Opus 5.5 is pretty insane for Excel read/write I have a few dozen models that are 4-8 quarters out of date. Prior LLMs had gotten better at simple, single quarter updates, but were still sketchy with multi-quarter refreshes (and failed on from scratch builds, to a typical hedge fund standard). Opus is really the first model that can one-shot my Refresh Modeling Skill Pack in a way that passes my Turing test (i.e. I couldn't tell if this was updated by a human or an agent) and passes final validation in a way that I find reliable Exciting progress.

Please join me Thursday to walk through a live webinar: Building a Credit AI Workflow. A common & recurring challenge in enabling AI for fundamental investors is the blank page problem. In this webinar, I will walk through my approach of taking Fundamental Edge Credit Academy Instructor Alex Goston's raw curriculum and turning these workflows into a pre-configured dashboard powered by Agent Skills. Our view at Fundamental Edge is that the "hacker era" of AI & investing is over, and AI has reached sufficient utility that a combination of pre-configuration and custom curation can help make your investment process AI forward in a matter of weeks not months. Registration link in the thread below.

I laughed at loud at Gabe explaining low reception to the Series A, saying "they didn't invest because the product was terrible". I first got my hands on Rogo in fall of 2024, and can confirm: the product was terrible. The remarkable thing about Gabe and the Rogo team has been the accelerated pace of improvement. Fast forward two years, I had a moment this summer where I ran a number of my investment research workflows through Rogo and ran evals vs. my Claude system and Rogo routinely won in head to heads. Rogo's Felix is this sort of malleable agentic substrate that has become flexible to deploy and delightful to use. And importantly, I'm too obtuse to predict the future but I've learned to trace the curve of the exponential. I am super excited to see what this team builds for investors, and think the next 6-18 months for investors deploying AI is going to get very exciting.

This experience will only spread, in my view. The historical best practice to get smart on a name was read the filings, read the transcripts, and read a big stack of sell-side research, then model out the company. This hadn't really changed much in decades (outside of innovations in alternative data & expert network transcripts). Having a comprehensive offering of sell-side research was important at the institutional level. We have reached threshold on numerous fronts where a public market investor can achieve similar or deeper comprehension on a name with AI, and doesn't necessarily need that same comprehensive sell-side offering. That random sell-side report that went deep on a certain aspect of the business or industry can now be created with an AI agent sitting on the right data pipeline. One example I've shown in the past on DKNG...in the past, if I'm trying to get smarter & sharper on a deep dive on state level taxation, the right sell-side note dropping at the right time is supremely helpful. Now, I can run that analysis when I need it at the push of a button. The cohort of investors who build off of sell-side models will, very soon, be at push-button AI capabilities (and more may move modeling off Excel into JSON). The moat of the sell-side is melting. And I believe the sell-side has, collectively, overplayed their hand in being adversarial to the agentic path. My view is they will eventually fold, but not before many clients learn to build around the commercial friction and, maybe, eventually come to the same conclusion as Just Another Pod Guy. Like most things the top decile sell-side analysts will be fine, decades of investor trust and relationships will continue to monetize. But what happens to the 16th best analyst on a name? It think it's obvious the industry just needs fewer voices on a name, so how do you pivot? Corporate access has enduring value (if you don't believe it, be a fly on the wall when there is one seat at a key meeting and 5 pods wanting that same seat...). But I think it's deeper than that...how does the sell-side drive differentiated client insight? Not just regurgitate publicly available information (never much value, and now zero value). Cleveland Research to me is the working mental model of deep embedding of their analysts into the operational flow of industries driving a regular & valuable flow of investible insights.

