Last Week in Code E006

Its E006. YAY. It has been a week and a half. I spent most of it doing archival work. The last part of the week I sunk into setting up mediawiki. I did some interview prep type questions also and I uploaded some of the solutions. But that wasn't as serious as the archival tasks.

I am preparing to change locations. I am not so sure about my new location's internet efficiency or its affordability. So I think I will so much need to have local access to critical information sources. As usual, the whole process and other small tasks I did taught me certain things I'd like to record right here, for sharing and future reference.

Its a lot. Its disconnected. It will be sauce for distinct articles I hope to post in the future. But, I hope there's something to learn.

Documentation as Infrastructure: Discovered from the trenches

I am making a habit of documenting the process I take when installing new software for how unique situations turn out to be for me Vs what online guides imply they'll turn out like. Sometimes the guide is written based on a system that isn't the one I am running. Or maybe there have been a lot of system changes since the guide was last updated. Or maybe I run into errors that the guide does not document how they are to be fixed or what to do to avoid them.

And for that or for whichevrer other reason, I have been thinking more deeply about guides and how this whole guide writing thing can evolve. I know its an obscure thing to think about but boy do I like thinking.

Its how I discovered the official name for what I was driving to: "Infrastructure as code".

The thought that led to this

I noticed how easy it was for me to fix issues I had written solutions for. The more well written a solution was, the better. I also noticed how easy it was to go through software installations I had written guides for. Both written by me and scrapped from the internet. Also, I noticed sometimes the guides I had missed out on some steps and whenever they did, (a benefit I would never get from relying on online guides), I went online and read about what was missing then returned the processed info to update my guide.

So I was like, "By the way, did you know, since I use this guide very often" – (thinking about some guide I used a lot) – "I could turn it into a script."

Then that led to the aha momemnt or something of the sort. "Ohh, so that's how those Arch installation scripts come to be!" I mean I understand why it could be hard to infect you with the excitement I had about this small discovery but it was reassuring. What seemed to me to have been random self directed efforts to make software installation efficient and predictable turned out to have been what led to documented standards.

But I guess my point in one sentence is. Writing and maintaining guides is a documented good practice. Its the basis for "infrastructure as code". And, that I discovered that from first principles. More about that here.

AI IMO is useless, unless its used for these two things

Documentation aside, onto AI. I don't own a company or do anything that would need for me to be constantly rushing towards producing more and more code. For that, I like that I can afford to not use Generative AI. For most of what Youtube Videos claim it makes easy at least. That is, generative AI not just AI. I mean, as a learner, it so seems theres not any use for AI I can put a finger on.

One could say, GenAI is good at explaining because it rephrases things in terms you as an individual understand and to that I'd say, I always hate it when it pretends it knows more than it does about me. I get way better answers from it when its not piping my convo history into its answer production process.

Another could say it produces code fast and my response would be, I said that already, I want to learn and not rapid produce code.

Or how about quick error correction. I would say, for the most part, I can use stack overflow. It takes time but still (emphasis on the gains from the learning processs and what speed can come say, years down the line).

Reading error messages. I am not doing a lot of CPP now but while I did (during a Computer Security Course) I subjected myself to 2 months of pain. Reading GCC errors, learning how to use GDB. It was hard in the beginning but at some point I could not remember the last time I felt frustrated about a compilation error. My brain had caught up to the pattern. I'd argue this applies across all sorts of error messages. I mean its all anecdotal and I wish it wasn't but I'm sure you can find corroborating articles echoing the same sentiment online.

But , there are actual services AI provides which before it was as accessible as it is today, it was hard to get access to. You know when you understand what you want to say but just can't find the industry terminology for it? You just know what it is but to convey this info to someone else or to , yourself , find corroborating articles online, you need precise terminology? That is, at least as per my reading, tacit knowledge.

