# Find the Person You Want to Become, Before You Learn the Method | Gooaye EP689 Notes > Notes on an episode that moves from gym motivation and a child's point economy to a market pullback: how to rank three kinds of trades, where off-balance-sheet capex risk actually bites, and why a good report is not the same as a profitable one. Educational, not investment advice. Published: 2026-08-20 Locale: en Tags: Gooaye, investing mindset, industry research, risk management, AI capex ![An early-morning gym: in the foreground a man sets down a barbell and looks toward a far-off figure training in the light, the room stretching away into depth](/covers/gooaye-2026-08-19-ep689-cover.png) > Examine yourself and you may know others; examine the present and you may know the past. Past and present are one; others and I are the same. > > — *Lüshi Chunqiu*, "Chajin" (Warring States period) ## What this episode is about The first half of Gooaye's August 19, 2026 episode (EP689) barely touches the market. Asked by listeners which trainer he works with, the host gives a different answer: don't ask for the name. Go to a gym near you, find someone whose physique is already what you want to become, and train with that person. Along the way he shares an observation from a family trip — he has a "good behavior points" arrangement with his son, and the boy will spend one point on a Pokémon card pack worth a few hundred, and one point on a scoop of ice cream that was free at the buffet next door. The market discussion comes later. Tech names in both Taiwan and the US have pulled back, and two explanations are circulating: rising real rates compressing high multiples, and a report arguing that big tech's AI spending has roughly three trillion sitting below the waterline. He engages with both without accepting either wholesale, and lays out his own framework — splitting the market into three kinds of trades and explaining which one moves first and which one is the tell. He closes with two listener questions: how industry knowledge actually accumulates, and whether the quality of a research report has anything to do with making money. What follows is my own synthesis after listening, not a transcript. ## Key points **1. Motivation comes before method.** His advice on finding a trainer: search for gyms nearby, find someone whose body is close to what you want, and just start with them. That's how he began — he kept noticing one guy in the gym doing strange movements who spent most of his time training himself rather than coaching, and only later learned it was called functional training. He draws the parallel to English: his mother bought him an iPod that somehow had Coldplay on it, and those songs pulled him into the language. Before the interest exists, handing someone a textbook accomplishes nothing. **2. To a child, one point is both a card pack and an ice cream.** His son will spend a point on Pokémon cards, and then earnestly ask whether he can spend a point on a scoop of ice cream — which was free at the buffet. Adults don't behave this way; at an all-you-can-eat we calculate how much king crab it takes to break even. He tells this as a story about parenthood, but it's really a story about prices: the child isn't irrational, he simply doesn't have a price list yet. **3. He doesn't quite buy the "real rates" explanation.** Rates can be treated as a kind of gravity — the more you earn risk-free, the more compressed high-growth multiples become. As explanations go, it's serviceable. But his objection is concrete: July fell hard, yet lately more and more stocks have climbed back to their prior highs or made new ones, and not just one or two. After a rally of tens of percent, three down days of a few percent each is, in his words, genuinely fine. The reasoning method is worth borrowing: to test a macro explanation, he checks how broad the cross-section is, not how plausible the story sounds. **4. The same headline reads as bullish in an uptrend and as a ghost story in a pullback.** The report about three trillion of AI spending below the waterline concerns not-yet-commenced leases and purchase commitments — obligations that haven't hit the balance sheet, meaning real leverage is hidden. His comment: in a strong bull market, most people would have read this as bullish — look how much money they're committing, the supply chain is going to rip. The pullback is what turned it into a ghost story. He calls it the elephant in the room, admits nobody knows when it detonates, and guesses it won't be soon; it needs to simmer. The tell he offers is specific: when AI-related debt gets packaged into retail investment products, when even local outlets are selling "data center bonds, principal guaranteed," the explosion isn't far off. **5. Risk control by drawdown, not by forecast.** He gave back roughly thirty percent in July and stayed in, because the year had been strong enough that even after that giveback the remaining return was something he could hardly have imagined after consecutive bull years. His rule keys off annual performance: if a year's gains get erased in a single pullback, he'll consider sitting out even when the indicators still look positive — much like how a fund house restrains a portfolio manager whose monthly drawdown gets too large. He also punctures a marketing myth: selling at the top and buying at the bottom — show me once. Go back to June and ask whether you could have sold. When his wife wanted to buy something in June, his answer was "later" — because money thrown into the market was going up daily, and nobody was watching year-to-date; people watch the month, or the week. Normally you ride it to the last moment and eat part of the drop on the way out. **6. Three trades, in a specific order.