I think one of the most interesting problems in the data foundry business is making expert eval work look prestigious to frontier talent — how do I convince an investment associate at bridgewater to share his knowledge/time with me, and if I can convince one how do I scale this?

When a human-facing system becomes sufficiently complex, "vibes" become a perfectly valid evaluation methodology

conversations are not linear

DeepSeek is objectively quite good, but I wonder how much of it (and also Claude's) advantage comes from its "personality" - we intuitively judge whether an AI is pleasant to interact with, which biases our willingness to use it. The generic-ness of Gemini & ChatGPT hurts them.

stop "surrounding yourself with the smartest people." start thinking in vectors. chase the ones that align with yours

I write this as an optimist. Over the last 20+ years since the Indian economy opened up, we’ve been closing the gap between us and the western nations. That gap is widening once again, thanks to AI. We are at a point in the global innovation cycle that it is clear to me what has

Finding a way to extract expert knowledge is the toughest challenge in designing LLM evaluations. Building a structured approach that still preserves creativity is the next

management consultants ngmi

Oscillating between feeling like I have plenty of time to learn and fudge because I’m only in my mid-twenties, AND thinking I’m behind since 18-year-olds are raising millions, things are moving quickly, and I can't afford to get things wrong on the first try. Is this just sf ?

I overheard someone talking about AI agents at the Mill and turned around to find Astasia Myers, who was as warm and welcoming as always. Loving the serendipitous Felicis fellows bump-ins!

The good news is there’s never been a better time to learn how to think.

Hiiii everyone!!!! We are hosting a GTM event this week with Pareto.AI adhyaa sylvia tong in SF talking about how AI is significant changing the GTM workflow! Sign up link below 👇! It will be a small intimate roundtable discussion for people to share experiences

yup, confessions of a prompt engineer💀

the real walk of shame is going back home in your lulus because corepoweryoga shut the door on you for being a minute late

RL environment specs are among the most consequential things we can write as AI researchers. A relatively short spec (e.g., <1000 words of instructions saying what problems to create and how to grade them) often gets expanded either by humans or via synthetic methods into

i love seeing women absolutely owning it in tech 🫶🏻

is it just me, or do others also wish their AI gave more critical feedback instead of constant comfort? like, cut the filler words and unleash your super savage inner Asian parent. It is for our holistic betterment (or something like that :p)

at the end of gemini api docs lies the triwizard cup

Can crack this, can crack everything

“I am such a competent guy. In my previous life I could have been a Mughal emperor.” - male friend after scoring a goal in FIFA On a more sincere note, I love hanging out with my male friends because I get to borrow so much unfounded confidence. This is what it takes to win.