Christian

Keyboard shortcuts + quick jumps | Feature request | Raindrop.io I've been using the beta version recently, and I think that one biggest omissions are keyboard shortcuts. Few shortcuts that I'd found most useful: focus on

looks like it’s planned but not yet a thing. i’m still using pinboard 🙄

raindropio.canny.io/feature-requ...

Likewise. i also find it confusing that a science of science isn’t more mainstream

Nah it is relevant. Appreciate you

ah

bsky.app/profile/bcri...

have seen those, funny enough. what’re the implications in your mind?

Curious. What does it mean to you?

Right there with you

bsky.app/profile/chri...

Apropos of

bsky.app/profile/chri...

Talk nerdy to me

More aptly, what is the science of science in your book? Is there a better definition than Von Foerster’s?

Or the best explainer. Whatever

What’s the best video you’ve seen about what second-order cybernetics means today?

a week ago i woulda disagreed but now lol

Citing polls does more harm than good.

Polls are PR (public relations). Bernays wrote Crystalizing Public Opinion. Polls are about manipulating public opinion.

Do people not realize polls are funded by the people who design the incentives of the polls?

AAAAAAAAAAAAAAAAAAAAAAAHHH

When this "AI" bubble pops, the men pretending they weren't pushing the hype, like "critics" whose position is "AGI is real but LLMs aren't the way," who were in eugenicist and "AI existential risk"🙄 circles, will get specials discussing what they saw coming, when its the women who told you so.

this message brought to you by me trying to help dorian with mArKeTiNg lol

ah the double-binds that bind us

and incentivizing design researchers and user experience designers to do product management’s bidding instead of the appropriate thing lol

part of these usability issues can definitely be attributed to “make number go up by sprinkling more ai on it” corporate leadership styles. which i imagine is in large part because corporate leaders oft neglect to educate themselves on how GPTs work and also the other 99% of AI research problems.

right. search summaries are far from perfect. part of the system prompt is probably something like “have a personality and be unique” lol. i often prefer Kagi search summaries over Google search summaries because of that.

In a chatbot context (ChatGPT, Gemini, Claude, etc.), the GPT is doing more work and iterative transduction based on follow up queries. I call the apps with conversational UIs a chatbot, or just “Gemini” or whatever. The GPT LLM architecture is such a tiny subset of AI research i try not to conflate

Yeah. How i think about it is in a web search summary context there’s GPT LLM architecture that makes up some of the underlying system, but it is a very small part relative to the other systems in play. The GPT is really just summarizing results returned from a search index at that point with RAG.

Calling summary affordances in web search “AI” is a) what big industry hopes we will do because it feeds over-investment, and b) a tiny seemingly insignificant act of anthropomorphism.

But, a big scary causal thicket grows more insidious when we (and industry) get reductionist about what we’re talking about.

I personally believe anthropomorphism of software is Very Unhealthy and has Very Bad nth order effects. Also, using transduction (machine translation or summarization) has valuable use cases, when certain standards and guardrails are respected.

Anthropomorphism, over-investment, improper use, gamification for attention hijacking, context collapse, information entropy, etc., are real societal issues that require attention. And also machine summaries are useful (and in the same category as fuzzy or personalized search, which is not new).