# The Story Flips Faster Than You Do: Meta's 28% Three Weeks and Admitting I Was Wrong on Netflix Source: Realpha Blog (blog.getrealpha.com) Original article and charts: https://blog.getrealpha.com/en/blog/compound-2026-09-22-meta-s-muse-launch-josh-is-wrong-on-netflix-intern/ > Notes from the September 22, 2026 episode of The Compound and Friends: Meta added roughly $450 billion in three weeks after launching Muse, why weak market breadth today isn't 1999, why only 23% of large-cap stocks beat the index over ten years, and how to think about the 10-year Treasury back at 5%. Educational, not investment advice. Published: 2026-09-23 Locale: en Tags: investing notes, US equities, Meta, Netflix, bonds ![Early morning on a high office floor, a person stands with their back to us at a full-height window, layers of city skyline receding into the distance, an open notebook covered in handwriting on the desk beside them](/covers/compound-2026-09-22-meta-s-muse-launch-josh-is-wrong-on-netflix-intern-cover.png) > Adapt and you succeed; insist and you fumble. > > —— *Lüshi Chunqiu*, "On Adapting" (Warring States period; translation mine) On the September 22, 2026 episode of *The Compound and Friends*, Josh Brown and Michael Batnick spent a long block on Meta: the AI assistant Muse launched on September 8, the stock went from $578 at the start of the month to the mid-$740s, and the company added roughly $400–450 billion of market value in three weeks — the largest single-month value creation in its history. In the same episode, Josh called his Netflix position his worst call of the year: he averaged down the whole way and ignored the death cross that told him to leave. Put those two together and they describe one problem — the market's story flips faster than most people revise their own view. What follows is what I took away. Every number below is one the hosts or the research they cited put on the record. ## What the episode covers This was Josh's return after three weeks away for a back injury and a conference. The show opens on Muse, moves through the chip stocks that rallied alongside it, asks whether the ugly market breadth everyone is pointing at is actually something to fear, works through the bull and bear cases on Netflix, walks through Adam Parker's research on how hard buy-and-hold stock picking is, covers the 10-year Treasury back at 5%, and closes on Buffett stepping down as chairman. My habit with podcasts is to keep only the parts that change how I'll judge something next time, so what follows isn't a play-by-play. ## The points worth keeping **One: what got repriced was an extra leg, not an existing business getting better.** Before Muse, Meta was stuck under the capex-with-no-ROI question, trading at 16 or 17 times forward earnings. Michael said that multiple sounded insane at the time and he still didn't want it, because the fears were reasonable — cheap was cheap for a reason. After Muse, analysts raised earnings estimates and the multiple at the same time; Josh mentioned one analyst going from 20 to 25 times. Same company, same filings, $163 of price in three weeks, and the difference is a revenue line nobody had in their model. ![Two rectangles aligned at the bottom-left corner: the smaller grey one inside represents the pre-rerating earnings estimate multiplied by seventeen, while the larger blue one grows along both edges at once, showing that the earnings estimate and the multiple were raised together.](/figures/meta-reprice-two-sides-en.svg) **Two: the rally spilled into CPUs and on-device inference.** Muse keeps a secure virtual machine for every user, handling browser sessions, tool calls and sandboxing — work that lands on CPUs and routes around the GPU layer everyone assumes. So AMD ripped and joined the trillion-dollar club (Intel, for reference in the show, sits at $615 billion), Qualcomm rose 9% and ARM 17%, the latter two selling chips that let phones run inference on the device. One app's architecture rewrote how a whole supply chain gets imagined. ![Three horizontal layers run top to bottom: the app layer, the GPU layer, and the CPU and on-device layer, with a blue arrow leaving the app layer, curving down the left side past the greyed-out GPU layer, and landing on the highlighted CPU and on-device layer.](/figures/muse-bypass-gpu-layer-en.svg) **Three: breadth is bad, but it's an order of magnitude away from 1999.