Matt Humbaugh

@WaterworldCapi1 The accounting is fun too.

$BE at a $90 billion valuation for $3.7 billion in basically profitless revenue is a sight to see.

@partners_road Can you share the link?

MS “fine-tuned” the open-source Llama 3.3 70B model to extract sentiment signals from Russell 3000 earnings call transcripts. Backtesting results show that the QoQ sentiment change factor achieved a Sharpe ratio of 2.15 from 2011 to 2025, with a max drawdown of -3.04%.

@S_Kadakia @cyera_io @AnthropicAI @HeyEpic Qualitate

To celebrate our launch, we're dropping our 60-pg report on who's winning & losing in enterprise tech in 2026. Forward-looking buyer signals on $CRWD, @cyera_io, $TOST, $CRM, @AnthropicAI, @HeyEpic, and many more. Comment "Qualitate" on the launch vid and i'll send it to you.

Introducing Qualitate. AI built for primary intelligence.  Qualitate conducts thousands of expert discussions each month, delivering structured intelligence for the world’s leading investment firms and enterprises. Here's how:

@akramsrazor Think about this all the time when folks talk politics.

There is nothing admirable about demonstrating a fundamental lack of empathy. You are a citizen of the planet before u are anything else. Start with that and work backwards.

@FundamentEdge The context document must be amazing to deliver this level of quality output.

Agentic Management Deep Dives In adopting agentic approaches for investment research, one thing investors should be very careful of is sacrificing rigor for speed. In many cases this can mean producing more documents to read, but those documents being much higher signal to noise ratio. To me the objective of overlaying agentic approaches is to clearly deepen rigor by spinning up agentic system to systematically conduct analyses that would never get to the top of my to-do list. These can be simple things like a systematic guidance credibility tracker across a coverage area. I'll give you 19 more below when it comes to evaluating a CEO, below. In this sense it's remarkable how powerful agents have become. This is not an ad, and I am not sponsored, but the multi-model user-friendly interface at Perplexity Computer, feels like the future to me. All the technical complexity is abstracted away and it just works. Incredibly well. I don't get stuck in approval hell like I do with the coding agents, and everything operates in a virtual machine so it doesn't have read-write access to my drive. Don't take my word for it. If you want to see the outputs that are possible now, I'm attaching a 17-page report on the BSX CEO. Having covered this stock for nearly 15 years, I know Mike well. I was blown away by how comprehensive and rigorous this report was. And I assume it will get even better as more of these analyses can be done as we find ways to plug in the right data. Enjoy

Let me explain exactly why every new subdivision in America looks like the top photo, because the math is wild. A mature tree increases a home's value by 7 to 19 percent. On a $400,000 house, that's $28,000 to $76,000. A single shade tree produces the cooling equivalent of ten room-size air conditioners running 20 hours a day. One tree on the west side of a house cuts energy bills by 12 percent within 15 years. The bottom photo is worth more, costs less to live in, and sells faster. This has been documented by the University of Washington, Clemson, Michigan State, and the USDA. The data is not in dispute. Removing those trees saves the builder roughly $5,000 per lot. Concrete trucks need twice the dripline radius of every standing tree. Utility trenches need flat ground. A bulldozer flattens 200 lots in an afternoon. Preserving trees adds weeks and thousands per home. So the developer pockets $5,000 in savings and the buyer eats $50,000 in lost value for the next two decades. The person making the decision and the person paying for it have never been in the same room. The Woodlands, Texas is the proof of what happens when they are. George Mitchell bought 28,000 acres of Houston timberland in 1974 and preserved 28% as permanent green space. He forced McDonald's to build behind the tree canopy. That McDonald's became one of the highest-volume locations in Texas. The first office building, designed to reflect the surrounding forest so you couldn't see it from the street, leased completely. The Woodlands median home price today: $615,000. Katy, a comparable Houston suburb that clear-cut: $375,000. Named #1 community to live in America two years running. Fifty years of data. The trees are worth more than removing them saves. Developers clear-cut anyway because they sell the house once and leave. You live in it for 30 years.