I’ve commented on other people posts. I’ll synthesize what I believe. Fundamental risks are not the only risks but they are the most important risks unless you have a lot of leverage in which case, risk is risk and trying to pretend one matters and one doesn’t always ends the same way. If you have no leverage the only thing that can force permanent capital loss is your or your LPs tolerance for pain. If you have leverage your lender can trigger permanent capital loss. If you work at a platform you are being lent callable money, not managing capital, and you should manage money like the bank can take it back whenever they want. The best way to manage non fundamental risks is through a rigorous fundamental force ranking process that compares similar ideas to one another and limits exposure to similar things. Most fundamental research is very good but doesn’t go far enough in demanding absolute and relative return thresholds or in creatively comparing ideas to one another. There are generally high correlations between fundamental and non fundamental risks. If you limit your exposure to a fundamental risk (like no current profits) you will manage the non fundamental risk as well and naturally de risk the exposure and improve effective breadth. You can reduce the number of ideas which increases concentration and results in less non fundamental risk through the same process. You can reduce non fundamental risk through a proper fundamental process and outperform.

I think the small, concentrated, high-conviction fund thesis is increasingly hard to square with what has actually happened in the industry. LP dollars have been moving toward passive, systematic, market-neutral and multi-manager structures while traditional LO/Tiger/mid-net L/S has steadily lost share. More importantly, the success of the pod model has already shown that fundamental investing is not some intractible problem. You can break the process into forecasts, exposures, risk budgets, sizing rules while reducing many of the behavioral and portfolio construction problems inherent to the traditional Tiger model. Acadian's work ( supports these same findings: concentrated portfolios have not produced consistent alpha, while more systematic processes can scale through greater breadth. If quarterly fundamental alpha can be industrialized this way, I don't see why longer-horizon fundamental alpha should remain permanently exempt, particularly when prop firms are spending billions on compute, data, domain-specific transformer architectures while continuing to push their forecast horizons outward. The chart below makes the concentrated portfolio argument even harder because the market has become materially more factor driven through most of recent history. A 10-20 name portfolio can therefore very easily amount to a few large latent factor bets with idiosyncratic noise layered on top. This is also where I think the idea that the answer is a smaller fund with more conviction starts to break down. If the data show that concentrated fundamental portfolios are not producing durable excess returns in an increasingly factor-driven environment, then the natural advantage shifts toward firms capable of extracting many smaller signals, controlling their common exposures and combining them systematically rather than relying on a PM's ability to identify 15 exceptional stocks. I also strongly disagree that fixing this is as trivial as onboarding Arcana and building an automated factor hedging layer. The software interface may be easy; building a risk system that is actually useful in production is not. Off-the-shelf models have materially underpredicted realized volatility in 2026, which pushes you into needing custom factors and covariance estimation and the much harder question of whether the factor structure you are using in your off the shelf model is even correct. Axioma itself distinguishes estimation error from specification error and shows how basic choices around estimation window, frequency, weighting, outliers and autocorrelation materially change the resulting exposures. Then you still need to turn those forecasts and risk estimates into positions. Portfolio optimization is a separate technical discipline involving constraints, turnover, estimation error, transaction costs and the interaction between all of them. None of that is trivial, particularly inside concentrated fundamental funds where many PMs remain skeptical of factor models in the first place. The capacity problem compounds this. Passive and pod capital are heavily concentrated in liquid large caps, so the obvious place for a small fundamental fund to look for less competed longer-horizon alpha is further down the capitalization and liquidity spectrum. But a 10-20 stock portfolio deploying meaningful capital into small/mid/micro caps is going to move price against itself both entering and exiting positions. At that point market impact and optimal execution are part of the alpha model rather than implementation details because they determine how much of the forecast survives into realized PnL. MOSEK treats transaction costs, market impact and portfolio constraints as explicit parts of the portfolio optimization problem for exactly this reason. So the concentrated fundamental shop ultimately ends up needing custom risk models, portfolio optimization, impact modeling and execution research anyway. Which are precisely the capabilities where the large prop firms have spent years building an industrial advantage, which is why I think the more likely endpoint is that they continue moving outward in horizon and subsume more of the longer-horizon fundamental alpha rather than leaving it permanently protected for small discretionary funds. Just my take though, would be interested in @TheStalwart @tracyalloway @__paleologo to opine as well

