@ShadyJosh5@PaperBagInvest Every AI proof point in the Q2 letter is on the expense side — LAE 7% → 5%. Meanwhile the attritional gross loss ratio went 54% → 59% QoQ and 58% → 59% YoY.
Is there any evidence yet that the AI edge shows up in pricing, not just cost? $LMND
@PaperBagInvest Every AI proof point in the Q2 letter is on the expense side — LAE 7% → 5%. Meanwhile the attritional gross loss ratio went 54% → 59% QoQ and 58% → 59% YoY.
Is there any evidence yet that the AI edge shows up in pricing, not just cost? $LMND
@Upstart $UPST Q1 adj EBITDA was ~$40M. SBC that same quarter was ~$36M — essentially all of it. @paulxgu has stressed minimizing dilution. How does ~$140M annualized SBC square with that? And when does cash actually start accumulating after it?
@Upstart $UPST OCF includes HFS loan purchases/sales, so reported FCF is hard to read. What was underlying FCF in Q2 ex-loan flows? And when does reported FCF turn sustainably positive — gated on the balance sheet step-down, or on EBITDA scale?
@Upstart UPST @paulxgu framed auto/home margins as always-improving, never “mature.” Will you break out segment-level contribution margins for auto & home? If not — what internal metric shows they’re tracking core’s margin path?
@Upstart UPST Held-for-sale jumped ~84% to ~$724M in Q1. How much actually cleared via sales in Q2? And did the average time loans sit on the balance sheet before selling go up or down vs Q1
@paulxgu@macro_guru@Upstart If the lower take rate is the intentional LTV lever and not just mix, some data on the volume and lifetime-value payback it’s buying would really make the case☺️
In May 2026, the $LMND website had all-time bests for:
1) Web visits 3.7 million
2) Global website ranking 11,078
3) US website ranking 2,321
4) Finance website ranking 23
5) Lowest Bounce rate 38%
None of this means the market is always right. Markets make mistakes all the time. In fact, some of the greatest opportunities in investing come from identifying situations where the market has become too optimistic or too pessimistic. But investors should be careful about assuming the market is stupid simply because a stock continues falling while current results continue improving. Often the market is focused on a completely different question.
The real challenge is determining whether you are looking at a business whose future is brighter than the market believes or a business whose past is brighter than its future. That single question explains a remarkable percentage of investment outcomes. It is also the question sitting at the center of the entire $ADBE debate.
Ultimately, $ADBE is not just a story about software or artificial intelligence. It is a reminder that investing is an exercise in forecasting rather than observation. The present is visible to everyone. The future is where the debate takes place, and that future is what the market is attempting to price every single day.
For what it’s worth, $ADBE falls into the “too hard” pile for me. I am neither particularly bullish nor bearish, as I think the long term impact of AI on the business is difficult to predict with confidence. Since $ADBE is one of the most discussed stocks on this platform and generates strong opinions from both bulls and bears, I thought I’d share my two cents.
🌹
2/2
Thoughts on $ADBE
$ADBE is one of the most fascinating stocks in the market today because it highlights one of the most important lessons in investing. It continues to execute at a high level. Revenue continues to grow, margins remain exceptional, free cash flow is enormous, and millions of customers still rely on $ADBE products every day. Yet despite all of that, the stock has struggled for years hitting all time lows.
This confuses many investors, especially newer investors. They look at the financial statements and see a business that appears healthy. Then they look at the stock price and assume the market must be making a mistake. After all, if the business is improving and the stock is falling, shouldn’t that create an even better opportunity?
Sometimes the answer is yes. Some of the greatest investments in history occurred because the market became too pessimistic about a business whose future remained bright. But it is important to remember that the market is not trying to value what a company earned previously or even currently. The market is trying to value what that company might earn in the future.
This is where the story becomes interesting. $ADBE looked cheaper at $500 than it did at $600. It looked cheaper at $400 than it did at $500. It looked cheaper at $300 than it did at $400. Many investors looked at the declining valuation and concluded that the opportunity was becoming more attractive. Yet the stock continued to fall because investors were not debating the current business. They were debating what the business might look like in the future.
For decades, $ADBE built one of the strongest moats in software. Photoshop, Illustrator, etc became the standard tools used by creative professionals around the world. Entire careers were built around learning Adobe’s products. Millions of designers, marketers, photographers, and video editors integrated $ADBE into their daily workflow, creating an ecosystem that appeared almost impossible to disrupt.
Then artificial intelligence arrived and changed the conversation. For the first time, images could be generated with a prompt. Videos could be created automatically. Design work that once required years of expertise could suddenly be performed by almost anyone. The question investors began asking was not whether $ADBE remained a great company today. The question was whether $ADBE moat would be as strong five or ten years from now as it was five or ten years ago.
That distinction is incredibly important because stocks are ultimately claims on future cash flows, not current cash flows. Imagine owning a toll bridge that earns $100 million per year. If someone announces that a second bridge will be built beside yours five years from now, the value of your bridge immediately changes even though today’s profits remain exactly the same. Nothing changed in the present, but something changed in the future.
This is why investing can be so difficult. The numbers investors see today often tell a very different story than the future investors are attempting to price. A business can appear healthy while its long term competitive position weakens. At the same time, a business can appear expensive while its future becomes far more valuable than most people realize (ie $PLTR). The market spends surprisingly little time pricing the present and an enormous amount of time attempting to price a future that has not yet happened.
This is also why one of the most dangerous phrases in investing is, “The stock is down but the fundamentals are improving.” Investors have said that about newspapers as the internet emerged, department stores as ecommerce gained share, and cable television as streaming began taking over. In many cases the current business remained healthy long after the future business had already started to deteriorate.
1/2 👇
@HenryInvests@Upstart On the beneficial interest and co-investment positions Upstart retains in its capital deals — is the company’s downside capped at the carrying value on the balance sheet, or could realized losses exceed what’s booked through recourse, repurchase obligations, or guarantees?
In my opinion, this is true for $UPST $LMND $OSCR $TEM $TSLA etc - these advantages will only widen & accelerate over time.
This is the application layer. What's interesting is that Agentic AI caused a sell-off in the application layer b/c investors believed people could vibe code basically anything and therefore software / unique tech was practically worthless.
There are several reasons this is wrong - the most obvious of which are the proprietary data sets Ellison mentions. And no, just because legacy incumbents have lots of data doesn't mean they can catch up. 1) Data is not valued unless trained w/ relationships examined over long periods of time measured against outcomes and 2) As Ellison said, public internet data has no alpha and will be commoditized.
Agentic AI will accelerate the operational efficiencies of companies with true proprietary datasets. And it will enable even greater dynamic understanding of the underlying data and relationship mapping.
The future will be dominated by AI-native application layer companies with proprietary data sets. These companies will have such an advantage on both underlying efficiency and data that I predict many will slowly grow into behemoth monopolies. Legacy incumbents will crumble and their downfall will be fast and unforeseen to most - but not to me.