Posts tagged: thought

All posts with the tag "thought"

843 posts latest post 2026-04-15
Publishing rhythm
Apr 2026 | 17 posts

vim usage is becoming normie level. Just like archinstall made it too easy to install arch and brought normies into the ecosystem. It killed ArchBTW^TM^, distros like lazyvim have killed vimBTW^TM^. It used to be that to run arch, vim, nvim you had to read the docs, and go deep on understanding. running archinstallor lazyvim make it so easy to get started that you miss all of the details, you no longer have to understand ctags, quickfix, what an lsp is, or even how to set your own keybindings. You just use the damn thing, like you would with VSC****. No shame to anyone who does this, but you are probably missing out on a bunch of really useful features of a very core tool in your workflow.

Just discovered Sylvan Franklin in this post and he is cracked, sub now.

I just never quite understood why the word just can send people over the top. I get it when you don’t know someone, you don’t have history with them, and they come in saying you are doing something wrong.

I pulled this out into a full post just

2025 is not the year to get put on the market, its rough out there. Junior’s have little chance, senior+ are even struggling. We had it easy from 2020-2023, now its over saturated and you have to want to be in this industry to be here and stay here. It used to be a fine place to get a good job to pay the bills, the bar has been raised and if you don’t want to be here you are going to struggle. Theo covers this in this linked video deeply [[ thoughts-472 ]].

David’s design on his blog is fantastic likely from years of small improvements like this converting ugly quotes to pretty quotes and optimizing fonts.

It’s common for markdown libraries to convert the first to the second like my build script does.

This is new to me, I had no idea that markdown libraries did this, I’m now interested if markdown-it does it.

For subsetting I use the fontTools library but I’ve no idea how to setup Python environments. I got it working once and failed to document the process.

...

pypi yanks suck, they are rare, this one got me today as it was a pinned dependency in my dependency chain. The latest release broke python 3.6/3.7 (which 3.6 has been EOL for 3.5 years btw), and it claimed >=3.6. In order to allow users to still install xlsxwriter without pinning down it needed yanked. I’m not sure if there was another way around it as pypi releases are immutable, so you cannot fix

This now has me wondering what the heck is using it with old pythons.

It appears to have broken builds on Canonical/checkbox for ubuntu 18.04. Checkbox is a device compatibility testing framework.

https://github.com/canonical/checkbox/actions/runs/14644718138/job/41098549191#step:8:125

I try to use conventional commits on all of my commits, but I often end up only using feat/fix. I need to keep this page handy and get new verbiage worked into my language

Optionally include a scope fix(parser):

A bang indicates a breaking change note. For example …

Great breakdown of nextjs. I was highly unaware of its performance optimizations before reading this. The smell of vendor lock in from next/vercel has been there from the start, this is the first real claim I’ve seen.

I’m out on modern js front ends, complex builds that change every 6 months, design patterns are out of date just as fast. Its hard to keep up, especially when you don’t have the use case for highly interactive apps. Libraries like htmx or plain ol js gets the job done on the majority of sites and everything I tend to work on.

I’m totally with Prime here, there is something about the read only, mouse clicking part of my brain that causes me to be more critical of the code at a different level. It doesn’t hit the part of my brain thinking about the edit or how to do the edit, it hits a part thats thinking about how I will have to deal with the code moving forward.

Vendor lock in disguised as performance. Nextjs aparantly now streams all of your metadata on the fly with js. This would obviously kill all seo right, well not if you’re on vercel they automatically detect search crawlers and serve the metadata. Why the f do they need to do this and not just serve everyone the metadata. The Web is this beautiful place where anyone can create and build amazing things with a relatively low skill. Js is meant to be enhancement, not degrade the experience of its users.

I thin a lot of us have this issues, especially on side projects. At work therre are expectations, jira tickets and so on, keeping you shipping. I think there is something to be said about getting that quick and dirty POC to the right group of people early for feedback before you add redis caching, kubernetes, auto scaling, disruption budget, distributed nodes, high availability, backups, disaster recovery. At work you kinda have to have the right person to shoot ideas by that can understand that you probably need some of these complex things for your app and it will take time to get right.

I would love to have a browser based video editor I could throw on a server and do quick edits from anywhere. I tried to get this one to work and struggled to get front end to send api requets to backend. I think the root of it was their redis wants to run on 80, this caused a permission error so I tried to run 8880:80, but redis was still unable to start due to a config permission error.

I love this idea of tiny useful apps for yourself. In fact I’m working on a project to built out tinyapps for myself to replace my common needs. I absolutely love that all of the state is stored in the url bar, nothing is stored server side. As much as I love to hate js, I really appreciate that things like this can be built to just live on the web, be accessible from anywhere, and live practically forever as they require such little hosting demand.

This looks like a very useful formatting tool to keep in the back of my mind. I do a lot of python and our tool tends to be pre-commit, named after the git hook pre-commit. It specifies a bunch of tools to run, you can run them in ci, manually, and opt into doing it before commit. I like the simplicity of this one not needing a whole ecosystem, but rather just leveraging the cli commands from those tools. This would probably be something that would get in the way of setup for new devs and not something I would throw on one project by itself, its another thing for everyone to figure out how to install and run on every platform, I’m sure its not hard, but being on python teams pre-commit just fits in.