Posts

Recently I was helping a family member setting up a Revolut account in Hungary. A step in this process included verifying their address. It seems to me that – unlike many fintech apps – Revolut does not accept utility statements or bank statements, but instead requires you to upload a picture of your residential address card (in Hungarian: “lakcímkártya”, or “lakcímet igazoló hatósági igazolvány”). The residential address card is one of the pinnacles of Hungarian bureaucracy – it is both quite pointless and also very important, you often cannot get things done without having it with you....

Since the acquisition of Kùzu, a number of forks have popped up: LadybugDB/ladybug: Ladybug is an embedded graph database built for query speed and scalability. predictable-labs/ryugraph: Ryu, a fork of Kuzu, is an Embedded Property Graph Database built for speed with vector search and full-text search built in. Kineviz/bighorn: Integrates with Kineviz’s GraphXR visualization platform. Supports both embedded and stand-alone server modes. Vela-Engineering/kuzu: KuzuDB fork maintained by Vela Partners. Embedded graph database for AI agent memory with concurrent multi-writer support....

This tutorial is an adaptation of “Public DuckLake on Object Storage” to the object storage of Leafcloud, an Amsterdam-based cloud provider. Hosting DuckLake on Leafcloud Setting Up the Object Store Navigate to the Leafcloud dashboard at https://create.leaf.cloud/. Go to Object Store | Containers and create a new container. We’ll use the name ducklake-storage. Tick the checkbox for Public Access and copy the link. This will be e.g. the following: https://leafcloud.store/swift/v1/AUTH_f84982a3c5d04bd0846197d8e8ce3ddd/ducklake-storage Setting Up the OpenStackClient Fetch the credentials from Leafcloud....

(Updated on 27 June 2026 to reflect my latest experience.) Installing Windows 11 on the Framework Laptop 13 turned out to be trickier than I initially envisioned. The main difficulty was that the Windows 11 image does not recognize the disk manager of the laptop’s motherboard. So, without installing that driver first, it was not able to create the partitions required for installation. Configuring the boot During boot, press F2 and pick Administer Secure Boot....

“Benchmarking gone wrong”, one of the best and funniest benchmark talks I’ve ever come across – it’s well worth 6 minutes to watch it....

This year I have taken a few long train rides across Europe, such as Berlin to Amsterdam. I found that these need a bit more preparation than the typical short flights within Europe, so here’s a(n incomplete) checklist for these trips: Gear: Dress comfortably. Take noise cancelling headphones. Take multiple charging cables. Some companies just rolled out USB-A ports, so I would definitely take both a USB-A and a USB-C cable for phones....

Following the recent retirement of KùzuDB, I published an updated version of my Graph Databases slide deck. I originally presented this back in February at FOSDEM 2025. Now I updated it with new systems such as KùzuDB’s fork, LadybugDB....

In the last decade, I had my fair share of bad sleep. Often, I pushed through it helpless and the usual tips – aim for a regular bedtime, stop caffeine intake mid-afternoon, reduce screen time in the evening – did not help much. However, over the years I built up a repertoire of techniques. This is obviously not medical advice. All I can say that if you’re struggling with bad sleep, I encourage you to experiment....

I recently discovered the typos command line tool and created the following workflow around it in Git repositories: typos -w # writes the changes to file git add -p # allows you to review the changes interactively Some git clients use line-based diffs, which make the differences difficult to see. In these cases, you may want to switch to the diff-highlight tool, which is shipped with Git. On macOS with Apple Silicon and a Brew-installed git, you can use these commands to make git use it:...

I recently attended SIGMOD and chatted with people in the data management and graph processing communities. These conversations and other interactions led me to come up with a few research ideas. I lack the time to actively pursue them, but I list them below for future reference. If you are working on something similar or would like to collaborate, please drop me a line! Single-node LDBC Datagen: I have long thought about developing a single-node variant of the LDBC Datagen....