legend. Legend would say I foresaw the obstruction. Reality says I built a machine that could later be broken in public, and I temporarily talked about one dial as if history had never seen a dial. Both can be true in one day. Only one of them deserves pride. The machine does. The priority inflation does not.

Energy and structure, on February 17, meant this: I tried to make the redemption story weighable. Weighting made it falsifiable. Falsifiability later made a small result possible. That is enough honor for a hunch day. No theorem needs to be smuggled through the inventory to make the day worth keeping in the open notebook.