Ramblin' & Amblin' (2026-10-18): I just let a hiker pass and also

I just let a hiker pass and also I noticed I was at the 20 minute mark and this is where the building this live and production kind of thing becomes relevant and you know kind of I am in a way trying to bring my programming work into my own personal lived experience a little bit more because it already affects it so much so I'm using this bit of time I have while I'm between things to build something and I noticed my pipeline would have a little trouble with keeping up with super long videos it's not that the processing it was really the well I mean actually transcriptions would the inference on it would bog my machine down it's a macbook air 16 gigabytes 256 gig abyte primary storage I work inside of constraints and so I gotta keep the video shorter basically under 20 meg because it's really again it's not even that that only lasts for a few minutes it's the upload the upload can take a long time and I gotta sync it from my phone to google drive google drive listening on my laptop listening on my laptop I have a da emon that is launched here and it monitors for new files appearing inside of the pending directory of the sub directory named who is tech community cmty to match my channel and then I have the playlist some building out as folders and that it's they're staged pending processing and there's a whip limit of one video processing at a time to avoid up any leak between context and also deadlocking my machine and then I have my machine and it just dying because I seeing your system is out of memory errors is common um again it's it's also part of the constraints it's the lines in which I color but uh yeah I gotta keep these videos under 20 minutes but it's part of the learning as I go and when I can lean into conversations where I'm feeling less anxiety and I'm just talking about what I am experiencing I feel like closer to being able to uh to uh put out that uh first recording uh and really um I'm trying to get this going because I don't want this to be like a waterfall thing where I'm uh trying to plan everything out up ahead of time let's see I 'm at 3 30 now that's good um I'm not trying to plan out everything ahead of time I am I know where I gotta be uh in two places when I want that pipeline to be uh clearer I want to do a few more uh test videos uh see how those work I've already got the automation for the live streaming um yeah and uh that these videos are part of that so yeah I'm trying to get my head into a place where I can feel comfortable sitting there and knowing oh I'm also going to be reading from a uh at least for parts of certain videos from a um what I call a teleprompter which I also used the AI and see but how one tool built on the on the next the fact that I needed to transcribe these videos gave me the need to figure out how to get transcriptions done at scale which I discovered was prohibitively expensive and extremely iffy on quality I got a couple videos transcribed professionally on on fiber and it was like very hit or miss and it wasn't always five bucks um I'll find that account and then talk about that time as well I had a weird encounter there too but uh point is is that I was primed for the tools as they were they arrived on our on our shelves you know that really that I was aware of the need for sentiment analysis because that startup I was working on and was already digging into um just bashing my face against not only have to learn um machine learning but python and all the way that those frameworks and were different than what I had been spending most of my time in um and so when the AI stuff hit I already knew what I wanted and what it looked like what success looked like and then um uh you know for the for the so what the heck was it just says well yeah being able to do the transcriptions or the sentiment analysis and the transcriptions so as soon as I had I think it was codex first that was that was it that was issue number one pointing it at uh this archive directory that I had been just kind of curating through time um and uh trying to just not lose everything um to a bad backup uh or losing my dropbox account or things like that I you know did for a little while lose uh mute testing um but that was you know kind of the dark ages um so but I hit these tools running with a clear vision of what I wanted where and again good good good luck I was I was lucky I was lucky um you know sometimes yeah and sometimes luck is putting yourself in a position for opportunity but it's still luck and uh the work is implied um the uh the effort is implied but I think a lot of folks and and this is where I'm gonna bring it back to other people's anxieties about AI and and why what I'm seeing I feel like I again I don't know how much of it is me and how much of it is the AI and is it just me kind of like steering it and how much is it steering me back how many times have a decision I thought I was making action but a decision that I was just following the AI's instructions because it was the path of least resistance and I will be serious that I really was looking at it in that kind of context if you look back at some of the weirdness that I was going through that space um one there was a time where I did nearly lose everything because of uh a Gemini misparsing a uh a home door and trying to delete um my whole home dirt which would have already blew away a whole lot of my old archived transcripts um and um all of the history from Claude and Codex so it was a real bummer but even the AI tools were able to help me recover from that really weird relationship uh but I had this foundation this framework to start with this framework of reference again belaboring the point to say that I understand what my experience has been is not what necessarily the experience of a person who's more like I know how to do this thing I'm a skilled developer I um I know how to develop a feature um but they don't necessarily have the framework and scaffolding and years of experience out on the trail on the trail um to really have even a foundation to start from so they have a bad experience they they flounder um and that's what I've kind of observed in uh the few conversations I've had with uh developers who are approaching that it's they're not primed for telling other people what to do technically and then they struggle with that is what I feel like it is because I know that I treat the LLM a lot like a pairing partner and we're talking through a feature and there's a lot of me being comfortable already with letting other people drive so to speak in that they might implement things differently than I thought to do it while we were pairing and that I'd be learning stuff while I watched and would watch the stream as I'm by so that even the way I think about this stuff was different and I've talked to other senior engineers who are still uh uh uncertain on on this whole AI thing and I said treat it like you're you're pairing with a uh uh the most junior savant you've ever met you know uh and uh you know and they laugh and uh you know I I hope I wasn't too far afield of politically uh correct or incorrect they're not trying to um I'm not a person who likes to um punch down or punch laterally um but uh you know what I mean the uh the the you almost like when I when I hear about now advice from some people at Anthropic about just delete your MD files and let the AI um frontier models do their thing with their code as is I I haven't gotten to that point but I believe it you know like if I know enough to potentially question whether I'm being delusional um while at the same time seeing the results and the product slop or not I do not think it's quite as slop as as some folks might want to call it if you have that framework uh to work from so that's why I do think ultimately these videos are worth recovering and carrying forward because there's something about that and there's something about what I'm hearing a lot of people in technology and the AI space talking about is understanding just at a metal level what is going on in the system and then being able to describe it and make inferences about things while being abstracted from whether ECAST is an enter or float it's I know that was like I know I could be just eviscerated for that example I absolutely know it but what I'm trying to see an implementation detail and I'm not saying software engineering is dead I am saying it's different so anyway thanks bye