I don't think there has ever been a time before AI when one could dump some fuzzy writeup for what they think in a search box and get results precisely pointing to what they were trying to say. Search engines without AI ain't built to handle that. Even those with semantic search enabled. But with AI, if you can explain what you think you want to say, it will almost always clarify what you mean and give you terms you can use to then continue your research with a search engine.

Another use case is documentation. There's a site called deepwiki that I'll link in the cool finds section that does a real good job of introducing the reader to a code base for how articulate it is about explaining what's going on in some messy repo of code.

Control your sources, or lose your mind: A rant on media consumption

I also did some thinking about my consumption. My media consumption that is.

I have had written down a list of sources I find myself always returning to. Sources whose value I vetted myself. I recently decided to upload a copy of those to my site together with other information. I did that for how easy it is for me to revisit most of what I put on this site. Also, as a means to easily share with whoever will find it. I am already liking the effects of that.

The internet , for me at least, does not feel like it has my best interests in mind. It probably once did. But a quick run through youtube while I was giving thought to my internet consumption, questioning if I really needed Youtube as much as I thought I did, had me noticing most of what was being shown to me was not anywhere close to as useful as I thought it was.

And that sentiment I found to be as true for Youtube as it was for any other online platform I frequented for "updates". I have since been consuming less and less even from people I follow. Most of what is on there seems to be opinions about opinions about previous opinions. Most of which aren't contain no actionable advice. You know, the type of content you consume and 5 minutes later you watch it like it its new. (You forget you consumed it already). I don't think its healthy to indundate one's brain with random shallow opinion 24/7.

Trying to train my brain to be more intentional about what I consume. A source driven consumption pipeline. Consulting my sources for any thing I need info about and expaning the source list when none of the sources can provide a valid response. Its hard but its worth it and with time I am sure it will become my normal.

There's never been a better time to do that than now anyways. A lot of fakes online. Non follow based recommendations. Aggressive attention retention campaigns. More about that topic here and here.

Don't wait till its perfect to publish it: A rant on evergreen notes

I was searching online to find if anyone had done added a notes section anywhere on their site like I was planning to when I learnt about the ever green notes idea. Its basically notes that grow with you. Looking at it from the perspective of online articles (since my articles are an extension of my notes), its articles that keep getting updates as I learn more.

Well, what is worth sharing about that? A lot of people intending to write have quite a load of drafts that never make it to the internet. And something about that, can be to blame for why for some, nothing ever does. I'm of the opinion, based on this evergreen notes insight, that publishing one's half ready notes would make it more likely they have "ready" notes to post as articles in the future. There, to me at least, is more push to edit what's public than what I have locally stored in the drafts folder.

The future of this site: The medium for thought can be free thought's worst constraint

One last thought I'd like to log in this one is on my writing journey. Or, my recollection of it. Where I started and my plans for the future.

I moved from writing on the Whatsapp status tab to doing so on substack then moved onto Blogger and now here I am on Github.

Each migration was triggerd by some major challenge. From whatsapp it was the lack of access to old posts and lack of acess to a wider number of 'interested' people. From Substack it was the tight competition incentive that took away from free writing. (for me at least) From Blogger it was the feeling I did not own my writing. It was authored and kept on Google servers by design and I needed some architecture that had ownership built in.

On Github, all the infrastructure problems seemed to have been fixed. Until recently when I noticed another one. I hope it'll be the last one. Don't want to be doing massive changes for ever. I am using Hugo, a static site generator. And using the mainroad theme for the styling. I would like to have a site that mirrors the kind of expression I'd like to have online, and not stiffle it. I looked online for what that would possibly look like and found this very thoughtfully designed site: Gwern.net. Its a lot of work building that am sure so I was like I could add some sections to the blog and solve the problem.

But it now seems that's the only way I'll have what I want, taking on the gwern.net project.

I thought I'd throw that out there was well. The platform is no longer the problem. The site structre I hope, is the last writing infrastructure problem I will have to solve. Maybe until I augment my writing with podcasting.