the first month of CS studies again. the first 1.5 weeks went really bad in terms of giving enough attention to studies. the whole month shoving aside studies a bit too much has been a situation (but there were also very little deadlines of tasks so far… and i only missed 2 out of 6)… although it is hard to notice, seems to go on in the subconscious.
and we are mostly over any boring classes and there is actually fancy stuff being taught. (i realized LLMs is a real gap in my Computers Understanding Chain)… so that is the goal for this semester.
the most fancy thing is, that my execution skill battle seems mostly solved. (meaning we now move on to prioritizing and productivity… that still needs work).
- I have been taking vitamin-D and folic-acid the whole month (yes, that actually changed it a lot (if it is placebo i dont care, it works)).
- for the later 2 weeks of the month a friend lived with me because of construction work at his place. this did introduce the expected issue of making log periods of focused work harder… but the amount of time wasted on youtube, movies or just feeling bad have completely vanished. and finding time to focus improved the longer we lived together (I have 2 rooms after all)
one the day 2026-10-27 was about to go to bed, but had the thought of searching for “torrent like systems, but made for keeping data around”. this led me down an Internet rabbit hole to logos.co and eventually i searched “is there maybe a large community gathering around the problem of keeping data around?” which pointed to the Archive Team → Jason Scott → json’s textfiles.com → The Nomadic Research Labs → their youtube channel → and then this amazing video about a 90s computer bike BEHEMOTH bicycle console walkthrough - 1993 making me like taking a (modern) bike like this around the world (or other vehicle)… is the thing most sounding like my dream… (yes as mentioned prioritizing is the next issue to tackle)
also a very exciting thing is: I finally figured out how to write code with AI Agents without producing slop. the line about “right detailedness of commands to ROVER” from We are Legion (bob) turned out to be ON POINT!!! when giving an agent a rigid structure for architecture, clarifying questions, then the right amount of detail when prompting and tackle small enough tasks at a time… i can make it write code which I can say is ok.
this also means that the idea of MiZe gibing architecture for agents to make software with might work out really well.
speaking of which there will be downloadable binaries uploaded soon.