Escape rooms and equity exposures: Thoughts on Q3
New team members, accounting robots, LLMs, big hairy audacious goals, and more…
Dear readers,
We started this newsletter series to share what we’re learning and building as a team. Speaking of which, we’re thrilled to have Zsolt and Valerie onboard!
In this Q3 edition: Peter reflects on public equities and automation wins; TJ and Zsolt unpack the systems behind our daily financial statements; Val shares her first impressions; and Justina explores content, culture, and the possibility of a “TB-style” investment firm.
We hope you enjoy this edition of the Titanium Birch newsletter.
— All of us at Titanium Birch
Peter 👽
I’m quite happy with some of our recent wins:
P&L attributions on the entire portfolio have become much easier and faster by using our fully automated, updated-daily financial statements. Using double-entry accounting lets our robots accurately describe the various inputs to a P&L, such as transfers, paid interest, accrued interest, various types of fees, price changes, FX changes, etc., and the daily reconciliation process our robots run on our books also means that we can place a certain amount of trust in the accuracy of the numbers. I’m very glad to have reliably accurate answers for everyone on the team to self-serve, explore, and find insights, and to have done so without duplicating efforts across “investment vs accounting teams”. I think TJ should be super proud of how he designed a solution and led us to build it. It was also great to see how quickly Zsolt onboarded onto our team, shipping features even in his first week.
Justina and I learned lots of lessons about investing in funds and have tried to feed those back into our due-diligence and decision-making processes. Our newest joiner, Val, brings lots of experience in this area, and she’s already sharing a stream of observations and ideas for improvements with the team.
We’ve come to put more and more trust into Interactive Brokers’ VWAP trading-robot. It reliably splits orders into many small pieces, trades automatically in relatively even chunks throughout the day, and usually delivers as its name promises: over the course of the trading day, it delivers to us a “volume-weighted average price” that’s extremely close to the market’s overall VWAP over the same time period. Here’s a screenshot of our dashboard that visualises the order-execution timeline: nice and steady throughout the day, all automated, and executed at prices that are almost exactly “market average”. I’m frequently amazed at how much functionality IB offers.
A recent theme for us that’s continuing into the summer: we’re getting better at understanding the exposures in our public equities portfolio, and we’re slowly trying to become more confident in our answers to “what exposures do we want and why?”.
Val 🌚
I’m excited to share my first update as the newest member of the Titanium Birch team!
I’ve just returned from a rejuvenating two-week holiday in Georgia (the country, not the U.S. state!). From hiking in the Caucasus Mountains to exploring the streets of Tbilisi, it was the perfect reset before diving into this new chapter.
This quarter, I’m focused on getting up to speed with our portfolio—deepening my understanding of the investments we’ve made, thinking critically about how the portfolio could evolve, and reconnecting with people across the industry. I’m also familiarising myself with the various tools we use day-to-day, which has been an insightful (and occasionally humbling!) part of the onboarding process. There’s a great deal to absorb, and I’m looking forward to contributing meaningfully as I settle in.
I’m especially grateful to everyone on the team for being so warm and welcoming. It’s made the transition not just smoother, but genuinely energising.
What I’m reading: The Caesars Palace Coup by Max Frumes and Sujeet Indap. It’s a gripping account of the high-stakes battle over Caesars Entertainment’s bankruptcy. The book unpacks the complex legal and financial warfare between creditors, hedge funds, and private equity firms, offering a sharp look at the inner workings of distressed investing and the power dynamics behind the scenes. A compelling read for anyone interested in corporate drama and financial strategy.
Zsolt 🌶️
print("Hello, World!")
The first few weeks after joining Titanium Birch in May felt like being back at school. I got to brush up on and learn a bunch of new concepts in one of my favourite uni subjects: accounting. We even found some surprisingly entertaining material that made IFRS 9 feel (dare I say it?) fun. What made it click for me was being able to apply it directly in a live project.
For Q3, I’m diving into how we can apply factor investment principles and make some upgrades to our infrastructure and data pipelines—so that as our data (and ambitions) grow, LedgerPigeon (our general ledger) can run at turbo speed.
TJ 🤖
Financial statements are harder to automate than one would think (but it is possible!)
Last quarter, we worked hard as a team to complete automations for bookkeeping across the entire portfolio. It was a great team effort, especially with Zsolt joining the team!
When we embarked on this journey to automate bookkeeping, it initially seemed like a straightforward task, but there were some significant blockers to overcome. Our biggest takeaways:
Use vendors to structure data prior to ingestion. While we could get automated feeds from most of our banks and brokers, there remained our illiquid investments, whose information stream was mostly in unstructured PDFs. Integrating with a vendor like Arch helped greatly (check out this blog post I wrote about how we did that).