Thank you, @FundamentEdge for hosting our CEO and Founder, Jack Kokko, for a packed fireside chat last night on SuperAnalyst and what's actually working in AI-driven research and portfolio management. A quick thread on what he said 🧵

I have been of the view that the exact opposite will happen. @pmje73 recently tweeted a view that he believes concentration will tick back up, and I couldn’t agree more. Experienced fundamental investors will know the degree to which your tier 2 and tier 3 ideas act as a tax on your best 10-15 best ideas. How good is your 67th idea? Generally pretty crappy. Yet the idio-based risk models have demanded breadth for risk management purposes, so many stock pickers have been forced to play that game. I lived this personally going from a seat where I ran 6x12 (and consistently generated double-digit spread) to seats where I had to run 30x70 to make the risk model happy (and could never figure out how to scale research depth needed to have alpha in my 67th best idea). Strip out a talented stock picker and just have them focus on one thing: give me your 15 best ideas. I can’t logic through how that approach to portfolio construction doesn’t get meaningfully more compelling given the evolution of tool set and current state of equity markets: 1) the ability to build scaled research systems distilling then tracking the 15 best ideas from 8,000 stocks has been transformed with AI…a single stock picker / PM can truly now have close to institutional-grade research intelligence, which was never the case before 2) historical periods of index concentration have been compelling starting points for bottom’s up stock picking / and with more asset classes facing high valuations and low forward return prospects, the ability to find 15 great equity ideas that can CAGR 15%+ for 15 years is a more compelling value proposition to LPs on a relative basis. There are 26.6 quatturodecillion (!) 15 stock portfolios I could construct from 8,000 underlying global equities, and, definitionally, roughly half of those will outperform a broad global index. AI can’t and won’t change this fundamental reality of dispersion (will probably make overshoots and factor/thematic violence worse, a benefit to the longer duration investor). People will do this in regional banks, Nordic small caps and thousands of other approaches that aren’t necessarily owning AI winners. There’s always a bull market somewhere. 3) Agree with @0xFaust12 point on risk systems, but the friction to onboard a tool like Arcana and build an automated AI factor hedging system is now fairly trivial and with relatively little effort can take that 15 stock portfolio that probably looks like a giant factor bet into a tapered idio-primary portfolio. The bottleneck is more awareness than technology. There are lots of reasons there hasn’t been this explosion of talented stock pickers starting $100-$350m 15 stock concentrated long biased funds (coming from a guy who thought about building one in 2018, i can list all the reasons in detail). Specifically, many of the most talented stock pickers have had a durable bid from the multi-complex with compelling guarantee terms. To me, what the Jane Streets (and many others) will be able to do is to start industrialize the broad alpha of the discretionary multi-manager. The same way that quant firms compressed alpha signal in alternative data, quant firms will compress alpha signal currently harvested by swarms of pod teams. Except for the fact that the large multi-managers are run by the sharpest, most adaptive people in our economy, to me the “peak pod” argument becomes more real over the next 3-10 years. And, multi-teams will be able to do with 8 investors what they historically did with 12. What do the 4 who attrit do? @chamath had this idea a few years back where he gave small portfolios to single analysts/PMs. It was a super intriguing idea but the time and tooling wasn’t ripe. I think it’s ripe now. Probably some centralization makes sense (build the AI and data pipelines centrally), whether it’s sponsored by a quant firm with a buy-side or by an LP directly. Absorb the displaced multi-manager talent and just ask them for 15 best ideas

