Hattie Zhou
@oh_that_hat
I want to understand things deeply and explain them well. Building friendly AI @AnthropicAI Give me anonymous feedback: https://www.admonymous.co/hattiezhou
we wanted a diffusion tutorial that's gentle and fun because we banged our heads for a long time trying to learn
A beautiful paper that goes through Diffusion step by step, explaining the entire math of it from the beginning.
A beautiful paper that goes through Diffusion step by step, explaining the entire math of it from the beginning.
Claude code feels like magic. Is this how people felt when calculators were first invented??
The reason meditation is naturally appealing to tech people is because both science and meditation are truth-seeking religions. One is searching for objective truth and the other for subjective truth.
In jhana they teach you to notice that the default state of your nervous system is piti, or pleasure, but why? My theory is this is how evolution made us — existence is fundamentally pleasurable and should thus be fought for
Opus: reports of my death have been greatly exaggerated
Introducing the next generation: Claude Opus 4 and Claude Sonnet 4. Claude Opus 4 is our most powerful model yet, and the world’s best coding model. Claude Sonnet 4 is a significant upgrade from its predecessor, delivering superior coding and reasoning.
Things I would work on if I was in academia: - memorization / generalization circuits - dataset interactions - learning dynamic differences b/w PT, FT, RL
Has anyone else had experience with this?
Taking testosterone as a performance-enhancing drug has made me sharper than I was a decade ago, and along the way improved basically every aspect of my life. More below.
Luckily, I think Claude doing most of the "routine generalist work" will also speed up people ability to become deep experts or high-level meta thinkers. It used to be that you have to trudge through hours, days, weeks of grunt work to get a big learning moment -- where you are…
Claude raises the waterline for competence. It handles 99% of general discrete tasks better than the average person - the kind of work that used to require basic professional training but not deep expertise. This creates an interesting split: routine generalist work (basic…
📢Curious why your LLM behaves strangely after long SFT or DPO? We offer a fresh perspective—consider doing a "force analysis" on your model’s behavior. Check out our #ICLR2025 Oral paper: Learning Dynamics of LLM Finetuning! (0/12)
Today, AI can generate tons of code—but how do we know if it's good? That's why we built Sculptor: the first coding agent environment. Sculptor helps you catch issues, write tests, and improve your code—all while you work in your favorite editor.
It’s scary to dream because we don’t want to be disappointed. But if it turned out to be possible to transcend our biological limitations, would you take it? @michaelandregg talking about Eon’s mission at @Abundance360
It’s likely that humans reach immortality via uploading their minds into the cloud + humanoid form before true biological immortality is solved Maybe only 5 years away What type of mechanical/humanoid form will you upload into?
Michael Andregg (@michaelandregg) announcing that Eon is looking to achieve whole human brain scan and emulation by 2030 🤯
Michael Andregg (@michaelandregg) announcing that Eon is looking to achieve whole human brain scan and emulation by 2030 🤯
IYKYK
@michaelandregg publicly announcing Eon’s plan to emulate the first human by 2030 🤯 Last year, scientists achieved the groundbreaking feat of emulating a living fruit-fly, proving that emulation is much simpler than we expect. The future will be so much wilder than we imagine…