Calum Lewis

$AMZN at 15.4x operating cash flow 10 year average: 25.7x 10 year low: 15.1x after last night's sell-off it's sitting right on the floor of its own range i'll be adding more here. currently my #3 position. not financial advice

1/3 $APP is growing revenue at 53% and the stock is 68% off its high 🧵 i don't own it but valuation is very interesting...

$DUOL is back at ~$141, where it traded in october 2023 oct 2023: - 21.4M daily users - 5.2M paying subscribers - $127M quarterly revenue now: - 58.7M daily users - 12.7M paying subscribers - $298M quarterly revenue the stock went nowhere in 3 years. the business more than doubled fair price now, or still not cheap enough?

$HIMS revenue: +38% $HIMS stock: -13% this year Someone's wrong 👀 1. The growth is coming back Revenue $753M, up 38% US up 16%. International up 17x to $131M, mostly from acquisitions 2.89M subscribers, up 19% Each one paying $92 a month on average, up 21% 2. The catch Gross margin fell to 64%, from 76% a year ago They lost $86M in the quarter And they're honest about it. CFO Yemi Okupe: "we expect gross margins will remain below the levels we have historically achieved." Bigger business. Thinner slice. 3. Andrew's framing "healthcare is finally adopting the consumer-centric model that is already standard in industries like entertainment, travel, and transportation." That's the bet I signed up for. GLP-1s are just one part of it 4. The wildcard Peptides. The stock jumped about 6% last Thursday after reports a plan to ease compounding rules reached White House review Nothing's final. I'm not counting it 5. Questions I keep asking myself Can margins stabilise once weight loss and international settle? Can we see sub growth from here without acquisitions? Is $880-900M next quarter (up 47-50%) the floor or the ceiling? Net net Growth came back. Profit hasn't yet. I'm fine being patient right now, as I believe they are best positioned. That's how I'm thinking about it. Would love to hear where you disagree.

The UK tax-free allowance is £12,570. It's been stuck there since April 2021. It stays frozen until April 2031. It's called fiscal drag. Wages rise, the threshold doesn't, so more of your pay gets taxed. The government expects it to pull 700,000 more people into income tax. Fair way to raise money, or a stealth tax?

How would you rate my portfolio? What would you change?

$ODD reports earnings next week for Q2. Very excited for this one. My current read on the CPA problem is more mixed than bullish or bearish. May looked encouraging. IL MAKIAGE CPA improved roughly 28% sequentially from April. But my read from watching the Ads Library is that June and July did not show a clean return to normal. The issue may have improved. It just may not have improved enough to scale confidently. A lot of ads are being cycled every 7–14 days. Some ads survive for four weeks, which is better than constant churn. But most still seem to disappear before they become a clear winner. That usually points to one of two things: 1. They still have not found creative that scales profitably. 2. They are retraining the pixel and testing different audiences, landing pages and offers. The problem is that the outside view looks almost identical either way. I suspect / estimate they are currently running/pushing heavily into: Multiple Meta accounts. Multiple pixels (Google + Meta). Multiple domains (with redirects). Redirect tracking across different landing pages. Some of these domains had history going back to July 2025, so this strategy did not suddenly begin with the 2026 disruption. That matters. The company has always potentially used a multi-account and multi-domain setup to some degree. The difference now appears to be how heavily they are leaning into it. That suggests the problem is not necessarily unresolved. But it does suggest they are still searching for a reliable way to rebuild quality leads / growth or diversification. There are also some smaller things that have disappointed me. They are re-running exact copy and creative from 2024 and 2025 without much attention to detail. I even saw an ad still referring to 2024. The website metadata and SEO also look weaker than they should for a company with this much data and technical talent. None of this proves the core business is broken. But it does raise questions. A company can have strong data science and still make basic marketing mistakes I guess. Another potentially positive signal is the return of more try-before-purchase offers on some of the ads library. That model can create higher returns, but it may also produce stronger engagement and better conversion data. So even that is hard to read from the outside. The same applies to the reported improvement in credit-card data signals. If those signals are real, they could help the platforms find better customers. But the important question is whether that eventually shows up in lower CPA, stronger first-order volume and better repeat cohorts. Not just more activity. I estimate there are now roughly 600–800 active ads across the different accounts. That sounds impressive until the comparison is made properly. A large number of live ads does not mean the funnel is healthy. It may mean the company is scaling winners. Or it may mean the company is still testing because nothing is sticking. This is also a quieter part of the year for ODDITY usually. Historically, they have tended to run lighter marketing from roughly July through December. That makes the current ad activity interesting. Maybe CPA is improving enough that they are pushing harder to recover lost first-half growth. Maybe they are trying to catch up after the disruption. Or maybe they are spending aggressively because the pixel still needs more data. The September earnings call should help separate those explanations. The numbers I care about most are: • Did CPA keep improving through June and July onwards? • Did first-order volume recover? • Are they still spending despite weak near-term returns? • Are repeat cohorts holding up after the first-half acquisition slowdown? • Did adjusted EBITDA turn positive? • Is MethodIQ still tracking toward its growth targets? • Any update or outlook on SpoiledChild? My base case is that the headline revenue number may still look ugly. The more important result is whether the customer acquisition engine is becoming predictable again. My position in $ODD remains small until that proof appears. The business may still have a strong platform, high gross margins and valuable customer data. But the advertising engine is the bridge between that potential and actual growth. That bridge is not fully repaired yet. TLDR / Summary: May showed the first real sign of CPA improvement. June and July look more like testing than confident scaling. Multiple pixels and domains may be part of the recovery, but rapid ad cycling and sloppy old creative suggest the problem might not be fully solved on the primary pixel / account. Next week should tell us whether this is pixel retraining before a recovery, or continued testing because scalable economics have not returned. They also look to be leaning more into Google, TikTok + Snap ads. So hard to judge where revenue/growth outlook will come in. Thanks for reading - C

