Donald Lee-Brown

"I don't always check my code, but when I do it's in production"

"The difference between 'looks right' and 'is right' is enormous." That's how Strib Walker, our Head of Product, framed a core challenge of building AI for finance in a conversation with @Anthropic. To close that gap, we embed former bankers, investors, and research analysts alongside our AI researchers and engineers. Read the full Q&A: claude.com/customers/rogo

The third best-selling computer platform in history, after Windows PCs and the Mac, began as a recruiting tool for Cambridge University's computer science department. It's the size of a credit card, has no case, and costs less than a pair of shoes. Eben Upton built Raspberry Pi in 2012 to get more applicants into Cambridge's computer science course, then the easiest to get into. He thought kids needed the real thing: a general purpose programmable computer (like his childhood BBC Micro) to fall in love with the unbounded creativity of coding. He was more right than he could have imagined. On launch day in 2012, he sold 100,000 computers. A million shipped before Raspberry Pi hired an employee. Computer science is now the hardest course to get into at Cambridge, and Raspberry Pi is a $1.5 billion public company that has sold over 73 million units. 80% of its revenue comes from industry. Every digital display at Heathrow runs on a Pi. Schindler uses them in its elevators. The International Space Station has carried one in orbit since 2015. You'll also find the tiny computers wherever the next thing is. Bitcoin mining farms ran on them. So did the first wave of hobbyist drones. Pis now run LLMs. In five years, Upton thinks Claude Sonnet-class intelligence will fit in your pocket. Most tech conversation is about the frontier: the newest chips, the biggest models, trillion-dollar training runs. Raspberry Pi is a case study in the opposite. It shows what cheap, general-purpose, and performant-enough can achieve. It's also a rare British hardware success story, designed in Cambridge, manufactured with Sony in Wales, and reshored from China a decade before the rest of the industry caught on. Read @TerranMott's interview with Upton below. It comes with extraordinary photos of Pis baking in the Welsh factory, and covers the journey of automation, teaching children to program in the era of agents, and putting foundation models in your pocket.

Some early thoughts after building real apps by myself for the first time… We built an internal tool called Conveyor It’s an app builder, and internal App Store It is connected to all of our data, context, and external data APIs I’m completely and utterly useless as an engineer, but I’m good at knowing what I want a tool to do. I’d previously struggled to make useful programs with pure CLIs. Our wrapper made it easy for me. In the first 3 days of having this tool, I’ve built several fairly complicated applications, two of which I’ve used a ton for real work. I’ve only used a couple hundred million tokens so far. Some early feelings: 1) It’s obvious to my that my companies Positive Sum and Colossus will have fully bespoke operating systems, built in house. They will manage as much of our work as possible. This is already exploding for things like research and reporting. Every business will want this for themselves. Sure we won’t built our own slack, but we will built everything that pertains specifically to our shape as a firm, which is a lot. 2) x402 protocol (which enables AI agents and users to pay for API access and digital services instantly, without accounts or subscriptions) is immediately interesting to me. Many times I’ve wished I could just stream payments for individual data points. 3) right now each loop of prompt to output takes 5 to 15 minutes. As models and ASICs (@Etched !) make this faster, it’s going to be so much more fun. Even 5 minutes makes it hard to get in the flow. Can’t wait for seconds instead of minutes. 4) it’s so much easier to design things by starting with a shitty first draft of an app and seeing what’s wrong and iterating than nailing a full design ahead of time. When I had directed the design of software before this was always maddening and slow. 5) this has made me realize that my imagination had atrophied. Use it or lose it is real. Very quickly I’m finding it easier to have good ideas by building more stuff. I encourage everyone to do the same. So fun and rewarding. 6) We need more compute

top shelf product

I wanted to share what we learned over the past few months, building agents 🧵

If it turns out that context engineering is the killer use case for data catalogs but they've all shuttered b/c the market timing was wrong, I'll lose my mind.

We’re excited to announce our $75 million Series C led by @sequoia, with participation from @jpmorgan, @WellsFargo, @ThriveCapital, @khoslaventures,@BoxGroup, Mantis VC, Positive Sum, Stonecroft Management & @_altcapital to build AI that makes the best financial professionals smarter, faster, and more ambitious. There’s a long road ahead and we’re grateful to our clients, team, and investors for their partnership as we set out to transform finance. rogo.ai/series-c

Blake read all the books in the “Silicon Valley Canon” and wrote about his takeaways. Fascinating!