** He splits the market into components (shortage and price hikes), bottlenecks (whichever stage has the most material piling up in front of it — back-end packaging and test, for instance), and value-add (the premium that comes from chip iteration, software updates, genuinely doing more for the customer). The classification matters less than the ranking he attaches to it: memory is the scarcest thing with the clearest visibility, so if even memory can't get its multiple lifted, what argument does any other component have for a higher one? Same for back-end packaging — what could possibly be scarcer? If those two stall, money may rotate back to value-add. He also notes bottlenecks get actively attacked: accept the price hike, buy out the capacity, or have the customer purchase equipment outright and place it in the packaging house with margins pre-agreed, so declining isn't an option. Their read is that Nvidia over-ordered certain optical components not because it needed them immediately, but so competitors couldn't get any. **7. A good report is not the same as a profitable report.** Asked how deep research needs to go, he starts with technique: don't model a whole company's blended average selling price — get the unit price and volume of the *new* component. If the earnings call won't tell you, read the calls of US and European peers; up and down the chain, companies leak clues about how many margin points the new product carries or how much it contributes to a segment. Then he turns: however good the research, researching a stock that won't go up gets you nothing. Conversely, a scrappy report with numbers all over the place that nonetheless catches the point is, to him, a good report — because every time a theme appears the market overshoots. Every company winning outsourced work claims a large share; step back and the claimed shares plainly don't add up. But the report still lifts the stock, because someone loves it and someone keeps revising upward. **8. Industry knowledge accumulates by island-hopping.** Asked how he knows about busbar cross-sections, cable bend radii, and connector mating cycles, the answer is unglamorous: all of it is residue from researching other things. Busbars came from a stealth company working on immersion cooling. Cables came from having to learn what every wire in a system does — eventually he could identify a vendor by wire color. Connectors came from studying socket and pluggable component makers, which is how he learned early AI servers could fail after only a few insertions, so testing simply minimized insertions to get through. His analogy is online gaming: once you know how to gank as a jungler, a new character takes minutes — learn what the abilities do, the logic is identical. The hard part is always the first island. As for information edge: it lets you know sooner, but it can't do the foundational homework for you. ## Going deeper ### "I got trapped in July — cut or hold?" — Replace the question with a number you can compute This is the most common question since July, and it's hard precisely because as phrased it demands a forecast. You don't know whether tomorrow falls further, so cutting scares you and holding scares you, and you end up deciding on the most painful day. The episode swaps the question out: don't forecast direction, look at giveback. He held through a thirty percent drawdown in July not because he was bullish but because the year's earlier gains were large enough that what remained was still a result he was happy to hold. Conversely, if a year's profit gets wiped out in one pullback, he'll rest even with positive indicators. What's really happening here is a substitution: "what will the market do" — unanswerable — becomes "where is my position relative to my starting point this year," which you can calculate tonight. And the arithmetic is asymmetric. As he notes in passing, recovering from a thirty percent giveback doesn't take a thirty percent gain: 1 ÷ 0.7 is about 1.43, so you need better than forty. Which means the line to define isn't "how far can it drop before I sell" but "how much of this year's result am I willing to hand back." Draw that line while calm, because the moment it triggers, what you'll feel is refusal. One more detail that's easy to miss: his composure comes from a strong first half, not from courage. He contrasts 2022 explicitly — that year you were losing from January onward, the felt experience was entirely different, and strategy adjusted accordingly. The same percentage decline sitting on a different starting point is not the same event. Before you copy someone's calm, check whether you share their starting point. ### "The news says AI spending is three trillion — bullish or bearish?" — Ask what it changed The same article reads as supply chain opportunity in an uptrend and as a crash warning in a pullback. That alone is the clue: what changed your view probably wasn't the article, but whether your account was green when you opened it. To read news as information rather than mood, ask what it changed. What the three trillion figure reveals is off-balance-sheet: leases not yet commenced, purchase commitments — real leverage larger than the books show. That's information, not a verdict. It enlarges the scale of "if demand plateaus, someone gets hurt," but it says nothing about when demand plateaus. The episode's posture matches: acknowledge the elephant, and still judge the blast isn't imminent, because AI monetization is currently