Book information as it comes, and worry about balancing the books afterward. Booking with eventual consistency allows us to immediately reflect straightforward information (such as cash flows) while worrying about later information (such as classification or accruals and deferrals) later. Most manual accounting software enforce balancing on every transaction, but that only works if you close the books once a month.
Following dbt best practices, especially the one about separating source transformation logic and accounting logic, allowed us to add more than 16 different sources of transactions and consolidate that into neat accounting objects (expenses, payables, journal entries, instruments). It definitely would have been a tangled mess otherwise, hampering the ability to integrate new data sources.
We still have a lot to improve, but it’s a great time saver—and assurance—that the robots are booking, reconciling, and reporting daily with little human intervention required.
Building a long-term, comprehensive, and shared institutional memory
During a recent research sprint, I dove into LLMs and their state of development over the past few years. A few things struck out:
LLMs, by default, are text-in, text-out machines with no inherent memory. A “conversation” with an LLM is just sending the whole history over and over and asking it to complete the next likely response. Having just read The Man Who Mistook His Wife for a Hat, I could not help but draw parallels to a man who had Korsakoff Syndrome, could not form memories, and constantly confabulated memories to fill in the gaps. For LLMs, it seems one of the most important “jobs to be done” is surrounding the LLM with the context it needs to solve problems, just like you need to equip teammates with the accountability and autonomy they need.
LLMs greatly reduce the costs of building and accessing a shared institutional memory. We’re big fans of documentation at TB, and while before we would have spent a lot of time cataloging and organising this information in (hopefully sensible) hierarchies and triggering documentation updates, we might not need to do that anymore. Additionally, text documentation and quantitative data usually live in separate places, so LLMs might have a role to play in bringing them together.
What would it take to build a system such that TB, as a firm, never forgets anything? If we assess an investment opportunity in 2050, could we remember:
if we made a similar investment in 2025,
what the portfolio manager thought at the time of due diligence,
what the expected and actual J-curve looked like, and
how performance looked in years 1, 3, and 5?
I think LLMs are going to play a key role in making this information accessible even as the people, process, and technology have drastically changed over decades-long timescales.
Justina 🐯
Reflections on people and technology
Last quarter, Peter and I reviewed each of our private-market investments. It was energising, fun, productive, and full of lessons (especially around how we can improve our due diligence processes moving forward). On that note, I’m happy Val has joined us; I have a feeling our team is about to level up big time with her here.
We also leaned on tech to make that work more efficient. Our toil-killing Arch integration spared us the pain of manually collecting documents from our fund managers, and we used LLMs to comb through material and quickly flag things that needed deeper digging. I still have lots to learn about optimising my prompts (and found Claude’s prompt engineering tutorial particularly useful).
My biggest takeaway from last quarter: investing really is a people business. From a team perspective, I’m chuffed with how we’re growing. From an investment perspective, we’d love to get to know other LPs. If you see us at SuperReturn in September, say hello!
Big hairy audacious goals
We kicked off this quarter with the ultimate test of teamwork: an escape room. (Good news: we escaped.)
Earlier this month, Peter challenged us to write BHAGs (big hairy audacious goals). Here’s mine: Imagine that by the 2030s, people refer to a “TB-style” investment firm as its own category. Wouldn’t that be cool?
But what exactly is “TB style”? That’s something we get to define, evolve, and share over time. As the team’s official “keeper of the brand”, my job is to help everyone at Titanium Birch contribute to that definition… by sharing what we’re learning as we go.
So: we’re revving up the content machine. Expect more blog posts in the months ahead. :)
Stray thoughts
It’s been two years since I moved to Singapore and joined Titanium Birch full-time. I’m proud of the team we’re building and excited for what’s next.
On a personal front, last weekend, my cousins and I picnicked on a lawn to watch the National Day Parade rehearsal. As I ate corn on the cob and watched a Chinook helicopter drag the Singapore flag across the sky, I felt a strange sense of belonging in this city. I’m grateful for the friends I’ve made here through soccer, music, and volunteering communities!
Final update: I’m about to move homes. My dream is to get a wooden dining table and collect random chairs to put around it. Maybe an industrial-looking chair from a retro store, a wooden stool from the curb, an orange metal chair from Carousell…? Then I’ll throw dinner parties and invite friends from different parts of my life to eat together—strangers to each other—and trust that they’ll all get along.
ICYMI: Blog post roundup
Using Arch to streamline oversight of our alternative investments by TJ
On being a generalist at an investment firm by Justina
If this newsletter resonated with you, we’d love it if you could pass it along to a friend!
Until next time, thanks for tuning in!
Disclaimer: The content in this post should not be taken as investment advice.