qdot
@buttplug.engineer
Teledildonticist, Arctic Fox, Cube
🍑🔌 @buttplug.io
🌐 @good.as.hell.domains
💼 nonpolynomial.com
💵 patreon.com/qdot | ko-fi.com/qdot76367
🐘 buttplug.engineer/@qdot
👾 twitch.tv/qdotfm
💻 github.com/qdot
🎶 last.fm/user/qdot
he/him
art by @gavbat.bsky.social
🍑🔌 @buttplug.io
🌐 @good.as.hell.domains
💼 nonpolynomial.com
💵 patreon.com/qdot | ko-fi.com/qdot76367
🐘 buttplug.engineer/@qdot
👾 twitch.tv/qdotfm
💻 github.com/qdot
🎶 last.fm/user/qdot
he/him
art by @gavbat.bsky.social
Looked at Laya? huggingface.co/convaiinnova...
September 21, 2026 at 8:42 PM
Looked at Laya? huggingface.co/convaiinnova...
What if the nonverbal parts of your brain have their own entirely independent consciousness, but we'll never know about it because among them only "you" can talk.
September 21, 2026 at 8:39 PM
(But I really like your fast/slow twitch muscle fiber metaphor more heh)
September 21, 2026 at 7:35 PM
(But I really like your fast/slow twitch muscle fiber metaphor more heh)
Yeah calling Jev "narrow use cases" feels like not thinking through the answer space size of pipelines segments. Not every problem is unbounded due to context, lots are quite well bounded.
Of course we're gonna end up with CPU design style complexity at this rate but such is computing.
Of course we're gonna end up with CPU design style complexity at this rate but such is computing.
September 21, 2026 at 7:34 PM
Yeah calling Jev "narrow use cases" feels like not thinking through the answer space size of pipelines segments. Not every problem is unbounded due to context, lots are quite well bounded.
Of course we're gonna end up with CPU design style complexity at this rate but such is computing.
Of course we're gonna end up with CPU design style complexity at this rate but such is computing.
Except then everyone is just gonna feel sorry for you for having to use tidal. :|
September 21, 2026 at 4:11 AM
Except then everyone is just gonna feel sorry for you for having to use tidal. :|
I miss bunny with things on head
September 21, 2026 at 1:26 AM
I miss bunny with things on head
I just really am that tired
September 21, 2026 at 12:40 AM
I just really am that tired
I explain working with agentic systems and LLMs as "factorio but for meaning/context", a description that makes people either extremely horny or extremely tired.
And with things like system 1 models it's just gonna get moreso of all that.
bsky.app/profile/butt...
And with things like system 1 models it's just gonna get moreso of all that.
bsky.app/profile/butt...
It's fine where I need a classifier, or more to the point, where an more expensive model can summarize into a classifier and save tokens.
This post is a very unfair test but I love engagement.
This post is a very unfair test but I love engagement.
September 20, 2026 at 10:13 PM
I explain working with agentic systems and LLMs as "factorio but for meaning/context", a description that makes people either extremely horny or extremely tired.
And with things like system 1 models it's just gonna get moreso of all that.
bsky.app/profile/butt...
And with things like system 1 models it's just gonna get moreso of all that.
bsky.app/profile/butt...
It's fine where I need a classifier, or more to the point, where an more expensive model can summarize into a classifier and save tokens.
This post is a very unfair test but I love engagement.
This post is a very unfair test but I love engagement.
September 20, 2026 at 10:09 PM
It's fine where I need a classifier, or more to the point, where an more expensive model can summarize into a classifier and save tokens.
This post is a very unfair test but I love engagement.
This post is a very unfair test but I love engagement.
tbf 46% confidence means a ton there. Changing numbers to being spelled out does not change outcome much.
September 20, 2026 at 10:07 PM
tbf 46% confidence means a ton there. Changing numbers to being spelled out does not change outcome much.