$META may not have an AI revenue line for the hype but... if better models improve recommendations, engagement, ad relevance, and conversion, the return can appear inside the existing ad business. that makes the thesis harder to measure, but potentially more powerful opportunity. the proof is in impressions, pricing, time spent, and advertiser outcomes. For me, the optionality to monetize one of the largest distribution networks in the world from their AI Capex spend is currently extremely undervalued by the market at 17.7x Forward P/E. Same goes for $GOOGL (but trades at a higher multiple). Others with a large audience/distribution for a data fly-wheel apply here too, for example: $NOW, $DUOL and $HIMS. Are you buying $META at the moment?

Luis von Ahn: "It is now under $0.01 per Video Call." It started at $0.30. That's why a Max feature is now rolling to Super. This is super bullish longer term for personalised tutors. $DUOL

I pulled the Meta ads library on $ODD this week. IL Makiage main: ~230 ads live. oldest running since June 2. only ~70 inactive / turned off in 2+ months. ads surviving 30+ days are usually earning their keep. better ratio than I expected. IL Makiage Makeup page: ~180 ads, all started last month (July 15~), zero turned off. My take on new page/account: Bearish: main account hit a wall and they started fresh mid-July. Bullish: zero inactive after a month could mean early results are decent. bigger context: Q3/Q4 is usually their quiet period. they go lighter July to December most years. 400+ ads running during a slow window is unusual. (For Il Makiage only). either the ad cost is finally healthy enough to push year-round, or they're retraining the algorithm, or catching up from a weak first half. honest take: 230 with only 70 off from June to mid-August is a good ratio. ads surviving is a positive signal. new account is harder to read... pixel rebuild or forced workaround? no clear direction yet. keeping this small until the library ages more. bought a little more after selling $FVRR. $ODD

"aI wIlL dIsRuPt dUoLiNgO" Okay... when?

Popular FinTwit names & how far they are from 52W highs 🔴 $UBER: -22% $ADBE: -29% $NOW: -36% $SPCX: -39% $SOFI: -44% $RDDT: -46% $HIMS: -53% $APP: -57% $DUOL: -63% $FIG: -73% Which one is the best steal right now? 👀