Today, we’re introducing Rely: the fastest way to close commercial real estate transactions, starting with multifamily. You’ve probably seen a lot of talk about AI in real estate. We’re using it to fix the part of the business that no one talks about. Multifamily transactions are held together by late nights, messy data rooms, conflicting numbers, and manual work that no one fully trusts. The process is slow, expensive, risky, and it’s been accepted as “just how diligence works.” Rely changes that. Rely ingests entire data rooms and runs full-coverage audits across leases, rent rolls, contracts, financials, and more. Every output is traced directly back to its source, creating a single place to understand, validate, and manage the entire transaction lifecycle. The result: up to a 95% reduction in diligence time. Now that Rely is live, I’d love your help. We’re learning directly from the best operators in the industry. If you’ve lived this pain, or have a strong opinion on where diligence should be headed, I’d value your perspective. Drop a comment or reach out directly. We’re listening. Huge thanks to our early partners and backers: @CardinalGrpMgmt , Elm Grove, Tarragon Property Services, @btv_vc (@pitdesi @iamjakestream @jbahrdestefano @nbobba @lsodaro @digmonster), Del De-Windt (Byld), Howard Smith (former President, Walker & Dunlop), @chasegilbert_ (Built Technologies), @jakebolling (Scotch), @tomloverro, @JaxGeller , @PhysicsBliss, @pk_iv and more. Your belief helped make this possible.

Tomorrow.

We are hiring a research analyst Here are some signs you might be a great fit: - hearing “if we capture just 1% of the market” in pitches makes you wince - the harder something is to learn, the more you want to learn it - you read wikipedia for fun - you’ve said “this would be so much easier in Python” - you have a PhD but hate academia - you feel restless if you haven’t built anything in a while - you know the difference between being accurate and right and being precise and wrong.  - you use phrases like “strong prior” and “null hypothesis” in casual conversation - lateral thinking is second nature for you - you appreciate complexity and strive for clarity - debating your ideas with others fills you with excitement, not dread - it’s obvious to you that working harder AND smarter right answer - You believe the trope that too much research for VCs deals leads to bad decisions is ridiculously dumb - you want to work in a unique office in NYC (we work out of same office as two other incredible investors and their teams) - you value evidence more than eloquence Apply below!

I'm hiring a researcher at @psumvc Truly high quality research in VC is rare. That makes it valuable, not just for us, but often for founders, too. Our goal is for every founder to say our work was by far the most insightful view into their market of any firm. More ⬇️

@andreylebedev29 I’m a huge @Speedwell_LLC fan, imo one of the absolute best in quality and depth. Agree that potential impact goes up the farther down the EV scale you go. Way bigger space of potential deals, tighter diligence economics, value creation a different game, etc

Appreciate that! Love what you guys do! I suspect ai will have the greatest impact in unlocking the sub-$100m EV space. And that we’ll see regulatory changes enabling more liquidity in late stage venture (high growth software / network effect businesses). But I’m sceptical that vanilla PE (mid-single digit growth, $500m-5bn EV) will see the same transformation. Those businesses are operationally complex and illiquidity matters more when you’re holding billions vs millions. Re CoStar there’s a great long form podcast from @Speedwell_LLC. In fact I discovered it through a Colossus recommendation. Worth a listen (if you haven’t already!)

Thought provoking article from Colossus (I'm a super fan) but I think it misses the structural nature of private market illiquidity. Bloomberg isn't the right analogy as it automated price discovery in an already liquid asset class. The right analogy is CoStar Group. CoStar brought price discovery to commercial real estate, an illiquid asset class. And although this increased institutional capital and introduced new strategies in CRE, buildings still take months to trade. And companies are even more idiosyncratic and harder to transact than real estate. Illiquidity and transaction complexity are the core risk factors in private markets. And better price discovery doesn't fundamentally change legal restrictions/processes.

More takes: 1. Using AI for tasks like summarization and call notes is a consensus take. If you are here, you are behind the curve. But it's easy to catch up - @NotebookLM , @meetgranola , etc. 2. Most 'deep research' tools do deep *search*, not research. Don't confuse them.

micro-integrations

@spacecasetayl0r @patrick_oshag 👋

@patrick_oshag Nice work on this. Very solid research and I largely agree with the conclusion Is @PhysicsBliss the author?

Can you make any money investing in “the final frontier?” We found ourselves looking at tons of space companies before finally deciding to answer our most basic question in writing… IS SPACE INVESTABLE?! I learned so much from this investigation: joincolossus.com/article/is-spa…

March 2023: An Open Letter signed by 30k luminaries calls "all AI labs to immediately pause for at least 6 months the training of AI systems” to avoid “loss of control” April 2025: still not entirely clear how we reliably analyze a bunch of PDFs in corporate folders with AI

1/ SpaceX is one of the most valuable private companies in the world. Launch costs have dropped by nearly 45x since the days of the Space Shuttle. Over $75bn of investment has been poured into companies building space infrastructure since 2015. But investors with long memories will think back to the turn of the millennium and ask, "Haven't we been here before?"