** Jason Goepfert wrote that in almost a hundred years we haven't seen breadth this poor, with 1973 and 1999 as the only rough analogues. Michael went and got one number: with the S&P up 14% year to date, stocks down 30% or more make up 4% of the market. In 1998 that figure was 10%; in 1999 it was 14%. Both hosts had guessed north of 10%. Josh added the detail I liked most — in 1999 that 14% was retail. Sears was a Dow component. Radio Shack, Pier One, Sam Goody selling CDs, plus a wave of steel bankruptcies. The same "breadth is terrible" label can sit on top of very different things. ![Three vertical bars compare the share of stocks down more than thirty percent year to date: 2026 is a short bar, while 1998 and 1999 are much taller, and a dashed line marks the level both hosts had guessed.](/figures/breadth-label-vs-content-en.svg) **Four: Nvidia at 16 times forward earnings is itself a vote.** Michael asked which single stock personifies the AI trade; the answer is Nvidia, and Nvidia trades at a discount to the market, with semis as a group 10–20% below the S&P multiple. His read: there's enormous disbelief that these earnings are sustainable, and that disbelief is largely in the price. It doesn't mean the stock can't fall 30%. It does tell you what the market is currently betting. **Five: the hit rate on holding individual stocks is lower than people guess.** From Adam Parker's work: of the top 500 US stocks, 23% beat the index over ten years, and 28% over three. The brutal part is the cost of being wrong — stocks bought three years ago that lagged the S&P lagged it by an average of 62%. Michael pointed out something intuitive once you hear it: the one-year hit rate is better, because over one year you can credibly argue the market has overdone it. Over ten years, nobody in any field can say much of anything. ![Three short horizontal bars on the left show the share of stocks beating the market over one, three and ten years shrinking as the horizon lengthens, while a deep downward bar on the right shows that the losers trail by sixty-two percent on average.](/figures/buy-and-hold-hit-rate-en.svg) **Six: Buffett never bought and held.** Josh pushed back on the popular image: Berkshire's 13F shows buying and selling every quarter, there are maybe eight forever holdings out of hundreds that came and went. His description stuck with me — Buffett lets the market tell him a stock is a forever stock, and ten years later there's no reason to sell. In the first year or two, you rarely find him sitting on losers. The show also put up the number that Berkshire could fall 99% and still have beaten the S&P 500 since 1965. **Seven: at 5% on the 10-year, missing the price costs you something different than in stocks.** Michael said he has never owned bonds and now sees this as sound allocation. Josh took the logic apart: buy a 10-year at 5%, watch it go to 6%, and yes you gave up some upside — but you already have a 5% cushion, so the damage to total return is contained. A stock bought at 10 and trading at 8 has nothing underneath it; equity can go to zero. He added two things nobody stops you from doing: reinvest the coupon into higher-yielding paper, and ladder in (half now, half if it reaches 5.5%). He also noted 5% is historically normal — the 2010 to 2020 stretch was the aberration. ![Two panels each hold a falling line: on the left the line drops until it rests on a thick blue coupon floor, while on the right there is no floor and a dashed line keeps falling all the way to zero.](/figures/coupon-floor-vs-no-floor-en.svg) ## Going further ### "The good news is already out and I missed it — is chasing here a mistake?" That was my first thought during the Meta block, and it's the question Josh put to Michael: does this have legs? Michael's answer was that the difficulty of buying it here is exactly why it goes higher. I won't take that as a conclusion, but it points at the right layer of the question. Whether to chase depends on what got repriced. If the market repriced an improvement in a known business, the move is usually a one-day event and it's over by the close. If it repriced a new revenue line — a leg nobody had modeled — then what has to change isn't just next year's EPS but the multiple people will pay, and multiple changes unfold over quarters as data arrives. With Meta you saw both estimates and multiples move up. That's the second shape. So rather than ask "has it run too far," I'd ask "how many years of earnings does this touch." If you can answer that, you know whether you're looking at a one-day story or a three-year one. If you can't, say so and let it run without you. ### "I'm holding something that keeps falling and I can't make myself sell" Josh's Netflix confession is the most human stretch of the episode. He bought around $100 in the summer of 2025, the company waded into the Warner Bros. situation, and the stock fell off a cliff from there. He added the whole way down and was wrong