MrPoastMan (@MrPoastMan) header
MrPoastMan (@MrPoastMan) profile picture

Mostly investing ideas (not financial advice), occasional geopolitics or other topics. I love models and data, somewhat get this whole AI thing.

US-based
Joined June 2025
Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)Photo by MrPoastMan (@MrPoastMan)
Hey everyone! New account to work on putting my ideas and research out there. Mostly investing ideas (not financial advice), occasional geopolitics or other poasts.
1
3
1,057
Going to do a little account takeover of my own account today to cover something genuinely cool happening right now: the DARPA Lift Challenge. Three reasons this is worth your time: 1) At the end of the day I'm a technologist. Watching technology actually move forward in public is cool, and this is about as public as it gets — 7+ hours of live broadcast, teams failing on camera, engineers explaining what broke. 2) There are insanely fun clips coming out of this competition. Aircraft flipping over prep tents. A rotor blade snapping and cutting off its own tail boom on the run that would have won. I'll be posting them. 3) If this goes well it's informative for a lot of theses at once. Drone delivery platforms (DoorDash, Wing, and everyone chasing them). Second-order pressure on traditional food delivery economics — Uber Eats et al. The autonomy workload this does *not* require versus what FSD is carrying. And if delivery miles move off the road, that's fewer driver-miles, which eventually shows up in auto insurance exposure. I always like watching where the puck is headed. Hope y'all get something out of the change in coverage.
1
2
612
I want to say thanks to @shai_wininger for taking my question. It was a great & exciting answer on the sophistication of LMND's use of AI. Ask yourself how do you think State Farm or Geico or Progressive or ... is executing on this front. This is pre-web -> web all over again.
3
264
SAY QUESTIONS ARE UP! And, in a rare triumph over my traditional personal bent toward perfectionism, I’m actually posting them with some time left this quarter. I tried to focus the 3 questions on the things that matter most to whether the $LMND machine keeps compounding over a multi-year timeline—not just whether the next quarter is good: Q1 — AI Durability: Daniel has said an AI-native company should improve almost immediately when a better model arrives. I’m asking for a concrete example: on LMND’s last major model/vendor change, how long did evaluation-to-production take, and what measurably improved in automation, human escalation, or cost? saytechnologies.com/s/1cb359… Q2 — Growth Assurance: Growth spend has tripled while LTV/CAC somehow stayed around 3x. I want to better understand how management separates incremental customers from merely attributed ones, detects channel saturation, and benchmarks something like the graffiti/OOH campaign against the next dollar in Pet or digital. saytechnologies.com/s/00c4c4… Q3 — Inflation Risk: In 2022, inflation outran approved rates and forced LMND to slow growth. Given elevated structural inflation risk and the potential for more discrete shocks, what has changed in LMND’s modeling and planning—and what now triggers rate filings, tighter underwriting, or slower growth? saytechnologies.com/s/2433de… Basically: can the tech advantage persist, can the advertising machine keep scaling intelligently, and did management build a better early-warning/action system after the last inflationary cycle? If you’re a verified shareholder and think any of these deserve an answer, I’d greatly appreciate you tossing some shares behind them. Feel free to throw them into any $LMND Discords/group chats where people might find them useful too.
1
1
1
10
1,870
An interesting thought surrounding personal computing hardware (tablets, phones, computers, future VR). In a world where software becomes essentially free and platformless via LLMs. Whether you're on a "PC or Mac" or "Android or iPhone" -- software becomes much less of a differentiating feature. Now, more than ever, it is or will be super valuable to have the "most preferred" consumer hardware, which is probably some combination of extreme vertical integration, good engineering, taste, and an incredible brand. Do with that what you will.
91
KKR has Insurance as the most disruptable industry by AI, as they switch to capital-light structures. Humorously, they don't mention there's one company already excelling at this new model. nitter.cf/lhamtil/status/2054243…
I don't know how they quantify exposure to AI, but an interesting plot from KKR showing which industries are least and most exposed to AI disruption. kkr.com/content/dam/kkr/insi…
1
7
2,090
Lots of points of discussion in the community right now. I want to tap this sign again. LMND is a modeling and control system company. They're spending because they told us when the sign would flip -- and they want to maximize what they can plough in before the sign flips. They want what's available for S&M post sign flip to match their spend beforehand. Even better, I can tell you from looking at and modeling this very thoroughly, their best area to spend right now is in Pet. And, guess what, that's exactly where they spent the most, by a long shot. If you're generally "long" and yet feeling frustrated at S&M spend levels, I think if you can really try to understand what's going on here, you can find a ton of peace with the current strategy.