Over the past few years, I have tried a number of things to encourage my kids (12, 10 & 8) to become AI native, including downloading Gemini, Claude & ChatGPT to their devices and walking them through how to use it. My 12 year old will occasionally ask ChatGPT a question, but nothing has really stuck. This weekend, I got my 12 & 10 year old set up with Instinct. > They can now turn the hot tub on & off with a text > Created an "Ebay sniping" approach to have Instinct pull auctions on sports cards ending in the next 60 minutes priced meaningfully under comps > Pulled from ParentSquare to know which days they have to wear sneakers for PE (apparently wearing sneakers when they could wear Crocs is like corporal punishment for a 10 year old). "Jarvis, do i have to wear Crocs today?" > Connected Google Calendar so they can see their practice & game schedules for the week But the big winner: > Connected their Instacart family account so they can order what they want ($100 weekly limit) from Instacart anytime, with a text My 12 year old ordered 4 bags Tru Fruit from Instacart and I could see the "aha" moment in his eyes. "AI is amazing...I can't believe this, Dad"

As an ambitious former hedge fund PM, now current entrepreneur, it's always been hard for me to turn my mind off. I got to a point maybe 10 years ago where I could really never relax, and it got to be too much. Even on the weekend, with my young kids, my mind was active with work problems and it was hard for me to be truly present to the moment. Not only is this a hard way to live, but I found myself showing up Monday morning intellectually fried. I eventually wore myself out so severely that I needed many months away from work to feel "normal" again. Maybe you feel this way. If you do, ahead of this holiday weekend, I wanted to two few rituals I've learned along the way that helped me. Ritual 1: Empty Your Head on Paper. Write for 3-5 minutes (pen and paper always best) about all the work-related problems still bouncing around in your head. People to call, ideas, problems, things you're worried about. Get it out of your head, onto paper. Even better, ask whatever higher entity feels personal to you to help you untie these knots over the weekend ("dear God/guides/Universe, here are a few things I am struggling with, please help me on these over the next 48 hours"). It's woo-woo, but it works, and I instantly feel better, like I've done something crafty but temporarily assigning my struggles to a higher, wiser entity. Those "aha" moments have a way of finding themselves into my brain come Sunday afternoon, and this sort of work can be just as or even more powerful than more hours behind a computer screen. Ritual 2: Self Talk, "I Did Enough This Week": Look at yourself in the mirror, and tell yourself, "you did enough this week...I did enough this week". Celebrate the progress you've made and what you completed, get into gratitude and thank yourself for all the progress you made. This shifts the anxiety of "I didn't do enough this week" to "you did enough this week", and that kindness with yourself will cause a subtle but important energy shift. Just be kind to yourself. High achievers are very good at flogging themselves into hard work. There's a time for that, too. But be kind to yourself as well. For me, when I'm busy (and I've been quite busy lately, working long hours), this stuff just makes me feel better. And even in the context of long hours, tight deadlines & big opportunities, just slightly, makes work feel a bit more like play. Try it out. Maybe it will work for you, too. I hope everyone has a great holiday weekend with their families. Mrs. Caughran on a sister trip so the Caughran boys are excited for the first ASU football game tomorrow night and riding our new jetskis on Lake Saguaro. Go Devils!

Jevon's paradox, buy-side edition We've seen many research productivity tools emerge over the last 20 years: > Modeling: India outsourcing, Canalyst, Daloopa > Data: Credit card panels, web-scraping, app download trackers > Expert Network Transcripts: Tegus/etc. > Consensus Aggregation: Visible Alpha I started my career in '08 on a 7-person Consumer investing team at a Tiger Cub before the invention (or at the infancy) of these tools. Armed with only Excel, Bloomberg and GLG/Coleman, we needed that many bodies to do all the work we had to do. My whole reason for existence on that team in the first 1-2 years was to help & assist with these (now automated) administrative tasks. It would be absurd today to have a Tiger cub need a 7-person team to do what we did. So, the first order impact was labor destruction. But, with these tools evolving over the last 20 years, it became much easier for multi-managers to scale enough research rigor to generate P&L, and the evolution of these tools (particularly alternative data) created a brand new business at quant funds of data/systematic fundamental investing. People forget that quant funds also employ lots of people. I believe AI will directionally be similar. Sure, there are specific instances where teams can shrink from 7 to 3 and maintain research rigor. But the prize of alpha is so powerful at large firms that, more often than not, those teams will simply use these tools to go deeper and broader with more research rigor. And, as alternative data spawned brand new investing models, AI will almost certainly do the same. As quant firms do today, those models will need (lots) of bodies.

