^ this

all the other stuff around it is bloody unhinged, including how kent reacted when someone promoted his bot to reply in a manner that didn’t fit his deluded conception of it

from a distance, it looks like the kind of unhinged that doesn’t take much to progress into being dangerous

slowly burning through the latest BTB eps (on epstein), getting to the point in ep3 about the discussions/events around ~2016 and just … god

the remarks and then-recent actions, wrt affecting the internet and the social technological comms structure of humanity as a whole, and then rapidity of a bunch of shit starting to turn to shit 2014~2016 (as I’ve remarked on in previous posts)…

I’d want to see more threads checked into and researched in depth (and I know that some stuff (partly?) also had their own drivers), but fucking hell there’s a lot of apparent overlap. dunno if I can take on that investigation (my stats derivation/calculation skills border on a warcrime), but other than that would be interesting to see some analyses

and the W is for “wailing”

just 100 more billion bro please bro just 100 more billion bro, we’ll solve statistics bro you’ll see bro just 100 more billion bro

(and then in a year it’ll be 500b…)

ye, and they only notice when e.g. food production (that impacts them) is interrupted. outside of that you get a brief fascination with disaster porn, but never enough to do anything concrete

there’s a couple chairs I’ve seen where they’re designed for a different style of sitting, stuff coming from medical and similar worlds. I’m not familiar enough with them to have picked a specific example, but I’ve seen a couple of baytypes pick these up because someone else had one

nasb, video from a climate scientist going over the claims by promptfondler ceos

rolls 1, 1

I behold the paperclip calculator, and tremble in fear

the only revolutions these people are capable of is spinning their office chairs

…when it’s one on a swivel instead of some fixed-position ergo thing they acquired after reading a thotleader blogpost

I regret to inform you that the promptfans have a new fucked up way to thotpost:

https://nvie.com/posts/15-years-later/

Other than that, I find this whole thing mostly very saddening. Not because some company used my diagram. As I said, it’s been everywhere for 15 years and I’ve always been fine with that. What’s dispiriting is the (lack of) process and care: take someone’s carefully crafted work, run it through a machine to wash off the fingerprints, and ship it as your own.

So take your codebase, compress it as a zip at normal compression level, and then think how large the output problem space is, shrink it down quadratically, and you have a good estimate of how much ctx you need for LLMs to solve the hardest problems in your codebase at any given point during token generation

wat

I can see what they’re going for but that seems … wildly guess-y?

Something like “innovations in parasitic enclosure” may perhaps be a phrase that can give a handle on it, yeah

ime much like the SAFe diagrams, this diagram is all over a certain type of “this is how your corporation should be developing software” thotleader posts

(although I imagine the lag 2~3y all those heads have pivoted to promptpraise)

I guess that doesn’t emphasise the “innovation” aspect much

parasitech?

okay I can’t argue with that outcome

morgin’ all muh featues

Timn

yes, I certainly do know how to handle software development over Timn

it is actually kinda incredible that this shit has invented a way to be terrible that we can’t actually easily riff off by what’s expressable in unicode. an unholy clusterfuck of what would otherwise be be joked about as keming (but isn’t because it’s straight-up an artefact of the process used to encode visual data from source data, badly), a mindless automaton outputting garbage, and then also the shitty model

and people keep telling me this shit is good

in today’s news about magical prompts that super totes give you superpowers:

We introduced SKILLSBENCH, the first benchmark to systematically evaluate Agent Skills as first-class artifacts. Across 84 tasks, 7 agent-model configurations, and 7,308 trajectories under three conditions (no Skills, curated Skills, self-generated Skills), our evaluation yields four key findings: (1) curated Skills provide substantial but variable benefit (+16.2 percentage points average, with high variance across domains and configurations); (2) self-generated Skills provide negligible or negative benefit (–1.3pp average), demonstrating that effective Skills require human-curated domain expertise

I am jack’s surprised face

…and given I have other yaks, I shall not step on my “software and tools don’t have to suck” soapbox right now