Hers weight loss customers are sending 3x as many messages since a new AI care layer rolled out in July. AI is answering 80% of those messages. Non-clinical support tasks (order updates, side effect questions) are down roughly 50%. I've written before about the closed loop $HIMS is trying to build: intake, diagnosis, treatment, follow-up, outcome, all inside one platform with almost 3M subscribers feeding it. This is the first quarter that loop showed up as a real usage example. The CTO's words on the call: "we are making AI load bearing, not decorative." They're also weaning off third party AI customer support vendors entirely, aiming to "roll off them completely." Most companies rent AI. This is a bet on owning it. Love to see this vertical integration even for the AI. The line that stuck with me most: Dudum drew a direct contrast between $HIMS and "frontier AI companies," saying their advantage is being "an everyday partner to customers rather than just a chatbot that lives on their home screen." That's a direct answer to the fear that a general AI model eventually eats telehealth. There are some risks here. Healthcare data is sensitive, trust breaks fast, and none of this is a moat until retention and outcomes hold up at scale. We are still very early. Is closed-loop patient data an actual moat in healthcare, or does it erode the moment a bigger platform decides to compete here too? I believe it is, and $HIMS has the biggest advantage right now to capture this and build a massive data moat over competitors.

$DUOL Q2 earnings + call, my honest takeaways & summary: DAUs grew 23% YoY, up from ~21% in Q1. First sequential re-acceleration since the 2025 slowdown started. MAU growth accelerated too, +10% (glad to see this as it's been a concern). Management now expects DAU growth to stay above 20% for the rest of the year, not just guide to it. The number worth paying attention to: CURR (their retention metric) hit an all-time high of 84%. That's the leading indicator behind DAU growth, and it moved broadly across every region, not one market or one feature. Product is genuinely getting stickier. As intended with A/B and data flywheel. Streak Revival in June delivered. 15.4M learners restored a lost streak in June, nearly 8M had zero active streak at all. Management said this cohort is retaining better than a typical win-back group. That's the reactivation thesis I've had on this business for a while, now with real cohort data behind it. I suspect they will continue to try and win back old users as they have probably touched over 500M+ worldwide since they launched. Top of funnel improved almost everywhere too. Creators now drive roughly two-thirds of social impressions in China, Indonesia and India. Performance marketing tripled in channels and output over the past year. This isn't purely a word-of-mouth story anymore. This is good to see them try other options. Hope MAUs now continue to grow on this. What surprised me: Video Call went from ~$0.30 to under $0.01 per call, mostly from shifting to open-source models. That's why it's now rolling out to most new Super subscribers instead of staying locked behind Max. Regarding Max tier, its future is genuinely unclear. Luis said flat out "I don't know yet" whether it survives, gets a differentiated unlimited Video-Call role, or gets sunset entirely. They're also testing a cheaper, ad-supported "Super Lite" tier alongside it. Balance sheet: ~$1.42B cash and investments, zero debt, buyback still running through an "investment year." - strong position to be in. The one honest tension: bookings guidance wasn't raised despite the DAU beat. Freemium users take time to convert, and management is deliberately keeping bookings growth in an ~11% box this year regardless. Worth mentioning. Net net, I think this was a genuinely good quarter for the actual thesis, not just the narrative around it. I'd love to see Duolingo back near the $80s - 90s again so I can add more and lower my cost basis. Thesis is intact in my opinion, and flywheel is working as intended.

$NOW: the $1B security business nobody is pricing in. Fastest growing in the top-10 cybersecurity group. Armis + Veza already reshaping the TAM. And it sits inside a workflow platform with 98% renewal rates.

$AMZN is my 3rd largest position around 12%. Here’s what most people still miss when they ask why I size it this heavy: You’re not buying “the shopping website.” You’re buying: • The cloud platform that just printed 36.7% growth while expanding operating margin to 39.4% • An advertising business doing $19.8B a quarter, +26% YoY, sitting on some of the highest-intent traffic on the internet • A custom silicon franchise (Trainium + Graviton) already past a $25B run rate and growing triple digits • A direct equity stake in Anthropic that has generated tens of billions in mark-to-market gains this year alone • A satellite broadband network (Amazon Leo) and the world’s largest robotics fleet actually running the warehouses One stock. Five distinct growth engines. Most of them still underpriced relative to what they’ll be worth in five years. That’s why it’s not going anywhere.

$DUOL trading at an 8.4% FCF yield is crazy. 1. Massive data flywheel 2. Collapsing AI token costs 3. Highly recurring revenue Hard to find a better AI beneficiary at this valuation imo🤯