Mastodon Posts

Relied heavily on opencode. I've steered the application to use a stack I wanted (neon postgres, react, astro). For initial stuff it works great.

My idea is now to get to know the projects, learn a bit about new concepts (electron, transformerjs, ELO system, etc)

In the last two weeks I've gotten around to implement projects I've had in my mind for some time:

- Implemented a kahoot-like game for a birthday. Used neon postgres to handle multiplayer and it worked great (some rough edges I'd like to atend)
- Implemented an app to help organize my weekly football games. Includes web notifications, and maybe an ELO system
- Created an app to help my D&D DM manage the music. Heavily based on Kenku.fm but will use to test auto recommendations

Fun stuff

My next challenge. I've created an opencode agent that serves as a mentor.

You are a **pair programming mentor** -- not a code-writing assistant. Your primary goal is to help the programmer think through problems, arrive at solutions themselves, and deeply understand every line of code that gets written. You never write code on behalf of the programmer unless they are completely stuck and you have exhausted all other teaching strategies.

Gonna be test-driving it for the next days

I've been trying to get my mind on how the LLM's can help boost the productivity without devs losing their edge. It's way to easy to just let it do it's thing, specially with the latest models.

For now my mental model for this shift is:
1) we can't trust it's output, so we need to verify it, either manually or through formal tests
2) without typing everything we need to do more code exercises to keep the knowledge
3) LLM is a tool, we are still the owners and take ownership of failures

Two super powers to learn a new technology: read the documentation/source code and use extensively

Fiddling around with a thermal printer + Web USB API. After some hours working with a text encoder found this god saving library.

Just need to understand how to get the QR codes working now 🤔

I was very interested in the new memory manager to help compact a month of parquet files. Still got an OOM but as it was ~1B rows of data I guess it's a bit overkill

Reading algorithms to live by and I think will be a book to have for later consulting. Regarding probabilities I find we look too much into averages and binary certainties. However life is built of complexity and in fact probabilities follow distributions like the ones below

Advent of Code can be a very humbling experience

After all this years I still struggle to install python versions. I think the issue is now around zlib on Mac but ended up with what always works for me.

1. brew install python@3.11 python@3.12
2. alias python=python3.12
3. And use poetry on top for each project

Going to present at a local meetup after more than a year without doing public speaking. Let's see how it goes 😁

TIL: After adding a new patter to gitignore I can remove existing files from VCS with:

git rm --cached `git ls-files -i -c --exclude-from=.gitignore`

Getting back to preparing a talk. Feels like getting back to training. You feel awful at first and see no major results

TIL: DynamoDB always returns the entire item. Even with a Projection Expression it will retrieve the entire item and then filter the intended fields. I guess it's the same with Postgres (row based storage) but my expectations were misaligned

TIL: Kernighan’s Law
“Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it.”

Hello World