@TechFundies I cringe every time someone starts a sentence with “Respectfully” because you know what’s coming is anything but. :)

Went through $XYZ to rubberneck. I've followed this company relatively closely since it came public as well as $TWTR so generally have a good feel for what's going on. Jack is an incredible product CEO. Amazing hits with Twitter, Square point-of-sale, and CashApp. Truly a product genius. That being said, his ability as a business operator has been consistently disappointing. $TWTR never evolved past group discourse and was just monetized as cheap ad tonnage. It was an incredibly bloated company that meandered until it was bought. $SQ (original $XYZ) had a great low-end point-of-sale system sold online. However, the company failed to make meaningful moves up-market largely due to the strategic decision to not have sales reps which opened the door for Clover and $TOST to seal off that opportunity. Then they completely missed most of the online opportunity which $SHOP dominated. They literally only started hiring field sales within the last year which is like showing up to a party a decade late. $SQ CashApp did great for peer-to-peer payments but the product literally did not evolve for a decade. It added bitcoin and stocks, and eventually a debit card. Despite having the initial lead, they lost P2P mindshare to Venmo / Zelle, lost asset trading to HOOD / COIN, lost BNPL to AFRM, APT, KLAR, etc. They lost the ability to build a closed financial network to Stripe / V / MA (though never really had a chance given their assets). Then $XYZ purchased APT which was orthogonal at best at a massive price and distracted the business. It took years for it to be integrated and still isn't there. Then $XYZ bought Tidal to have Jay-Z on the board. Most recently, the company is still investing in proprietary Bitcoin wallets / servers, and focused on "neighborhoods" whatever that means. The internal organization of the company has been adjusted for several years now and the presented logic hasn't ever really made much sense. First they were separate BUs, then some shared resources, and now who knows how they intend to structure the business. So tonight we learn they are laying off 40% of the company. The cited reason is AI improvements to coding. They spend 3b / year on R&D! Even if fully-loaded cost is $500k / year that is 6,000 engineers. Doing what, I have no clue for a set of products that haven't evolved much at all. They also spend 2b / year on G&A which I'm guess has a lot of fat in it as well. Long way of saying I'm pretty sure this has more to do w/ the company just being far too bloated from all the mistakes made above, and not AI advancements scorching the earth. Respectfully, the CEO splits his time between SF and Costa Rica (apparently mostly remote in Costa Rica these days), spends 2 hours a day meditating when not disappearing for 10 day retreats, etc. That's all pretty awesome and I'm honestly happy for him. Also he controls the vote at $XYZ so he can do whatever he wants. That being said, I wouldn't hire him to run a business (definitely would hire him to be chief product officer or whatever). And I don't think this 40% RIF really has any broader implications. I don't know if their tech stack is mostly self-built or SaaS, but I would think their spending on cloud AI is certainly going up (Anthropic, hyperscalers), their spend on payroll is going down, and their spend on CRM / customer support sw is going up.

@JohnHolbein1 This is such a fabulous dinner party question. Also raises the Q of how much time we need before we can determine a policy has truly backfired?

What policy decision most spectacularly backfired in your lifetime?

In Kyoto and just saw a lady rinse down a pole with her water bottle after her dog peed on it. The level of respect in Japan is off the charts.

I've spent years struggling to prove that the sunk-cost fallacy isn't actually a fallacy. No sense in giving up now, though.

Excited for the 2026 Diddy Bowl tonight

@TechFundies Got any ETFs trading BELOW their LT multiple?

Also, if you want to buy the S&P500, buy $VOO which has lower fees than $SPY. Same thing, better performance.

Happy New Year. My occasional reminder that the best gift you can give yourself and everyone around you is the awareness of index ETFs. Seriously, I don't mean in this in a promotional way but read this tweet. 99.999% of investors cannot beat these incredible ETFs that provide exposure to the best companies in the strongest economy in the world, automatically rebalance to add winners / remove losers, can be held forever, and require you to pay little to no taxes. Also you can leave them to your kids / grandkids and they will not owe any taxes on embedded gains. Numbers updated for 2025. SPY: 17.6% in 2025. 13.8% compound return over 15 years. Trades at 23x forward P/E vs 10 year mean of 20x. EPS has grown at ~9% CAGR over the past 12 years so could also think of it as maybe SPY moves sideways for a year to "grow into its valuation". QQQ (Nasdaq 100): 20.7% in 2025. 18.5% compound return over 15 years. Trades at 26x forward P/E vs 10 year mean of 24x. SMH (Semi's): 49.2% in 2025. 24.8% compound return over 15 years. 53.2% compound return over 3 years during the super-cycle! Trades at 24x forward P/E vs 10 year mean of 19x. 2026 EPS is obviously going to be great w/ market figuring out 2027 / 2028 and what multiple is appropriate. IGV (Software): 16.0% compound return over 15 years. Trades at 26x forward P/E vs 10 year mean of 35x and down from last year forward P/E of 36. Even the dog has returned a solid return to date! I think everyone (kids, adults, endowments, pensions, institutions, ...) should have serious exposure here. It's just virtually the best option for everyone sitting in plain sight. [Also, if you think that the only way out of the US debt situation is via money-printing (which has been the answer for every other "empire" throughout history, then an effective way to protect yourself against that downside is via owning stocks / real estate. If minimum wage at McDonald's goes to $100 / hour and the cost of a hamburger goes to $200, then the company's EPS will also go up massively. This is why stocks move higher over time after devaluation events (like when the US abandoned the gold standard).] So get started on Monday. Set a plan. Put 0.5% of your savings into ETFs on the first of each month so you can dollar-cost average. Hold them forever. Best financial option for vast majority of people - but never recommended by the "pros" bc it generates no fees for them. Simple as that.

