When Content Floods In, What's Scarce Is Someone Watching

What night has no moon? What place has no bamboo and cypress? There are only few idle men like the two of us.

—— Su Shi, Night Stroll at Chengtian Temple (Northern Song, 1083; translation mine)

The moon shows up every night. Bamboo grows everywhere. What has always been scarce is someone willing to stop and look. Written nine centuries ago, and it reads like a note on the 2026 content industry.

What this episode is about

The first half of Gooaye EP694 (2026-09-05) comes out of a forum on AI and entertainment the host attended the day before. It moves from “will content become worthless once AI floods the world with it” into Taiwanese capital, the games industry, and what “going international” actually means. The second half returns to markets: a choppy September, a Broadcom report that landed well while the stock did nothing, Snowflake’s margin structure, and a job title called FDE. Then listener questions — a few of which have nothing to do with investing and are the most durable part of the episode.

The main points

Public and private markets, on purpose, run opposite playbooks. In equities his team runs momentum: they don’t need to fully understand every position, just that it rides the wave and carries AI exposure — because by the time a story is fully imaginable, the move is usually over. In venture they invert it and do value work: meet the team, study the pipeline, compound alongside the company. The reasoning is plain — momentum and imagination are already available in the day job, so doing the same thing in a second market buys nothing.

Content inflation is not content devaluation. Plenty of people in 2023–24 believed AI would make content worthless. By 2026 his read is that the predicted paralysis of choice never showed up: you still know which directors you watch and which studios you buy from. Abundance didn’t drag good work down, it lit it up — “like a lighthouse” is his phrase. He’s neutral on AI-assisted creation, but he can tell what was polished from what was slapped together because a genre happened to be hot, and so can you: “you smell it immediately.”

Reputation, once broken, no longer recovers. It used to be rebuildable, slowly. Now there is too much choice and everything leaves a record, so losing it is likely permanent. He weighs that above any single payday.

He redefines “international.” Reading foreign media daily doesn’t count — outlets carry their own ideology, and some of it is the “only my view is valid” kind, which is closer to importing a virus than connecting to the world. Once being international is just conversational currency, it’s the same as treating AI as conversational currency: you can talk about it and can’t use it at work. What’s actually happening, in his view, is Taiwanese capital building and investing abroad — and foreign teams walking up after a panel to ask whether you’d look at them.

Game studios are turning into chip design houses. He maps the semiconductor split onto games: an indie studio is the fabless design house, AI tools are the rented fab, the engine is the EDA stack, Steam is distribution and worth its thirty percent. That structure turns a hit-driven business — one shot, miss and you’re done — into something capital can spread across many bets, until one lands and becomes a long-tail IP.

A good print with no move needs no explanation. Broadcom’s numbers and guidance were strong; the stock sat. His read: this cycle is in its third or fourth year for capital markets, plenty of names have doubled twice over, and no report is obliged to spark another leg. The hazard is reading the coverage that rationalizes the flat tape — enough of it and you quietly put a cross next to the company, when the fact may be “hasn’t moved yet.” Price follows the EPS trend in the end.

Snowflake’s margin hides in who picks the model. Under the default setting the platform routes each request to a model of its choosing. If the customer pays ten and the platform spends five, the gap is margin. Pin the model yourself and that gap thins. He connects this to software firms staffing up FDEs — forward deployment engineers, a product-manager-plus-engineer hybrid who goes in, learns your workflow, and assembles the solution. A router picks your model, an FDE assembles your requirements. Both are someone doing your integration, and who gets trusted with that may decide how the next round of profit is split.

Going further

”The report was great. Why won’t it move?”

Probably the line most listeners felt. Your position posts strong numbers, maybe raises guidance, then opens high and fades, or drifts sideways for a week. Open the financial press and it’s wall-to-wall explanation: customer concentration, competitors taking share, growth decelerating. By the fifth article you’ve quietly downgraded it in your head.