I’ve become reflexively skeptical of the coordinated “VC launch slop” playbook on X But I actually have a fairly large spec of personal logistics things I’ve been trying to accomplish with a text based agent. I tried to do this with Ollie AI but it just made too many errors. Grok Bot was too slow and buggy (hope it improves). But Instinct, the fake “invite and waiting list” manipulation notwithstanding, is actually pretty awesome. Has quickly carved through many personal tasks on my to do list. For investing work, this sort of interface UX will be awesome for investors who travel 15-35%+ of the time and want to direct their investing agents to do work while they are away from the desk (in the same fashion a PM today will direct an analyst back at the office to do work)

When people ask me the (unknowable) question of how AI tools will impact alpha curves in long/short investing, I actually think we have an interesting sandbox to evaluate that question in the evolution of Consumer L/S investing. I started my career as a hedge fund analyst in an 8 person consumer team, and pure consumer investing was a flagship team for many Tiger Cubs in the 1995-2015 era. The process was pure scuttlebutt research and I spent a lot of my day calling Wendy’s franchises, Cabela’s gun counters, running quarterly surveys and slicing and dicing grocery scanner data and constantly updating a master list of global SSS data. Technology has dramatically changed Consumer L/S investing and, in particular, the emergence of credit card panels that can give almost a real-time signal of company fundamentals made so much of that scuttlebutt research obsolete. And you had a window of time, maybe 5-7 years, where investors with the “old” approach were able to adapt these new methods and absolutely crush it. Over the last 5-10 years, however, fundamental alpha in the Consumer L/S space has become much more difficult to harvest. Consumer is no longer a flagship team at most single managers, and many Consumer PMs have drifted into TMT. Some of this is the macro/industry evolution of where EV has been created, for sure (see: AMZN), but more of my friends simply won’t traffic in names where alt data is the deterministic resource. It’s an arms race, and quants on balance are winning that arm’s race. This has created a tougher playing field, but has also created all sorts of monetizable distortions and the domain knowledge of different nuances around these panels (which providers are included, Midwest vs. coastal bias, which providers are losing a key CC panel, etc) has adapted to still offer interesting “third order” alpha pools for investors who have adapted. But it certainly is not the same “data good, stock is a long” set-up that prevailed in 2008. What has been bad news for investors has been great news for the data ecosystem, as the arms race to stay current has included $5-15+ data and data engineering budgets. In the broad consumer investing ecosystem, data has taken share of wallet from labor, and quants have taken share from fundamental. Get ready for this prior to asset itself in other sectors, in my view. (My hunch is the “next Consumer” is Biotech investing, but we will see.)

I’ve watched this 3 times now 1) it’s so cool that @sequoia and @sonyatweetybird share these sessions. There is a lot of slop on YouTube, but these events are insanely high signal. So thank you for sharing! 2) @gabepereyra seems like an animal and I think, indirectly, gives one of the strongest arguments for why the vertical layer is set to exceed the horizontal layer in many domains, by effectively operating a Moneyball approach building on the frontier ecosystem (SOTA models, open weight models, but also post-training, eval architecture, etc). And how deploying AI agents to do things like create vertical eval loops is accelerating what a smaller vertical lab can do. Proof will be in the pudding, as they say, but it has been hard for vertical AI to keep up with the constantly improving pre-train bar from horizontal AI. It feels like we are nearing an acceleration moment across domains where the pre-training / generalization thesis hits a wall and the post-training & “Moneball” approach on horizontal AI ecosystem starts to actually solve real commercial problems in domains like law & finance. Exciting times.