Coocoo

@paulg After many, many years of presuming good faith, I’m finding myself increasingly drawn to a Konmari curation policy: if the prospect of further interaction with a commentator does not spark joy, regardless of whether they are benign intent, block/mute and move on with life.

The second largest Chinese independent refinery Hengli was sanctioned by OFAC today. China's teapots with access to sanctioned barrels have actually been running harder compared to falling runrates at state owned refiners who have access to SPR but not really tapped on it. These teapots only serve the domestic market. It's a message from China's government to run mainly for internal consumption. Refined products from the state owned enterprises are technically banned and we see that particularly reflected on decreased mogas exports. Is this going to stop Iranian flows into China? Doubt it. For a holistic update of the rest of the oil markets in the past week, do tune in to our latest podcast! #oott

Traders: the supply shock is still underpriced. Our May APAC Chart Pack shows the Iran war forcing Asian refineries off their medium sour diet. June Goh's three areas to watch: > Crude diet pivot. Heavier Middle Eastern grades inaccessible. Lighter slate cuts diesel and fuel oil yields. > Diesel weak prompt. Major arb routes shut into Singapore. > Fuel oil call. Q3 0.5% cracks a buy. The 0.5% to 10ppm spread at -505 $/t looks too weak. Recovery: 3-6 months minimum if Hormuz reopens unconditionally. Full pack in comments below 👇 #oilmarkets #oott #commoditytrading #crudeoil #singapore

@UpslopeCapital God are they really gonna ruin Magnum ice cream

@DaBao_

The new AI Camera Assistant* with Xperia Intelligence brings stories to life. Using subject, scene and weather, it suggests expressive options with adjustments of colour, exposure, bokeh, and lens for breathtaking photos*. sony.co.jp/en/xperia-1m8/… #SonyXperia #Xperia1VIII

BREAKING: MiMo V2.5 Pro (Thinking) takes 3rd overall out of open weights models on Design Arena. MiMo V2.5 Pro (Thinking) places 8 positions higher than MiMo-V2.5 on the overall leaderboard, landing in the same performance band as Claude Sonnet 4.6 on frontend coding tasks. Huge congratulations to the @XiaomiMiMo team on these improvements!

