Erik Lundevall-Zara’s notes
2025-02
The value of impact (2026-02-09)
For long parts of my career as a consultant, I have valued smaller organisations - both as employers and as clients. The idea is that these organisations tend to have less red tape, and you have a greater chance to make a meaningful impact.
Recently I had some time to reflect on two previous work positions.
One of them was in a pretty small organisation, around 30 persons. Let us call it organisation A. The other one was a slightly larger organisation, perhaps 500+ persons. Let us call it organisation B.
In organisation A, I was switching between many tasks and assignments. The cycles of development and deployment were pretty short. However, some parts of the job did not change, despite pointing to issues for years. It was not really part of the main business.
In organisation B, there was quite a bit more red tape and parts of the organisation was moving quite slow. Software However, there was a clear mandate to improve the way of working, and an expressed appreciation for the efforts I made in many parts of the organisation.
Did I make an impact in both work positions? Yes, I believe I did. However, I definitely felt that there was more of an impact that was meaningful in organisation B.
So even though the size does matter, the alignment with were the organisation moves or try to move matter also.
Make sure that what you do can feel meaningful.
Come for the language, stay for the tools, ecosystem, and community (2026-02-06)
In the past year I have dabbled with several languages outside of the regular ones I have been using. Some of them were new to me, other I have used in the past. These languages included Elixir, Gleam, Ocaml, Roc, F#, C#, Kotlin, and Zig.
All of these languages are quite nice, but from a tooling perspective I am a bit hesitant about Ocaml and Kotlin. It is not that the tooling is bad, but it does not have the simplicity of some of the other language tools.
I have also found that I value simplicity in languages more and more, this put Gleam, Roc and Zig on the top of my list. Zig seems to fit well with my plans for some command-line tools, and Gleam for web and I/O oriented projects. Roc will be quite interesting at some point as an embedded language, I think.
Move to Codeberg (2026-02-02)
I moved the repository of this website from GitHub to Codeberg. Codeberg is run by a non-profit organisation, and located in EU, so this is a big plus in my book. The software seems to have evolved a fair amount since I first looked at Codeberg a couple of years ago, and now fits much better for my use.
First microblogging entry (2026-02-01)
The notes section on this website is an attempt at doing microblogging. So most material here will be rather short items, as a complement to the regular long-form blog on the site.
I also added a few short descriptions on a couple of the more recent blogposts here, to make a transition between the long form and short form posts.
2025-11
LLM: Hallucination or Intelligence? (2025-11-19)
We tend to call LLM results “hallucination” when we don’t like them and “intelligence” when we do. But that hallucination is actually a feature of the LLMs. Also, what really causes problems in software projects has never been the speed of writing code — it is unclear requirements and poor communication. Good engineering practices help regardless of whether you embrace AI tools or not. Read more
2025-09
Pair programming with AI rubberducks - is it worth it? (2025-09-15)
I tried using Claude Code as a pair programming partner to build an F# command-line tool, with strict TDD discipline. The results were mixed but more positive than negative. Read more
2025-08
Rubberducking with Claude (2025-08-13)
I had a small exercise using Claude as a digital rubber duck for a coding problem. Claude’s suggestion was actually wrong, but the process of explaining the problem and arguing about it helped me confirm my own approach was correct. I think LLMs can work as rubber ducks — if you treat them as thinking aids rather than problem solvers. Read more
2025-07
Pair programming with AI - an experience report (2025-07-29)
I set up an MSK client EC2 instance using Pulumi and F# with some AI assistance. Claude helped me save time but struggled with F#’s strict type system and Pulumi’s input/output types. The real win came from noticing the code felt ugly, which led me to discover Pulumi.FSharp — something Claude never brought up. It was a good learning experience. Read more
2025-05
Support plans - AWS vs Scaleway (2025-05-13)
I compared the support plans between AWS and Scaleway. My impression is that Scaleway support is probably cheaper and more predictable, especially for small and medium-sized companies. Read more
Getting started with Scaleway (2025-05-08)
I started exploring Scaleway as a European cloud provider, and it looks like a promising option among European cloud providers. Read more
Considering European Cloud options (2025-05-06)
Should we consider European cloud providers instead of the big three? My take is that they do not need to compete everywhere — they need to be good enough in areas like simplicity, pricing, trust, and community. Finding a cloud partner that suits your company size may matter more than going with the largest provider. Read more
2025-04
Hands-on Infrastructure as Code with AWS CDK - the book is available (2025-04-25)
I published my online e-book on using AWS CDK with Python — it is available at no cost. It is based on a blog series I wrote earlier for TypeScript, updated and refocused on Python. If you are looking for a practical introduction to managing AWS infrastructure from code with Python, I hope you will find it useful. Read more