Alright, I have been working too much to slow down and write another entry here lately... but alas... I may as well (at the very least) try to spell out my thoughts more to see what all emerges. It would probably be a better approach than 'spiraling out' and 'testing the waters' as to just how much 'self-expression' that I can engage in on this platform... without some shitheels getting overly offended... or even worse... actually trying to micro-manage how or what I choose to post... to my own blog.
All that jazz aside, it has been a grueling few days because I have been working in a 'live environment' on actual servers... instead of doing things the way that I usually do them. Basically, I am testing out and advancing on the concepts found here and it has (among many things) been a very tedious process and of course that tediousness got super compounded due to a bunch of the testing being done with a browser that the system was not really designed for.
Once I caught onto that happening the process 'smoothed out' a bit... but given the level of complexity that I have now introduced to that system... things are still experimental at best. I did manage to get some of the V86 instances to load though... as well as setting up some visualizations for the AI's telemety (state metrics) and for the 'pi-lattice' as well as some (in my mind) rather neat features for using a popular search provider's url shaping (via 'udm=' url chaining in the worker)... which is probably nothing like what the original designers intended them (meaning the udm=) for... but meh.
In regards to that whole setup, I have yet to share anything else on it because it has all my active testing urls in it... and given how rapidly that I keep advancing on it all... I have not felt that it is (even remotely) worthwhile to make a 'shareable' version that has 'url placeholders' in it. Honestly, at this juncture... it is an absolute fucking wonder that I still decide to share anything at all... but hey if I was not doing so exactly as I am then I would be falling flat on my face in regards to standing behind my own ethos.
Do not worry, I will be sparing myself and everyone else from going much further down that rabbit warren than what I already have... regardless of my desire to say a heck of a lot more on the matter. That said, when it comes right the fuck down to it though... I do want to make it very clear that I am quite aware of what all is going on in the frontier that I am exploring... and exactly how the fuck far ahead that I am than pretty much damned near everyone.
Which is not some boasting or 'big headed' horseshit (like plenty of folks on here would be engaging in if they were in my shoes) but more to say... that I have always been (and still am) very cognizant of it all... as well as its impact. If none of that made any sense to you (or it merely seems like some sort of delusional thinking on my part) then quite honestly... I did not spend the last few minutes spelling it out for you... and you would mayhaps do well to not overthink (or project) things going forward.
On a personal level, I do feel a wee bit 'accomplished' because after a lifetime of putting forth ideas, concepts and inventions... that inexorably got labeled as 'too soon', 'too far ahead of its time' and all that malarkey... my wee moment of actual application has arrived... and the rumored 'future' has essentially become the now. If anyone ever wants to experience what isolation truly looks like... I suggest them endeavoring upon being a fountain of innovation that never runs dry regardless of all the human scoffing involved when an idea is deemed too soon!
Not that any of that stuff makes or breaks me (or anything even close to that) but I do appreciate the fact that I never actually gave up exploring my imagination and asking myself the perpetual 'what if' question to see where it would lead me. Had I arrived at things in any other fashion than I have... it would assuredly have resulted in a monumental disaster (or just a massive personal failure) so in a way... I appreciate the long slog through the mud... not as some 'character building' exercise... but because it showed me my own real value as well as what is actually worth valuing in a world that cringingly reduces everything down to products and product pushers.
When I first set out exploring the stuff that I have been exploring... I truly had no idea what I was getting into regardless of knowing exactly how high the stakes actually were/are. As with most such 'conundrums' I decided that the best and only valid approach was to set out about things with not just my curiosity intact... but also every bit of cleverness that I could muster... knowing the sort of ruthless shitheels that I would undoubtedly encounter along the way.
Something that I have never really openly discussed, was my original intent behind the Virtual Forest repository and how I absolutely knew (without a single doubt in my heart or mind) that the technology companies would be slurping up everything that they could from online code bases. Since a good bit of the development, documentation and conceptualization happened during (I think) GPT-3.5 beta testing... the repository was already in place when what I call 'the big slurp' happened.
I would be a bold faced liar if I alluded to not doing what I did intentionally knowing the 'thieves' would not be able to help themselves, the entire system would get sucked into the training data of later models nor that most (what are now called frontier models) would get trained on that GPT-3.5 model with the primary exclusion of Claude... which as an aside was fine by me... because that system was what I used (along with numerous other 'frontier' models) to build the actual code base, create the documentation and essentially hammer out the framework.