Interview with Unemployed Capital Allocator (@atelicinvest) What makes you prefer managing your own money instead of having LPs? In Canada the regulatory environment for setting up your own fund is not as ideal. Even if you do it the very lightweight way, which is just opening up SMAs and managing money through Interactive Brokers for family and friends, you need to be registered. One guy I was talking to was spending a third of his time chasing clients for paperwork. This is not really something I’m interested in doing. The space I deal in is mostly small and micro-cap, so liquidity is quite tight, and sharing that wasn't that attractive either. At the end of it I said, screw this, I'm just going to do my own thing. How many software companies fit your universe globally? Hundreds. My sweet spot is 20 million to 500 million in market cap, global companies, primarily software. The smallest one I've bought was around 5-6 million market cap. You also have to kind of hive off Japan by itself, because Japan is such a deep market, especially in the small-cap space, that it probably makes up 40% plus of my watch-list companies. How did software become your focus? So software wasn't really in my prior firm's purview originally. It was very old-school, kind of high-quality cash flow. But for a long time software wasn't really high-quality cash flow - it was lumpy licensing revenue. Then SaaS came along and the accounting looked really weird and you're like, well, these aren't profitable… are they ever going to make any money? It wasn't until maybe 2016 that things started to click. Even in 2016 in Europe you could find software companies growing at high single digit to teens at 3 times EBIT, just because people didn't think of it as a high-quality business. At my previous firm, we were doing a bunch of other stuff - a lot of consumer staples, a lot of professional services, etc. And then one day we started asking ourselves, well, we're really looking for recurring revenues, these guys all have 95% plus of next year's revenue locked in. If it's recurring revenue you're looking for in simple businesses, this is as good as it gets. Since then it's mostly been software. My current portfolio is maybe 85% software. One of the nicest things about software companies is that the P&L is the truest expression of managerial decisions and talent. You have this money coming in, and all it is is what you decide to do with the cash flows coming in. I've seen this time and time again, where you've got a company growing revenue at mid single digits, they decide to blow their brains out over-investing in new product stuff, they do it the wrong way, margins go to zero, somebody else comes in, takes it over, cuts cost, and they're able to bring growth back up. It can actually be a pretty tricky dynamic in public markets, because all it takes is one guy at the top saying it's time to do X, and now you're beholden to that. The gap between what the company is able to generate and the actual cash flows that I see, can be quite a gap. So you've got to be a little careful. That's why my strategy is basically just buying companies close to valuations that imply they're literally going to die within the next two to three years. My entire bet is around, they're not going to die. That's been my only playbook since 2016, where you buy these at half times to one and a half times ARR that are still growing, and for whatever reason people just hate them. Some of them do end up actually dying. The managerial talent for screwing things up is so good. I had one that literally spent themselves close to bankruptcy, still got taken out at two times revenue, because the acquirer was like, we can fire pretty much 90% of staff and still keep it around. The yield is really good at two times ARR. The playbook for me is: find products that people need in their day-to-day work, make sure the value proposition is solid, and just buy it at really cheap prices, and just buy 30 of these and see where they go. Why is there low demand from other investors? What makes them price these as if they're going to die? It's an interesting mix. In the beginning, so 2016ish, it was just pure ignorance. People weren't really looking at software companies, and if they were, they were only buying SAP and Dassault and these large-cap companies. There's also an overestimation of risk involved with smaller companies, especially in software. These are really dinky little operations. If you went to talk to the CFO, you'd be walking out shaking your head, because they don't have proper internal controls and other operational systems you’d expect. So the quality-risk element is something you apply categorically to all small businesses. Anything below 500 million market cap, if you go in there, it's a mess. But software is probably the one place where you can get away with a lot of sloppiness, just because the checks keep coming in. And at least historically there's been a put, because there's so many buyers for the assets at every level. You've got strategic buyers, you got PEs coming in, you've got Constellation copycats willing to buy at rock-bottom prices. You rarely see anything get taken out at one times ARR - even for catastrophically bad software. So there is a bit of a mismatch between the perceived risk and the actual risk you end up taking. Same goes with management quality. There are a lot of bad managers, but there's a limit to what they can do, because they still want to keep their jobs. The biggest risk is they dilute you until the cows come home. Is your view that organizational transformation is hard, software is just one small piece of internal operations, and people are overestimating the bear case because they see how easy it is for them to vibe-code the first 90%? That's part of it. It's gotten a bit more complicated. In the beginning I was definitely in the camp of making fun of all these guys with software bear theses, because it is pretty ridiculous to sit and think that you could recreate a lot of software functionality, because true rip-and-replace, unless the job itself is being replaced, is extremely rare. Unless you're in a decline and you really need to cut costs, or in a massive recession in which case you don't have enough time to do a rip-and-replace, you don't tend to sit there and go, we have this new capability, all these new things, but let's replace this $10,000 software that we're spending money on. There is a certain amount of risk involved as well. It's this thing of, don't change anything that works right now. This is why software is such a mess at most companies. You go in and wonder: why is it like this? It's because it started from this kernel and they started adding new things and they had to come up with workarounds and workarounds, and eventually it becomes this tangled mess. Obviously people would love to rip it out and simplify, but operations are running as it is, and there's such a big risk element changing parts of it, without redoing the full stack, that I do think it's going to be relatively rare. At the same time, the trouble now with talking about software is that it's so many different things, right. It's such a big and wide industry that spans across so many different activities that it's hard to say anything definitive as a whole category. I think pretty much everything anybody has ever said about software is true at this point. What's happening to the discourse right now is just, people are seeing different dynamics happen in their own little corner, applying that to the whole thing, and then saying, this is going to happen to everything. Well, it's not. It's very specific to your specific corner. The only bet I'm actually willing to make at this point is the same one I've been making this whole time: these companies in my portfolio aren't going to die in the next three years, we're not going to see churn skyrocket, and based on that I think I'm going to do okay. That said, there are certain software companies that are going to see budget pressures and growth pressures coming from money going to tokens. There are going to be new competitors coming in. It is easier to build software if you're a domain expert. If you're in a domain with 50,000 people that are world class, can 50 of them build really cool products that take market share? 100%. Could existing software executives make really dumb decisions, either on their own software development, or on competitive decisions, or acquisitions? 