visible — two years ago few dared imagine these companies would earn money this fast, and the fact that a slight deceleration is now being nitpicked shows the premise has already been proven. His warning sign, notably, isn't a financial metric but **retailization**: when AI debt gets packaged into products ordinary savers can buy, when someone is selling "data center bonds, principal guaranteed" in your neighborhood, risk has already been transferred to the last holder. The virtue of this signal is that you can actually see it — it shows up in your relatives' group chats, no footnote reading required. There's also the overcapacity thread. He deliberately translates the term into plain speech: overcapacity means demand isn't enough. Demand grew fast, everyone built generously, and then demand merely plateaus — not vanishes — but the plants, the borrowed money, and the equipment all still arrive, and somebody absorbs the pain. The 2022 correction looked like it started with war, but ultimately it was a supply chain inventory adjustment; in 2021 the belief was that whoever held the most inventory won, because one missing part meant no shipment. He sees a resemblance now — every US company tells you they're short, desperately short. So his watch point isn't "prices falling" but **prices holding flat and no longer climbing**, which arrives considerably earlier. ### "I do the homework — why don't I make money?" — Between homework and returns sits a layer called narrative This is the most demoralizing position: you read annual reports, follow earnings calls, decompose price and volume, and you're more diligent than ninety percent of people around you — and you earn less than someone who chased a theme. The answer in the episode is blunt: research quality and investment performance are different axes. A perfectly correct report may cover a stock that won't move; a sloppy one that catches the point can make you money. Because whenever a theme appears the market overshoots — his example is the current overflow of back-end capacity being outsourced, where every recipient claims meaningful revenue, though from a step back the claimed shares plainly can't all be true. Yet the reports overstating share and price still lift the stocks. So does the homework matter? Yes, but in the right order. His order is to pick the pond first — do your research where the industry genuinely has something, and even mediocre estimates can pay; pick the wrong pond and precision won't save you. This is the same idea as the three-trade framework: establish which layer money is flowing through (components, bottleneck, value-add), then pick names within it, rather than starting from a stock you happen to know and manufacturing a theme for it. The second layer is separating "is this report good" from "what is this report for." If you want understanding, clear industry logic is enough. If you want to judge price, add one question: are the assumptions in this report ones the market already believes? If so, they won't make you money again — they're already in the price. The third layer is technique. Public filings are historical; the parameters that decide the future are new orders, newly negotiated prices, and how much unit price steps up per generation. These are usually obtainable, just indirectly: if the company won't say, read peer calls, where clues surface — this product carries several more margin points, that line contributes such-and-such to a segment. Grind work, but reproducible grind work. ## Worth exploring - Gooaye EP689 (August 19, 2026) — the source of these notes; the segments on the three trades and on report quality are worth hearing directly - To feel the asymmetry of drawdown and recovery, plot 1 ÷ (1 − x) in a spreadsheet; it's clearer than any explanation - To track whether prices have stopped climbing, monthly pricing and earnings call transcripts from memory and passive component makers are the most direct primary material - To practice island-hopping: take a product you already understand (a phone or laptop), lay out its bill of materials, then look at what a server uses — the overlap is larger than you'd guess ## The one thing to take away If only one line survives: **find a person you want to become, and the method will follow.** In the episode this is about the gym — stop asking which training school is correct, find someone whose body already looks like your goal, and learn from them, because what stops most people was never an inadequate method but inadequate motivation. It generalizes to everything you know you should do and can't start. That's why the Coldplay on the iPod is worth remembering: what taught him English wasn't a textbook, it was wanting to understand those songs first. Method is what you need later; without the "I want to be like that," no method survives week three. **A practice for today:** pick one thing you've been putting off for three months or more — exercise, writing, changing jobs, a language, repairing a relationship. Don't look for more materials and don't make another plan. Instead write down **a specific person's name**: someone you know, or someone you can follow, who has already done it. Write the one attribute you actually want (not "they're impressive" but "they can still run ten kilometers at forty," "they actually finish the piece every week"). Then, within this week, take one concrete action connected to them — go where they train, send a message asking how they started, or read or watch their work all the way through once. You don't need the method figured out first. Methods follow people, not the other way around.