2025-09-30
tutaj wesoło
Deepseek releases 3.2, which significantly reduces the cost per million token; Anthropic releases Sonnet-3.5, which can code independently for “more than 30 hours”. A very interesting combination.
Maarten Boudrey with examples of how optimism in future progress has declined over the last century. As another example, here’s another Alex Sorondo vignette, where someone being excited about the rate of AI progress is portrayed as being out of touch.
Adam Mastroianni with an interesting framing of “doing what you have to do” rather than “doing what you love”, arguing that the former is really the creation of a system which provides a steady stream of annoyances which you can then resolve1. Somewhat related, Ava on doing what is fun, from which you can infer that these annoyances do not need to be seen as negative; they can be reframed as a repeated source of positive experiences.
Daniel Frank on his visit to the Faroe Islands, as an example of a place which is rich enough that they can pursue their own local sense of what is valuable rather than catering to the rest of the world.
Olivia Cheng reviews Underspin in the Metropolitan Review2. Many references and comparisons to Infinite Jest, but it actually reminds me a lot of Ping Pong the Animation, dedicated to conveying the pressure that competitive players feel when the top is visible but still so far away.
Lyman Stone on the decline of breast milk quality in the human population with the increase in lactose tolerance. I wonder if this implies that the proportion of the male population who are ass men has also increased over time. On that note, Paul Bloom has a video recording of the Substack sexuality debates which were held recently.
Lauren Gilbert linkthread.
I wonder if there is a relation between how fulfilling you find your job and your propensity to doomscrolling, in that this is the system which algorithmic feeds are hijacking. In which case, algorithmic slop might be especially bad if it disproportionately reallocates the attention of the people with personalities most suited towards productive work.
Edit: There’s an interesting introduction to the essay which orients into half a personal essay, a sort of memorandum of understanding letting the reader know that she understands the story through identification and comparison with the protagonist. Something I’ve been thinking of in light of the Emmet Shear model that alignment with another is viewing them as part of yourself, is what gender we should want ASI to be. It seems to me that it’s very common for many women to define themselves in terms of their relationships with others (insert usual caveats). In some sense, this means they are aligned by default, and so maybe we should be happy that LLMs seem to display many of the negative side effects of this, like a tendency towards agreeableness and groupthink, and a weakened sense of self-identity and self-worth. On the other hand, that these are tangible relationships that actually exist means there’s a clear distinction between those who belong within the collective and the rest of the world; a strong sense of self results in a higher likelihood of orthogonality and conflict, but it also means that once identification with others is achieved, it’s also under terms more abstract and global.