I go on with my small side-business of modding Skyrim. During lockdown it wasn't clear what Bethesda games the paid mods store will cover; had it been only for new titles, I had only vague idea that Starfield would be some space sci-fi game, so my idea was to make puzzle-adventures using the base game assets (and blending in the world game is nice but optional), and puzzle-adventure is a very broad genre in itself, so maybe I would end up making up distinct unique gameplay for each of the modded games.
Recently I came up with an idea for a puzzle-adventure under Playstation mod limitations. That means no new assets, and no new scripts either. Simply, let's treat it as a different game, and see what can be done with that. I researched the built-in scripts, and oddly, every script that could be used to enable a game object was not repeatable. Ditto for scripts to disable an object. The only repeatable scripts were toggle scripts (disable an object if it is enabled, enable if it is disabled). So last week I prototyped some puzzles using that primitive only, and while doing so, I figured out how to trick the engine into making those do-once scripts repeatable, too...
A solo gamedev project; upgrading my free Skyrim mods; thinking about learning vibe-coding for the little "web 2.0" side-project idea of old, seems could be fun to squeeze it in.
Funny they both originated from Sclaveni/Sklabenoi, just one of the tribe of south Slavs. But for the slave trade in middle ages, maybe you guys would call us Slovs (the jury is out on that)
Right now, a video game. If there ever will be AI that does the particular assignment no worse, I'll consider it the coming of the age of AGI. (modulo my game getting sucked into the training set, of course)
It appears that Bloomberg ran such a story this year spring.
I recall I was aware of lesswrong.com when it was still about bayesian stuff and not the hive of basilisk oil salesmen. So, before I stumbled on the fanfic (penned by "Less Wrong"), but for plenty of folks it might have been the other way round.
So, what's the performance of BG3 in Act 3, in streaming?
I had an unfortunate incident of my GPU going kaputt this summer, and while waiting for the warranty repair I tried playing games on an older machine. But not that old, at the onset of lockdown it was considered new. And on that rig, BG3 suffered even in Act 1. The game after a few minutes would go from nice 45+ FPS on ultra to 10-15 on lowest graphics settings, staying in that slow mode for up to hours. It wasn't correlated with GPU temperature, or fixable by some weird setting.
I wasn't aware of GeForce Now. Maybe I would shell out a few bucks. Assuming the savegames would be transferred eventually to my gaming PC with GPU replaced. There, BG3 works fine in Act 3 (except for crashing in one area, fortunately it's a side-quest that can be skipped).
Seems to me a software bug, and I suspect owners of rigs in between encounter it only in Act 3.
Avoiding such horror stories could be an incentive to use streaming, even though GeForce Forge doesn't seem to have a free tier the same way Stadia did, it might make for it in other aspects. (it has a free tier however for 1 hour daily, and right now I have 52 gamers ahead of me waitlisting. I'm curious about BG3 performance there, so I'll wait it out)
Recently I came up with an idea for a puzzle-adventure under Playstation mod limitations. That means no new assets, and no new scripts either. Simply, let's treat it as a different game, and see what can be done with that. I researched the built-in scripts, and oddly, every script that could be used to enable a game object was not repeatable. Ditto for scripts to disable an object. The only repeatable scripts were toggle scripts (disable an object if it is enabled, enable if it is disabled). So last week I prototyped some puzzles using that primitive only, and while doing so, I figured out how to trick the engine into making those do-once scripts repeatable, too...