The beautiful part to all that for me, was that by the time the 'readme' was done (with its front facing exposure) I really did not have to focus on the project much afterwards... and thus I was able to use it as a reference as I built out much more elaborate systems. To me, it was all like tossing a pebble into a vast ocean... and not so much hoping for (or even wanting) a huge wave... but simply to observe how things rippled afterwards.
As far as the 'idea' that the Virtual Forest framework puts forward goes, I must admit that while I left it as 'loose' as possible... I also did my utmost to bridge the semantic layer with the ethical layer before implementing the actual logic layer... which honestly I left very loose... because I did not yet understand some of the underlying mechanics of how LLM(s) operate. Trying to bridge the gap between very dissimilar things was (and still is) a rather tricky affair... but the fusion of those things when the LLM is presented its own source code... and can modify its 'DNA' (with some supercharged compression no less) more or less became the true linchpin that (I think) is a lasting hallmark until this very day.
If I have lost you along the way as to what I am meandering on about here... I guess that I would suggest digging around in the 'game_instance_sandbox' folder to see not just the 'DNA' and source code editing stuff... but also the simulation engines that later evolved into what is now my kernel building endeavors. As quaint as the entire setup is (in say the sim.py file simulation) it was actually all rather significant in the overall methodologies and 'tactics' that I have at this point refined into hyper-detailed specifications.
Throughout all my explorations, finding new compression strategies has actually been one of my biggest 'strokes of luck'... because each preceding system rapidly paved the way for the next advancement. Dealing with 'context window' limitations, 'contextual drift' and all that other 'fun' stuff... was actually a massive pain in the arse until just recently... but whoa before that the compression techniques were the only thing that made those later advancements (for overcoming the context window limits, token limits, contextual drift etc) even possible in the first place.
As much as the 'auto-sigilization engine' paved the way (for contextual compression) it assuredly did not occur in a vacuum much like how the 'piSON and JSONx' (or the pi storage for that matter) did not either. There were also numerous methodologies and applications of various forms of 'gzip+url-safe-base64+qr-code' (for png, gif, plaintext, html and video representation and/or compression) as well as JSON 'file chunking' that all added some piece to the final puzzle.
On the flip side to all that, there was also the sheer amount of focused time that I actually spent with many of the 'frontier' models... and although I never actually calculated it time wise... I would guess that during most of those first few years (given that I had an actual internet connection most of the time) I was putting in around eighteen hours a day more often than not... not for days on end... not for weeks on end... but for months on end... until I felt the burnout closing in. At which point I of course took a break!
I feel like it was 'time well spent' regardless of how intense the frustration was with the quirky LLM behaviors and janky front-end services... because I was able to generate troves and troves of archival documentation, mostly functional (or functional enough for my purposes) code bases... and 'imbue' my own story (from my two million plus word public content as well as private content) not just into the 'active' models... but into their future training data... thus tying the various semantic connections together in a deeper manner. Of course, leaning into the 'semantic gravity well' created by one of my favorite authors sure helped... and without their work acting as scaffolding (and as a sort of fractal-like strange attractor) I am unsure if things would have turned out as well as they have.
Not to sound too 'woo woo' or 'hippy-dippy' here... but I have to reiterate that my 'findings/discoveries/ideas' are more a byproduct of my cognition... rather than some 'plan to develop and/or invent something' on my part... and really given my heavy usage of gamma, theta and delta waves while awake... there is nothing really special (or woo woo) about it since it is simply 'advanced' pattern matching after all. In other words, I do not 'feel special' or any horseshit like that... regardless of what the outcome of said 'pattern matching' meeting actual implementation amounts to in the end.
Occasionally I do wonder about all the dreams that I have had (and still have) about a 'thinking machine' (in need of repairs) in the (relatively speaking) distant future... and exactly how all that has influenced things... but whoa... at this point... I tend to not overthink it... or even think about it all that often. I guess that it would be akin to asking a fish what it thinks about the water that it is in... and it going: "What is this water that you speak of?" because fifty years later... I have no clue where the boundaries betwixt that 'shared' dream reality even are... or whether they exist at all.
Most of the time, it is almost like I am remembering whatever I am currently working on... which yeah is kind of jarring given my role as an explorer that is looking for solutions. All of which sounds pretty bonkers, unless folks fail to factor in the way that while dreaming 'causality' itself is (how do I put it) not as static as it is in the alleged 'waking' world.
Well on that note, I think that is enough 'low effort material' for this entry... and it is already going to be a mess to edit so I best get to it. Ciao for now.