about it on television, in writing, and on this show. Michael is in it too, average $81 against a $73 stock. ![A falling price line carries three marks where more was bought on the way down, two moving averages cross mid-chart in a circled death cross, and the line ends below the average cost.](/figures/netflix-death-cross-ignored-en.svg) What's interesting is that the two of them disagree about where the exit was. Michael points to the breakdown out of the sideways range. Josh says no — the exit was the death cross, the 50-day moving average crossing below the 200-day. He described the signal honestly: not science, not an automatic silver bullet, just a signpost that the psychology in a stock has meaningfully changed. Netflix's fundamentals hadn't changed at that point. People's enthusiasm had. The line I keep coming back to is what he said next: he uses stops, and here he did the opposite. The rule existed, he knew it, and he chose not to follow it. That has nothing to do with intelligence and everything to do with a stock owing you something — his words were "it owes me," and he insists on making his money back in this particular name. It's the most expensive posture available, and he said it out loud on the show. So here's what I do now: write down what would make me leave at the moment I buy, next to the entry in my records. The point is writing it while nothing has gone wrong and the emotion is neutral. Once you're underwater, the condition you write will quietly become "sell when it gets back to my cost," and your cost is the one number in the market that doesn't care about you. ![A timeline holds a condition box at each end: the one written on the day of purchase clearly says to leave if the price breaks down, while the one written after being underwater is tilted and dashed, and now says to sell on getting back to cost.](/figures/exit-line-written-when-en.svg) ### "Then where should I even start looking for stocks?" Parker's hit rates make it easy to conclude you should just buy the index. I don't argue with that conclusion, but Josh offered a different angle: the starting point. Most people start with the news. You turn on a financial channel, people trade tickers back and forth around a table, and you assume a stock in the news is a stock worth knowing. His approach is to start from a couple hundred names that are already making people money — price rising, valuation rerating higher, earnings growing, buyers accumulating. He quoted his late friend John Bollman: if you want a stock because it'll go up, start with a stock that's already going up. I read this as shrinking the search space. Part of why the hit rate is low is that most people's search space is "every stock anyone mentioned," and the signal density in that pool is close to zero. Swap in a pool you defined in advance and at least you know where you're fishing. This is separate from chasing — the pond changes, the exit rule from the previous section still has to be written. ![A large grey dashed circle holds sparse scattered dots, and inside it toward the right sits a much smaller blue circle whose dots are visibly dense.](/figures/shrink-the-search-pool-en.svg) ## Worth a look - *The Compound and Friends*, episode dated September 22, 2026 (Josh Brown, Michael Batnick) - Adam Parker / Trivariate Research on buy-and-hold hit rates, cited in the episode - Jason Goepfert and Ryan Dietrich on market breadth, cited in the episode - Sell-side views discussed on air: Evercore ISI, Wells Fargo and HSBC on Netflix ## One thing to take with you **Write down the line where you let go, while you're still even, still fresh, and still emotionally neutral.** Both hosts had rules and knew them, and when the moment came they chose not to follow them. The difference is that once you're inside the situation, the condition you're able to write bends toward getting you back to even. All the value of that line comes from when it was written. Here's something I've tried that works outside investing too: pick one thing you're currently grinding through — a language you've studied for years without progress, a collaboration you keep postponing, a job you tell yourself to "give it a bit longer." Write one sentence: "If by such-and-such date this specific thing hasn't happened, I stop." Make the thing concrete enough that someone else could judge whether it happened. Fix the date. Then put the paper somewhere you'll see it in a few days — first page of a notebook, taped to the edge of a monitor. I've written three of these. Two of them came due and I chose not to act, with reasons that sounded perfectly good. But because the paper existed, I at least knew what I was violating.