select links from 2026-02-01

ClawdBot? 😪 MoltBook? 😴 Cow Tools? 😍

Feb 01, 2026

Hi friends! It’s -20 degrees Celsius over here in Lithuania and I saw four runners through my kitchen window this morning. As they say, there’s no such thing as bad weather! This winter kind of reminds me of the 2019 Polar Vortex in Chicago, though I can’t replicate the sensation of my eyeballs freezing. Generally, the Midwest has a similar climate to the Baltics, maybe it’s no wonder that a lot of Lithuanians flocked there. Did you know about the Map of Lithuanian Heritage in America? It’s lovely and I would love to have had it in 2019. The Map is one of the reasons why I love the US as a tourist – the abundance of footprints from different cultures could keep me entertained for MONTHS.

Anyway, ClawdBot, amirite? Or wait, sorry, MoltBot. No wait, OpenClaw. It was interesting reading the docs and thinking “wait, where’s the auth, how do I expose this safely”. Apparently, you don’t? Simon Willison raised similar concerns. My guess is the hype will die down and we will move on soon enough.

Understanding the API

HOWEVER, Moltbook did remind me about the necessity to understand one layer above yourself and one layer below. Basically the idea is that a lot of the things you’re interfacing with is an abstraction. Code assistants are abstractions around a set of technologies (LLMs, tool calling etc). However, an abstraction can feel like magic if you don’t understand its internals. There were so many “Moltbook is AGI” posts on my feed where really it’s just LLMs prompting each other to death. Boring.

John D. Cook has an old blog post about leaky abstractions too. Abstractions are easy to learn but hard to master, i.e. you will eventually run into issues when reaching a certain degree of complexity. A good example would be Django - the framework lets you do a set of operations VERY easily but at some point you might have to roll your own SQL queries. Code assistants are no different - very easy to pick up and use but you will need an expert or gain expertise at some point.

dplyr in duckdb

The tidyverse ecosystem is fun because you can write R code that directly creates query plans via duckplyr, you can also intermingle R and SQL code with the help of dbplyr. Now there’s another way to intermingle my favorite database with my favorite programming language: it’s the dplyr DuckDB extension. It allows you to write dplyr code inside a DuckDB session. I’ll repeat, you can open DBeaver, connect to a .duckdb database and write dplyr code. All without touching R itself. It’s mind bending, I love it!

Cow Tools

This would usually land in the other links section but just open the link and watch the video! Flexible use of a multi-purpose tool by a cow by Antonio J. Osuna-Mascaró and Alice M.I. Auersperg

High SQL

Discussion about this post

Ready for more?