- Why was lemonade just 5M from positive in Q4, why not do it?!? - If we're sooo close now, why didn't they up their guide and tell us they're bringing the first positive Q one or two earlier?!!!! - Bloody hell, we're being really patient shareholders, some have held from IPO prices, can't management's guide finally give us a hand? I actually think the answer to these questions are obvious, important, and telling. So here's my theory. Roughly speaking, you can model the efficacy of LMND's growth spend pretty well as just coeff*spend + quarter_effect -> new_customers. I think it's like a .7+ explanation with just an extra, additive Q3 effect dummy variable. Spend is roughly equally effective per quarter when you control for the impact of the quarter on its own. It's possible to also make a technical argument that Q4 has a tiny negative factor. You can make an argument that there's "signal" from a given quarter's spend that flows forward. But, primarily speaking, spend in a quarter is recognized as new customers in the same quarter. I think we all know/suspect that we could have had a positive EBITDA quarter, BUT management isn't ready to give us that until a fully-sustainable EBITDA profitable path is locked in. Furthermore, we know management's preference for growth over short-term profitability. So, to allow Q4 to flip positive (even though it is maybe a slightly worse quarter to spend CAC on) -- would then present them the dilemma of trying to stay positive or go through a sign flip (negative) for another 2-3 quarters (Q1-3), if they wanted to spend at similar levels in 2026. Furthermore, if they didn't spend at similar levels, it would take them at least about 1.8 years to grow back to where they were spending the same amount as Q4 2025. Whereas, right now, their exact quarter of transition should roughly allow them to spend the same amount in Q4 2026 as Q3 2026 while remaining EBITDA profitable, allowing for a clean transition in marketing spend. Obviously, if LRs keep getting better, technically they could match the current rate sooner, but why not continue to pull this forward until the absolute second possible, unless you're showing diminishing returns in growth spend. In all of my modeling of the company, I've really come around to the idea that it's not a "quarterly" company. Contracts are a year, they know ADR (annual), they know customer ages, they know pet ages, they know turn over probabilities, and some payments are even annual recognized pro-rata. So, fundamentally, even when a quarter is announced, they have supreme visibility into the next quarter (especially by the time of the report), and arguably have a decent lens into the non-growth-driven part 4 quarters out. I say all of this to say, that the comments that really stuck out to me on this call were Tim's. He said a lot about how Q4 is a really good mental model for 2026. He even dropped a projected number for growth spend, 225 M. I think we're going to see 3 more quarters where we all have sneaking suspicions they could flip positive in any one of them. But, I suspect we won't get it till Q4. They could of course surprise in Q3, but I actually kind of doubt this, as I don't think they'll want to forego their last chance to really plow into CAC in their last projected negative quarter. He said revenue is going to see another similar jump too, which is probably largely a factor of the continued decline in cede rate -- but, perhaps we might even see a further drop here, as they clean their book more and feel even more comfortable (what's a reasonable number is not really my expertise). I haven't really worked out the exact numerical consequences of this, but my intuition tells me is that basically 2026 is just going to be a ton of delayed gratification to where we then really benefit from expectations more being beat on the customer acq front, the cleaning the book front (and LR), and then this will translate to much nicer numbers in IFP, ADR, etc from Q4 2026 onward. What do y'all think?
4
1
24
2,974
HELP NEEDED! Say Questions: In my traditional personal bent toward perfectionism, I put off selecting my exact questions from my long list until the very last minute. This gives me a window of ~3 hours for some of you all to maybe toss some shares behind the Qs. Let's see if the Euros and the Early Birds can help me get a question or two answered this call. Q1: Daniel's called staying ahead of inflation a "cat and mouse game" with built-in lag. At 80% retention now versus 45% back then, that lag hits a lot harder. Has management modeled what 5+ years of persistent 4-5% severity inflation does to the path to and beyond EBITDA breakeven? saytechnologies.com/s/3c9226… Q2: Growth spend's gone from ~55M to 225M in 3 years and management's said they've been surprised LTV/CAC held up despite 10-15% annual CPC inflation. Has management modeled where the diminishing returns inflection is, and at what spend level you'd need fundamentally new channels? saytechnologies.com/s/f39a1a… Q3: Given the criticality of AI to LMND, what is its COOPlan for AI failures/cost hikes? If its provider went down is there a tested failover? How does LMND weigh providers vs self-hosting? How many non-actuarial models does LMND use and are any simple modalities like STT in-house? saytechnologies.com/s/22f283… I'll do my best to do better with timing next quarter. Thanks for the help! If you're any discord or something where people would vote between now and the deadline (3 hours away), feel free to throw them out there too!
