Think of a standard for classifying and regulating the self-hosting of open-source models similar to how an FFL works. You can do it, but you must have all your paperwork lined up, with background checks, a valid business license, and if you forget to dot an "i" or cross a "t" the Cyber version of the ATF shows up and shoots your fucking dog.

So, yeah. The "article" is incorrect from nearly the get-go about the "wild west" Atari age.

How have you set things up to have a good experience?

Now the software doesn't matter. The code doesn't matter. The hardware doesn't matter. Anyone can generate anything for anything, as long as they pay the fee. I think it can likely be argued that participation is now gated more than ever and will require usage of an LLM to keep up and maintain some kind of competition or even meager parity. Open weight models are not really a means of crossing the moat; none of the open weight models come close to the functionality, and all of them come from the same types of corporations that are releasing their models for unspecified reasons. The fact remains that the moat created by LLMs for open source software has never been larger.

I took it upon myself to write a library for my current employer (4yrs ago now?) that abstracted and standardized the way our Rust services instantiated and utilized the metrics and tracing fundamentals that OpenTelemetry provides. I recently added OTLP logging (technically using tracing events) to allow for forwarding baggage / context / metadata with the log lines. The `tracing` crate in rust also has a macro called `instrument` that allows you to mostly auto-instrument your functions for tracing, allowing the tracing context to be extracted and propagated into your function so the trace / span can be added to subsequent HTTP / gRPC requests.

We did all kinds of other stuff too, like adding a method for attaching the trace-id to our kafka messages so we can see how long the entire lifetime of the request takes (including sitting on the queue). It's been extremely insightful.

Signoz is newer to the game. I'm glad there are more competitors and vendors using OpenTelemetry natively. We originally talked to some of the big vendors and they were going to gladly accept OpenTelemetry, but they marked every metric as a "custom" metric and would charge out the wazoo for each of them, far in excess of whatever was instrumented natively with their APM plugin thingamabob.

The more the better. I love OpenTelemetry, and using it in Rust has been mostly great.

If you read many of the grimoires, there is very little NLP of any kind. The Papyri Graecae Magicae is one of the oldest explicitly magical documents we have from Greek Egypt, and it does have some manipulation spells (as most magical documents do) but none of this has to do with coersion to join a religion or join in a war, or to "do bad stuff". It's largely "technology" used by a practicing magician (a moonlighting Egyptian priest) to help the laity deal with their daily lives regarding helping their crops grow, animals not get sick, healing sick children, getting revenge on their neighbors and former lovers etc.

Magic is always a tool in the hands of the oppressed as a response to tyrannical hierarchy.

It was easy enough that I, as a single SRE (at the time) could write and implement across dozens of services in a few months of part-time work while handling all my other normal duties. OpenTelemetry has proved to be worth the investment, and we have stayed within the Grafana ecosystem, now paying for Grafana Cloud (to save our time on maintaining the stack in our Kubernetes clusters).

I would absolutely recommend it. I would recommend it and hopefully use it at any new future positions.

Science doesn't negate an interest in the arcane, esoteric, or occult. You can still find this stuff fascinating, and in fact there are practitioners who are actively involved in scientific circles simultaneously. It is not always mutually exclusive.

I love pretty much everything these folks do.

That voice is the voice of addiction. Once it starts it never fully goes away. It just gets less frequent.

However there is a gotcha. There are nearly infinite ways to setup your Pulumi codebase and there is no real prescribed structure outside of "create stacks". This can lead to a lot of time spent in architecting / implementing / testing different structures and abstractions which are not really a thing with Terraform. This can take time and can lead to refactors if you are not extremely intentional with your design.