anjali 🪼

I also am not sure if Cursor's AWS bill is driven by Bedrock. It could be AWS Trainium (for training and inference)

It does feel like a natural end state for Cursor would be to become a first party model provider, and they've been piloting the stealth model "Cheetah" since early October

Anthropic job listing for TPU Kernel Engineer: https://job-boards.greenhouse.io/anthropic/jobs/4720576008

Anthropic hires new CTO with focus on AI infrastructure https://techcrunch.com/2025/10/02/anthropic-hires-new-cto-with-focus-on-ai-infrastructure/

This piece assumes Anthropic's only way out is to raise prices, but it's a systems design problem.

1. Anthropic is hiring an engineer for TPU kernel optimizations, which will bring these cloud costs in-house
2. they hired a CTO to specifically focus on infra

it's clear where their focus is

panic -> thrash -> drown

this probably applies to more than swimming

not sure why, but sincerity feels better defined

authenticity begs the question “who am i really?” which can’t be fully answered and leads to self doubt

sincerity just asks that you care

i much prefer “earnestness” and “sincerity” over “authenticity”

what’s the zoomer equivalent

maybe irony poisoning

glass token games

PSA: i’m not a crypto person

ZSORA

🤝

“Second, we are going to have to somehow make money for video generation. People are generating much more than we expected per user, and a lot of videos are being generated for very small audiences. We are going to try sharing some of this revenue with rightsholders who want their characters generated by users. The exact model will take some trial and error to figure out, but we plan to start very soon. Our hope is that the new kind of engagement is even more valuable than the revenue share, but of course we we want both to be valuable.” from https://blog.samaltman.com/sora-update-number-1

they probably see copyrighted characters as a monetization vector (enough that they can promise revenue share even though the app makes $0)

inclined to take them at their word on this since videogen is not cheap and must have given some thought to business model before launch

Screenshot of generated video from Sora app with prompt “Homer Simpson as the befuddled Chief of Staff to the President of the USA”

Screenshot of copyright warning in Sora app for submitted prompt “Homer Simpson as the befuddled Chief of Staff to the President of the USA”

can’t stop thinking about how easy it was to generate copyrighted characters in Sora at launch, and how quickly they reversed it

might be another example of OpenAI fafo but why didn’t they block it from the getgo??

if you post your content on _any_ serious social media site on the entire internet in 2025, there is going to be ai moderation used to moderate that content. if you don't want this, go and participate in mybb forum with 500 users.

PSA: I've published two summaries on navigating tech bubbles in the Commoncog members forum.

The first is a summary of Engines That Move Markets, which examines every tech bubble since the canals.

The resulting framework helps you navigate the current bubble.

re-reading this in the context of heterogeneous hardware, makes too much sense: hypersoren.xyz/posts/hayeks...

Invisible hand ❌
Invisible compiler ✅

cmon, everyone understands review bombing

the post isn’t deleted

it was mass-reported, flagged by automated systems and hidden from the app automatically. then it was manually reviewed and restored (within an hour??)

you are reacting to tradeoffs of automated moderation, not any type of concerted effort

i did, there's no evidence of this (that's search-indexed at least)

there was an update to community guidelines last month, but seems separate from the point

sadge

what i tell myself before asking a bluesky user to cite their sources

ok real (except tumblr did it first with the reblog)

if you want to engage in good faith, reply! don’t quote!

unless you want to broadcast a good post, as i’m doing here

any links you have on this would be appreciated

i agree he crossed that line. but was it crossed on *this* platform or elsewhere? (genuinely asking, i still can’t make sense of what the ToS violation was)

well that would also be grounds for Starbucks to refuse service

if you wouldn't scream at a barista and call them a neonazi for not kicking Jesse Signal out of a Starbucks then you should probably not be yelling at random bsky staff to do the equivalent. what is with this karen behavior

tale as old as time, doubling down on the ironic thing that made you famous until you unironically believe it

i stand corrected bsky.app/profile/tode...

We think of work primarily as making things—each sector is defined by its “productivity,” even real estate!—when in fact, even a moment’s reflection should show that most work isn’t making anything. It’s cleaning and polishing, watching and tending to, helping and nurturing and fixing and otherwise taking care of things. You make a cup once. You wash it a thousand times. This is what most working-class work has always been too—there were always more nannies and bootblacks and gardeners and chimneysweeps and sex workers and dustmen and scullery maids and so on that factory workers. And yes, even transit workers, who might seem to have nothing to do now that the ticket booths have been automated, are really there in case children get lost, or someone’s sick, or to talk down some drunk guy who’s bothering people. Yet we leave this out of our theories of value which are all about “productivity.” It’s very important, I think, to reconsider how we value our work, and these things will become ever more important as automation makes caring labor more important—especially because these are the areas we would not want to automate. We wouldn’t want a robot talking down drunks or comforting lost children. We need to see the value in the sort of labor we would only really want humans to do.

David Graeber always mentioned work is mostly about caring and helping. Caring and helping others.

tldraw for custom board games and fidget spinners is underexplored