@TechFundies At what multiple would they NOT be buys?

3P contact works at JP Morgan on AI adoption / enablement for 60k engineers. Described a rising tide for all three hyper-scalers (AMZN, MSFT, GOOGL). They have set themselves up to buy all models via an agnostic platform so they can switch providers easily. Positive on Anthropic capabilities in coding, GOOGL advances in context windows / multi-modal. Open.AI still the majority provider but losing share. Just seems straightforward to own all three hyper-scalers as every AI inference trend is durable and flows through them. Highlights -AWS is by far biggest cloud for them bc got started there and have invested a lot in tooling / security. AMZN has 2-3 tech folks in each of our tech centers helping engineers grow w/ AWS. Want to move mainframe workloads but AMZN has missed promises as far as capabilities for years. -Azure: Mostly use this for MSFT apps as have had issues w/ non-MSFT tooling which is lackluster. -GOOGL: Trying to use more of their services – devs only have so much capacity to learn new platforms. Adoption driven by their context window length and multimodal capabilities. Were big customers of Windsurf – leadership moved to GOOGL and so will do more w/ Antigravity. Gemini context window is 2m vs MSFT GTHB is 128k – so very difficult to port an old program w/ 2m lines of code w/ MSFT bc breaking it into chunks reduces output quality to 50%. Also working on app where wealth advisor conversations are recorded and turned into action plans by AI. -ORCL: Just very late to the game with only a decent product. Our devs still not buying into them and even their HR app performance is subpar [positive for $WDAY]. -Models: Largest vendor is Open.AI purchased through AMZN Bedrock / Azure but have seen big uptick in Anthropic adoption as developers love it / quality is much better. Gemini models doing very well w/ document processing. Open.AI 5.1 model latency has been horrible. Opus is so expensive but great. Have built FinOps dashboard to closely monitor agents, model usage, cost to make sure model usage is efficient. -Do want majority of workloads to go through AMZN Bedrock or MSFT bc have implemented security / compliance guardrails there. Want monitoring control so low-value AI is forced to use older, lower-cost models. -A lot of AI usage in devs has been towards reducing tech debt, improving quality of programs but not yet necessarily on the pure innovation side.

We published this today. Phenomenal analytical work to understand how things could play out. give us 1,000 retweets and 5,000 likes, and we make it publicly available and do a space to discuss findings first week of Jan.

This man is going to buy the entire internet champagne. Let’s make it happen!

@modestproposal1 It was a special time. I miss it. Reality is that no matter how many teams are included someone is always left out. Just sometimes it was the #2 team in the country. New Years Day used to be amazing.

I was describing the old system to my son and it just sounds so amazing when you put it in words. Wait, there were two polls? Yes. And #1 and #2 almost never played each other? Exactly. And you liked this? It was incredible, 12 hours, 10 games, total chaos.

You will never convince me it wasn’t better when every good team played on the same day and then a bunch of overweight drunks voted for who was champ

If this 4% drawdown felt like 20%, the next 10% is going to feel like GFC 😂

@FishtownCap @evantindell Awesome! Thanks for your thoughtful and unemotional response.

@MattHumbaugh @evantindell The numbers you want are in that study. Maybe read it? Compare Figures 7 and 8 specifically for the horde of uneducated immigrants storming over the southern border.

"facts don't care about your feelings", right? 0.60% of Medicaid dollars are spent on "Emergency Medicaid" in NY, which is the only portion undocumented immigrants have access to.

@FishtownCap @evantindell Good point! I think you’re right. Is there a good study you would point me to that suggests the “net detractor” argument?

@MattHumbaugh @evantindell lol. You didn’t read the study, you just googled the conclusion you thought you wanted. This study is all immigrants, not illegals, which are overwhelmingly without degrees.

@MidwestHedgie What’s on the list today?

It feels like these three slides are worth sharing again: