As, probably, many of you, I did a long journey from manual FTP sync, to semi-automated tools like Capistrano, then manually managed docker deployments, then anisble managed docker containers and finally, met the orchestration system of choice - Hashicorp Nomad.
TLDR: Kubernetes is too heavy for my moderate needs. Hashicorp Nomad strikes the right balance between features and maintenance complexity.
At some point in time I felt a need for better container orchestration system that I could run by myself in Hetzner.
Rubu on Rails development changes a lot every year and sometimes it's not easy to keep up with the changes.
I've recently bought Sustainable Rails Book and it's hard to overestimate how valuable is it for me.
Here are couple of thoughts after reading first chapters.
Invest in developer friendly application bootstrap process. Running bin/setup and bin/dev should be enough to run on the new blank developer machine. Move business logic out of fat controllers and fat models.
The article provides overview of facts why Emacs is so different but also so
powerful comparing to other editors. It gives a good starting point for the
reader to understand the fundamentals and start own journey to effective work.
The way of learning is painful, but rewarding. The essense of Emacs is to codify your specific repetitive tasks so that eventually you realize, the editor contains a lot of customizations speeding up your daily routines, as no other editor does.
I do the following things in Emacs:
Coding in following languages: Ruby (Ruby on Rails), Python, Clojure, Javascript, TypeScript
Writing Terraform and Nomad configuration with HCL
Run Large Language Model on your PC in 5 minutes For the summary
I used model whisper to extract text from audio of this video and asked mistral to summarize it.
echo "this is a transcript of the video. Summarize this\n $(cat subtitles.txt)" | ollama run gemma:7b So, here you go, output from mistral 7b:
The speaker explains how to download, install, and set up LLMA, and then demonstrates how to use it to run a specific model called “mistrial” with a pirate-themed prompt.
While Ansible is a great tool for certain use cases, such as building new AMIs when used with Packer, it has some downsides. These become apparent when using it for container orchestration (which I have witnessed at scale) or keeping the configuration of a set of nodes in sync.
Another significant, though not dramatic, downside is related to Python version and dependency management. This adds extra complexity that I would prefer to avoid in a system management tool.
Recently, I stumbled upon a concept that sounded new to me: Compliance as a Code.
The core idea is to “embed compliance policies into the code that can be repeated and tested automatically.” But what does it mean in practice?
For example, consider PCI DSS, which requires measures to secure credit card information to ensure compliance. Tools like Terraform, Ansible, and others should contain the code for encryption, access control, and data obfuscation.
Very simplified about the AI revolution happening now. We are only at its very beginning. Next, automation and integration among different AIs will increase, leading to more and more complex tasks.
There are still many problems, such as - training a truly large model, and making it accessible for general use is still very expensive and very energy-consuming. It is still not very clear what to do with regulations, because it works both ways, making it significantly easier to assemble some kind of bomb at home.
SRE Simplified : Error Budget concept.
The concept of Error Budget is useful for setting up alerts of different severity. This video explains in extremely simplified manner how it works.