The trouble is that “not rising” and “getting worse” are two different things, and the coverage won’t separate them for you. Fundamental change lives in the numbers — revenue, margin, guidance, cash flow. Price change lives in the flows — how many times this theme has been played, what multiple the market will pay today, whether there’s something less extended to rotate into. A strong report moves the first set and leaves the second alone.

The real cost isn’t the trade you didn’t make. It’s that those explanations settle into your impression of the company. Three months later it prints something better and you can’t recall why you soured on it — only that “this one seems weak.” What I do is keep the two in separate columns: one for this quarter’s fundamental facts, one for price behavior, each line dated. The day the fundamental column turns is the day to act.

And the moment to admit you were wrong doesn’t come from the price. It comes from the sentence you wrote at entry — “if X happens, I’m wrong.” A position without that sentence leaves you running on feel when it drops, and feel betrays you exactly when you need it.

”The same playbook worked from April to July. Why is it bleeding now?”

He describes September: anyone running momentum has likely given back a chunk in a few weeks, because breakouts demand a chase and breakdowns demand a cut, and the round trip is buying high and selling low. Meanwhile the people who just held saw their equity grind upward.

What stings is that the losing trader executed correctly. Chase the breakout, cut the breakdown — that’s momentum done right. The market’s character changed and the method didn’t. His answer isn’t to stop: refuse to trade in a bad tape and, when it turns, the people who kept working collect the most while you discover the shift late and chase again. He shrinks leverage and size, keeps going, and scales up once the signal is clean.

That gives you a check on yourself. When a method stops working repeatedly, separate execution drift from regime change. Execution drift looks like the same mistake recurring — chasing, refusing to cut, adding to losers. Regime change looks like following every rule and still getting hit. The first calls for fixing yourself, the second for shrinking size. The prescriptions are opposite, and mixing them up makes it worse.

“AI can generate anything. Is what I make still worth something?”

If you write, design, or code, you’ve asked this. The episode’s answer runs the other way: once output is abundant, people care more about whether yours is any good.

Listening to that, I thought about the information environment in investing. A filing, a research note, an analysis — the cost of obtaining any of it is close to zero now, where a decade ago these were scarce. But as availability rises your bottleneck moves from “can I find it” to “can I get through it,” and attention becomes the binding constraint. Answering a listener, he puts it harder: attention is an asset, and most people hemorrhage it outward, above all into interacting with others. Some conversations you don’t want to answer, don’t. Some messages, don’t. Pull the energy back first.

Put those together and you get a way to judge others, and yourself: when everything is cheap, what carries value is someone having filtered out ninety percent for you. Snowflake’s router is that. The FDE is that. The few creators you trust are that. And if what you make filters nothing for anyone, the flood takes it — not because AI replaced you, but because nobody has the energy left to stop and look.

Worth a look

• Gooaye EP694 (2026-09-05): content and capital in the first half, markets and listener questions in the second
• The fabless semiconductor model — design, fab, packaging, and distribution split apart — used here as the analogy for game studios
• Forward Deployment Engineer (FDE): the product-manager-meets-engineer role software firms have been staffing up
• Content inflation: the surge in supply that followed generative tooling, and what it does to creators

One thing to take with you

One idea: attention is a stock, not a flow. What gets spent today doesn’t grow back tomorrow. You can leave it untouched for a year or burn it for a year, but you can’t burn it and expect to see clearly at the moment that matters. The listener who wrote in had lost interest in the world — good food stopped delighting him, good and bad news stopped moving him. The answer wasn’t to chase stimulation. It was to pull the outward-leaking energy back in, wait, and let the good thing find you.

Something I’ve tried, unrelated to investing, doable tonight: open your phone before bed and scan every message you answered today. Pick three where answering changed nothing — not the ones that would hurt someone, the ones where your reply got a sticker back. Tomorrow, when the same kind arrives, leave it. Three days later, check whether anyone minded. And check what you did with the minutes those three would have taken.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.