Sarah Guo

sometimes people are uncomfortable hiring friends. they don’t want to risk relationships, they’re nervous about screening, feedback, perf, firing, impartiality. all secondary issues if you want to win, hire your friends if you don’t (and win), they will ask why you didn’t

invest in artists https://t.co/uexVNsbalu

some propaganda from last night https://t.co/2DSjGGZXL9 a few more https://t.co/tiDJUvsEDx

thanks for coming, @conviction friends & fam we'll see you on the other side of abundance https://t.co/jblxWLxZR6 https://t.co/n4VLIX0rDm

one of the coolest things about ai products is sometimes you beat your head against the wall and then… https://t.co/sH3kn4ltzp not quite like implementing crud apps

deep research is so useful, even with its inaccuracies

New @NoPriorsPod on a national security strategy for superintelligence — from @hendrycks (with Scale founder-CEO @alexandr_wang, former Google CEO @ericschmidt) topics: AI safety, geopolitical implications, potential weaponization of AI, policy recommendations, state of evals https://t.co/kQyLm2AgB9 full discussion on YouTube: https://t.co/Lavz2yHcWR audio: https://t.co/ZeknSkh6EQ or wherever you get your podcasts more from Dan here:

Coreweave S-1: $15.1B of contracted customer commitments (backlog) https://t.co/1XSaZoBA4e https://t.co/CncPAJJKCe Microsoft #1 customer = 62% of its revenue in 2024 Who is #2 (15%)?

the realization that software only ate a small part of the world over the last decade, and it's about to eat a lot more of it

New dominant question is not “is this the end of scaling” but how does the base model quality interact with test-time scaling / (o1/o3 style RL)? Is it multiplicative (a little bit moar in the base model will you get advanced new behaviors in the reasoning version still)? Otherwise, next OOM of $$$ shifts away from pretraining (with big implications for infrastructure)

We’re either A) entering a AI paradigm where only “high taste testers” can intuit the progress of models, or B) GPT 4.5 is just incremental and uneven improvement

@michelletandler anecdotally many people using AI for conversations they don’t want to burden friends and family with a struggle - this seems valid. at the same time, wonder what that does to relationships long term

One will get further assuming good intent in others until given any evidence to the contrary. trust is earned. being quick to distrust folks who have worked for years to earn that trust is counterproductive. Paranoia makes one hard to support

See you at Upfront Summit! Questions? https://t.co/KFcHSTIOM5

amazing how quickly everyone reoriented around RL

1/ The biggest blocker to effective AI-in-bio is lack of quality data, but @vevo_ai is open-sourcing their gigantic #Tahoe100M single cell dataset (on @arcinstitute's new Virtual Cell Atlas)! I talk w/Vevo's founders and Arc CEO/CTO about "Virtual Cell Models" and significance https://t.co/ldR9S2DLGj 2/ Or watch @nalidoust @iamjohnnyyu @pdhsu @davey_burke @genophoria on @NoPriorsPod YouTube: https://t.co/mMGpb0KN1S or listen wherever you get your podcasts 3/ Access the dataset here https://t.co/IiMECoNuTs 4/ Some takeaways: - this is an "ImageNet" moment for biology - predicting protein structure (Evo, AlphaFold, Chai) was a big AI/biolog…

Coding agent might be the single most IQ-pilled use case of language models today. Claude 3.7 Sonnet smashing the field on Devin evals says a lot

Resurfacing this about neural retrieval, training embeddings models, and the state of RAG from @VoyageAI founder @tengyuma on @NoPriorsPod: https://t.co/1nA6ESh2if

1/ Congrats @dittycheria @tengyuma @sahirazam + the @VoyageAI team on their acquisition by @MongoDB today! As a founding investor, @conviction believed that better embedding and re-ranking models were critical to robust AI-powered search/retrieval (a key enterprise AI use case) 2/ It’s clear that’s the case — and that innovation in this domain was a powerful advantage for customers when combined with Mongo’s ubiquity for real time operational data. Thank you to this amazing research team for having us as business partners (along with @Wing and @CRV) 🙏 3/3 @eastland_maggie for @business covers the news https://t.co/sVWEbz7whx

An interesting paper that paints a picture for energy superabundance (though from 2022, before we even understood that energy access will = intelligence access) https://t.co/SKogt3cf1w

Efficient inference will enable companies to consume more intelligence. No consumption slowdown in sight https://t.co/2MoKhVkdha piece by @mims https://t.co/h2IxOMkdJY

the glamour of parenting: rolling over in your own bed trying to respond to a small child’s cry, stabbing yourself in the eye with a tiny rainbow shrinkydink hand on your pillow, and then tripping over an errant Lego piece in the dark object of violence https://t.co/PY5m6SzPWV

the gap between benchmarks and real-world use cases is…a chasm

Nothing better than user love!! Congrats @HeyGen_Official https://t.co/boMVrXQ0mg

Is ~log(15)x improvement in these benchmarks worth it for ~15x cluster scaling? In the end users will decide— by voting with their feet, and the markets— by voting to give Grok dollars It’s clearly a SOTA model, and folks who threw shade at Grok research team were mistaken https://t.co/Gz3YfnaaIK Still feels like open field with many vectors of potential improvement: expert data SFT, reasoning/RL and creative (more general) ways to generate synthetic data, efficiency in use of cluster (which will matter more the bigger the cluster is!), and user facing product Striking how quickly this was done, and with how few people. “You can’t start a…

new office pet https://t.co/YyaqjCdi5e

We’re extending the app deadline to Tues 2/18 for @Conviction EMBED, the community program for the best early-stage AI startups (10 at a time). Get $1M in cash+credits, and access to tier-one investors+mentors. Our first 3 cohorts have already raised $500M+ to build the future. https://t.co/LxRPDtG1P1 Apply at https://t.co/dlbc5yjpXS And obviously the billion dollar perk is hanging out with me, @pranavreddy @mvernal and the rest of the Conviction team. Come be a guest on “Between Two Servers” https://t.co/1kPubT55Pu

1/3 In the spirit of learning-in-public and transparency to founders, we're sharing the 4 LP letters we've published since launching @conviction. Here's how we see the AI landscape. There's never been a time of more opportunity. If you're a founder, please reach out! Links👇 https://t.co/uGf2OOan7s 2/ Full LP letters here: https://t.co/ANSatm8J1w 3/ Apply to Embed by midnight for community, cash and compute/calls! https://t.co/dlbc5yjpXS

demand is elastic, episode #41247 ❤️ https://t.co/8tIavGFuN9

.@harvey__ai is THE vanguard AI app startup 🆕 @NoPriorsPod w/ Cofounder-CEO @winstonweinberg on: *capability improvement *co-pilots, end to end task completion, and product strategy for AI apps *selling AI to enterprise *hiring philosophy *what lawyers do in 5 years links 👇 https://t.co/4Cpd4X6PwP Congrats to Team Harvey on their $300M Series D! Spotify: https://t.co/3etlkXF5aS Apple: https://t.co/ZCexCdN4mc YouTube: https://t.co/JvYKddYlGq