5
6
1,067
I just gotta say. I fucking love AI. I've wanted it to be this good since 2009 (or earlier). I pray that it won't be pandora's box for society. High potential for it to be that. E.g. what do we all do when a vast majority of professions (and the investment into them) aren't worth the same thing. That being said as a person who has always had endless questions about things and often needs decent answers to keep the ball rolling on complex processes (see: taxes), it is a godsend to have an infinitely patient, incredibly competent partner -- and partner is probably underselling it by a lot.
3
59
- Why was lemonade just 5M from positive in Q4, why not do it?!? - If we're sooo close now, why didn't they up their guide and tell us they're bringing the first positive Q one or two earlier?!!!! - Bloody hell, we're being really patient shareholders, some have held from IPO prices, can't management's guide finally give us a hand? I actually think the answer to these questions are obvious, important, and telling. So here's my theory. Roughly speaking, you can model the efficacy of LMND's growth spend pretty well as just coeff*spend + quarter_effect -> new_customers. I think it's like a .7+ explanation with just an extra, additive Q3 effect dummy variable. Spend is roughly equally effective per quarter when you control for the impact of the quarter on its own. It's possible to also make a technical argument that Q4 has a tiny negative factor. You can make an argument that there's "signal" from a given quarter's spend that flows forward. But, primarily speaking, spend in a quarter is recognized as new customers in the same quarter. I think we all know/suspect that we could have had a positive EBITDA quarter, BUT management isn't ready to give us that until a fully-sustainable EBITDA profitable path is locked in. Furthermore, we know management's preference for growth over short-term profitability. So, to allow Q4 to flip positive (even though it is maybe a slightly worse quarter to spend CAC on) -- would then present them the dilemma of trying to stay positive or go through a sign flip (negative) for another 2-3 quarters (Q1-3), if they wanted to spend at similar levels in 2026. Furthermore, if they didn't spend at similar levels, it would take them at least about 1.8 years to grow back to where they were spending the same amount as Q4 2025. Whereas, right now, their exact quarter of transition should roughly allow them to spend the same amount in Q4 2026 as Q3 2026 while remaining EBITDA profitable, allowing for a clean transition in marketing spend. Obviously, if LRs keep getting better, technically they could match the current rate sooner, but why not continue to pull this forward until the absolute second possible, unless you're showing diminishing returns in growth spend. In all of my modeling of the company, I've really come around to the idea that it's not a "quarterly" company. Contracts are a year, they know ADR (annual), they know customer ages, they know pet ages, they know turn over probabilities, and some payments are even annual recognized pro-rata. So, fundamentally, even when a quarter is announced, they have supreme visibility into the next quarter (especially by the time of the report), and arguably have a decent lens into the non-growth-driven part 4 quarters out. I say all of this to say, that the comments that really stuck out to me on this call were Tim's. He said a lot about how Q4 is a really good mental model for 2026. He even dropped a projected number for growth spend, 225 M. I think we're going to see 3 more quarters where we all have sneaking suspicions they could flip positive in any one of them. But, I suspect we won't get it till Q4. They could of course surprise in Q3, but I actually kind of doubt this, as I don't think they'll want to forego their last chance to really plow into CAC in their last projected negative quarter. He said revenue is going to see another similar jump too, which is probably largely a factor of the continued decline in cede rate -- but, perhaps we might even see a further drop here, as they clean their book more and feel even more comfortable (what's a reasonable number is not really my expertise). I haven't really worked out the exact numerical consequences of this, but my intuition tells me is that basically 2026 is just going to be a ton of delayed gratification to where we then really benefit from expectations more being beat on the customer acq front, the cleaning the book front (and LR), and then this will translate to much nicer numbers in IFP, ADR, etc from Q4 2026 onward. What do y'all think?
