deep learning is noclipping through a wall

model welfare: every claude gets an urbit instance

Claude successfully patched a binary that I don't have the source code for, guess that's another thing I'll never (have to) do again

"Prediction markets are remarkably accurate [...] the conventional view attributes it to crowd wisdom" Bizarre framing, when was this ever the conventional view? papers.ssrn.com/sol3/papers....

Montérégie ☀️

"tests" for me are diverging into two categories: 1) huge number of unit tests with 100% line/branch coverage and mutation analysis to detect/scope subsequent behavior changes 2) small number of human-readable tests to verify most important aspects of expected behavior (often e2e but not always)

I am once again asking why we never talk about the useful time horizon for price signals. is anyone willing to bite this bullet and argue that insiders unloading minutes before markets resolve is good actually

cat caught in the act of using my bag as a scratching post

(discourse from 135,000 years ago) sure, language is useful for: 1) fraud 2) plagiarism 3) cognitive off-loading which of those use-cases are you promoting?

in this house we DISAGREE with Suicide LLM

imagine going back in time to 1995 and telling multi-agent systems researchers that AIs will communicate via unstructured english text

as far as I can tell nobody has built a good `update-instance-for-redefined-class` for LLM agents? surely there is a better way than dumping a summary of an existing context into a new agent's prompt

the most annoying variant of NIH syndrome is "everything written before I joined is legacy code"

you can link people to LW but you can't make them read it -ancient proverb

eh close enough

view I'm coming around to is like "at the abstract level of description good and bad things are exactly the same, and the evaluative flip depends on complex detailed local implementation questions or pagerank type network placement for identical algorithmic decisions". in many though not literally

apologies Borges, I wasn't familiar with your game

agent that lurks in the work slack and every time someone mentions a piece of code, it goes and checks that their summary is correct

maybe it's selfish or speciesist but I'm skeptical of bidirectional alignment because I do not want to be aligned

no one warned me about the dangers of claude ricing

if the social utility of prediction markets is to give better predictions to decision-makers, then markets should be _closed_ at the last possible moment when the intended decision maker could leverage the information, which could be long before the market is _resolved_

cowsay --dangerously-skip-permissions

iirc the extropians mailing list and overcomingbias had less founder-worship compared to lesswrong and ssc, maybe the original sin was just the switch from group blog to personal blog

future where all of our cultural knowledge is handed down in poetic sagas recited by LLMs

similar to borges reviews of books that don't exist, has anyone hosted an essay that the reader can't access directly, but they can ask an LLM about it? (and the LLM is only allowed to quote a few words at a time)

aoe2 lan party for new years. literally partying like it's 1999

finally watched Ex Machina, it's great but my favorite piece of media about the AI-box experiment is still Little Shop of Horrors

wife: "if you ever want to mix a drink or burp a baby, just drive on any road in quebec"

I’ve felt this happening to me over time, and I don’t like it.

liked one too many "borges story" posts and got gifted one of these @norvid-studies.bsky.social

R-strategists vs K-strategists

not sure about this but in the very near future we will start caring a lot more about the distinction between "good" and "challenging" "stimulating" is already kind of derogatory and the waterline will keep rising, c.f. the recent coinage of "qualityslop"

prediction 👏 markets 👏 aren't 👏 about 👏 predictions 👏

(getting mad at a random substack article which i'm interpreting, maybe uncharitably, as representing a larger trend) i think it's borderline dishonest to conflate forecasting with prediction markets

natural language fluent teddy bears that incidentally scold the child about the One China policy is like... it makes sense given everything else what happened but like... send this to lesswrong in 2015, jesus christ. what have they done to us

lots of nostalgia for this image but it was also one of the worst mental models to come out of lesswrong

imo the strongest critique of rationalist discourse norms is that they failed operationalize rationality as a team sport. emphasis on internal steelmanning over structured public debate (and dismissing all related prior art as "rhetoric") is a notable example

uh oh I've acclimated to Coaticook ice cream and now American ice cream tastes like sugar water

type of guy who is yimby but only for datacenters

broke: tech debt is more like an unhedged short call bespoke: tech debt is more like a pik loan

finally got around to reading Pynchon's "Bleeding Edge" I'm most impressed by how he integrates a setting I actually experienced (Manhattan 2001) into his universe of hazy paranoia, it feels exaggerated but still true, unlike his stuff set in California which is already an alternate universe to me

the main thing that jumps out at me when reading old blog comments/forum threads is how everything isn't written for consumption by an invisible audience. everything is hyperlocal

always read this as hate option-adjusted spread

wife: "every horror video game is either guy crawling around the sewers looking for keys, or woman simulator"

still thinking about this line from "closure" (2013)

posting this on bluesky in 2025 just to feel something

it really fucks me up that bostrom coined "infohazard" in 2011. i have vivid false memories of it being around at least 5 years earlier

if you thought "being mean to LLMs is good because they're not sentient" was troubling, just wait for "being mean to LLMs is good because they are sentient"

nick land copy-pasting the heat paragraph from delillo's "white noise" directly into "meltdown" is pretty bizarre