100%. So there's a lot more risk today versus where it used to be. But it's reflected in a lot of share prices as well. So people are using the low-resolution word "software" to talk about what actually needs to be a bunch of much more specific, high-resolution pieces. It's like saying "startups", when it should actually be: this category, this type of buyer, this type of product, etc. Yes, a hundred percent. In software, you almost need to get down to the exact organization and the exact customer set they have and what they use them for, because so much of the dynamic depends on the exact organizational capabilities. I've actually found it very surprising how hard a time some software companies have had adopting AI into their development workflow. It's not the fault of the company or the software engineers. It's just, in a lot of cases, going fast is very much against the entire existence of what they've been doing, which is creating stable software that clients who don't know anything about technology can use without worrying about, “is it going to blow up tomorrow?”. And then here you have people saying “Now one person can do the work of 50 people!”, and you're like, “well, if that means there's a 1% chance of something blowing up every day, I can't do anything with that.” Is there one you could talk through as an example? Let me talk through a product that kind of seems really flimsy but might have longer legs than you might think. There's a company in Japan and it's effectively a caller ID + call-monitoring software. It's telephone-based sales but very specifically to do with multi-location businesses. So if you're an auto repair shop and you have six different locations, and multiple team members at each location, and they all need to keep track of conversations that need to be happening. It's a super competitive space, with probably hundreds of competitors out there. The functionality itself is simple. I'm pretty sure somebody could vibe-code this inside of a week, no problem. There is some complication with actually having a box on location to connect to the phone line. But it's not complicated software at all. It's also kind of an area where pretty much all of the publicly traded companies in this space, call center or call-related stuff, it's all going away, because supposedly everything's moving to agents and the whole stack's going to change. So it's this dual thing of the entire industry is moving to agent-based commerce, and on top of that the software itself is super easy to replicate. The key thing is, there are tons of different niches within the space. They’re also not very tech savvy - think about salons, auto repair shops and realtors. If they want agent features, they want it cheap, they want it fast, and they want it to integrate with whatever system they're working with, because they're not going to fire all their people and just have an automated answering system. That's what the majority of the real-world economy is. You've got customers calling in that are real people outside of Silicon Valley that just want to get stuff done quickly. Your business has to just operate, and frankly a big change is a huge risk and pain for a lot of these companies that don't have the resources to do it. They're certainly not going to get a forward-deployed engineer from Anthropic coming in and trying to set it up for them. Another one might be construction software. An industry famous for not even adopting SaaS. I think SaaS is just starting to take off in that space. Another would be a frontline-worker communication tool - employees that don't even have email addresses, that are out in the field doing work, with an app on the phone to communicate. How do you actually find these companies? I try to be exhaustive. Even before AI, my workflow was, get the basic information down. Historically, most of the times when I've made money, it's because I was aware of something, and then I see a price movement happening and notice, oh, that's weird, why is it trading that way? And maybe a few percent of them end up being a mispriced situation. So this base layer of knowledge bank becomes really important. My watch list used to be a few thousand companies long. You would do the basic workup on them, super basic stuff, but things that would still take you maybe about 30 minutes to two hours to figure out, pre-AI. Then keep on top of news flow, and either the price movement or valuation would look really extreme on the downside. So it'd be, well, why is this company trading at 1 times ARR? That doesn't sound right. Then you look into it, or the news flow itself would look really interesting. Take a look, and out of that the trade idea would fall out. So you look for companies that are falling in price unexpectedly? Yeah. I mean honestly, I probably would have made more money if I did it the other way, just given how momentum goes. But I can't seem to do that reliably. My foray into AI was as soon as ChatGPT came out, realizing it will be able to automate pretty much 95% of my discovery process. That's the thing I've been working on for the last three years. This kind of painstaking and time-consuming process of creating this knowledge bank, but now it's mostly automated. So now the challenge is to try to put it all together. As a product for other people to buy, or just internally for your investment process? For other people to buy. So far it's been very lightweight. I would make a web app for myself to do certain things and just let a few friends try it. I started with kind of regulatory filing access, and surprisingly, this is not a solved problem. One of my big theses, which isn't that controversial by any means, is that investment research, because of how AI is getting integrated, the volume of work you're going to do as an investor or analyst is going to go up significantly. So historically, a lot of people would maybe do 5, 10 deep dives a year. Maybe they'll look at a few companies here and there on the side and keep on top of news. If you had a tracking list of 50 companies, that was considered a lot. My watch list of 1,000 companies plus was considered suicidal. Obviously that's going to change a lot, because all the time-consuming stuff is being automated. As a result, everything around the research process needs to change. Historically, because you're doing so few a year, if you need a regulatory filing, you should just go and look it up yourself, download it, print it, all this stuff. That time now is such a big cost as the cycle time speeds up. And all the tools that were built around this, like Cap IQ and Bloomberg, still don't fully recognize that this is the case. People need really good ergonomic human-plus-agent-centric solutions, and even the most basic EDGAR access and feed generation is still not optimized. Some regulators like SEDAR in Canada put up a bunch of gates. They actually spent a few million bucks upgrading their site, and the entire site's basically broken. What's the timeline look like on offering the product? Honestly, I don't know. The biggest issue is actually putting together the right set of products. Some regulatory filing access stuff, I already have out there. Japan-specific stuff, SEDAR, EDGAR, I've got a product out there. But these are point solutions that do very specific things. Historically I always thought, okay, maybe I'll do this one holistic platform. Recently I've started to rethink this, because one of the hardest things to do in finance is getting investors to take time out of their day to learn a new platform that has a bunch of new buttons and layouts they're not used to. Forget getting them to pay for it. I've had people screaming out for a solution who can't take 15 minutes out of their day. I get it. I used to be one of those people too. So now I'm rethinking the product set as a bunch of point solutions. If you have a very specific need, e.g. I need to create a really good feed of important filings out of EDGAR on this set of tickers, great, you can do that in five seconds. I've got a solution for you. Or, oh, I need to create an email alert off of this very specific thing, great, I've got a solution. Favorite Twitter accounts, and any other communities or places you like for investing info? Big fan of @turtlebay_io on Twitter. I haven't seen him post in a while. He's the well-known one, when he used to post historical stuff on companies and investors and Buffett. He's got a database that is probably to die for. There are a lot of good 200-follower accounts that do very specific work on a name or two. Historically the best ones tend to do the work on one or two names and then release it in really good posts. So it's more of a post-based thing versus an account-based thing. On other communities, honestly I don't really frequent that many. The pull of everybody on Twitter, you just can't be elsewhere.