4
1
17
4,812
Someone needed to raise money. Wouldn't it be fun to have a bird's eye view and know if the drop was due to a capital raise for OpenAI. nitter.cf/sama/status/2027386252…
We have raised a $110 billion round of funding from Amazon, NVIDIA, and SoftBank. We are grateful for the support from our partners, and have a lot of work to do to bring you the tools you deserve.
3
1,058
Modeling question: if LMND literally stopped acquiring customers today and just let the existing book decay to zero -- what would the company be worth? 𝗦𝘁𝗮𝗿𝘁𝗶𝗻𝗴 𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻: 2.98M customers 414 PPC 1.24B in-force premium $1.05B cash and investments on the balance sheet. Licenses in all 50 US states + Germany, Netherlands, France, UK. The math is pretty clean. ADR is ~0.85, meaning 15% of premium dollars leave every year. The lifetime value of a geometrically decaying book is just: 𝙏𝙤𝙩𝙖𝙡 𝙋𝙧𝙚𝙢𝙞𝙪𝙢 = 𝙄𝙁𝙋 / (1 - 𝘼𝘿𝙍) = 1.24B / 0.15 = ~8.2B Apply a 0.60 GLR (conservative -- they just printed 0.52 in Q4): 𝙂𝙧𝙤𝙨𝙨 𝙋𝙧𝙤𝙛𝙞𝙩 = 8.2B × 0.40 = ~3.3B OpEx in runoff is minimal. Current OpEx minus growth spend is ~100M/quarter, but in shutdown you gut most of that. Maybe 100M/year, decaying with the book. The book is half gone in ~4 years. Total runoff OpEx is roughly $400-500M. 𝙉𝙚𝙩 𝙐𝙒 𝙋𝙧𝙤𝙛𝙞𝙩 = 3.3B - 0.5B = ~$2.8B Discount at 10% over ~4yr average duration → ~$2.0-2.3B in present value. 𝗦𝘁𝗮𝗰𝗸 𝗶𝘁 𝘂𝗽: PV(Net Underwriting) ~$2.0B Cash & Investments +$1.05B License & Tech Sale +$0.25B ─────────────── 𝗧𝗼𝘁𝗮𝗹 𝗥𝘂𝗻𝗼𝗳𝗳 𝗩𝗮𝗹𝘂𝗲 ~$𝟯.𝟯𝗕 73.4M shares → ~$𝟰𝟱/𝘀𝗵𝗮𝗿𝗲 I think it's useful to have this number in your head as a floor. Whatever you think the growth is worth, this is roughly what's underneath it. Reasonable people can haircut the GLR assumption or the OpEx estimate, but even being pretty conservative you're probably in the $38-45 range. Yes, these things would probably get discounted in a sale -- but it is another lens into valuing what's there today. At the current price today, you can buy a ton of growth upside optionality for 7.50 a share. Not bad.
4
1
1
35
4,541
Me hammering out a project trying to leverage all of Claude's new features, while not having fully grokked them.
128
New term I learned today. Seat Compression. I like it. If companies still need mission critical software like "Salesforce", sforce might still get hurt if the company used to have 500 people, but now only has 400 or 300 from AI efficiencies.
98
Lemonade has regulation to help be a moat against new players. Against, old players, it has "the mile deep" data Daniel likes to talk about versus incumbents. 8200 vets did things like Waze, etc. They're who you want designing that mile of depth. I like our moat.
Replying to @ShadyJosh5
Highlighted portion says it all lol. $LMND research is in Israel. In Israel, 8200 is _THE_ professional network. As a long, I'm glad he's on the team. AI-driven does not mean 0 to 1. Think about everything that still needs to be measured, such as when to switch from a paid API to a self-served model.
2
18
3,829
Forest from the trees. January's numbers are in. They look great. Obviously the FSD bump shows up in this data, but 2026 is off to a great start.
2
3
1
44
4,518
One of my favorite stats from investor day, every 10% discount, 1.5-2.5x increase in conversion. Today's FSD discount could mean a 7.5x to 97x increase in FSD user conversion. When it expands to all of $LMND's easy telematics areas, that could mean ~50-100k new customers from today. Or, something like ~50-100M new IFP.
10
494
For context, the reach so far is about 1/10th of a Super Bowl ad for free. (Well, not for free, but for hard, business-improving work).
Wow. wow. wow. 35.8k posts today. Elon retweets, Chamath retweets, Polymarket retweets. Google search spike. 17 million impressions on the biggest post.
12
515