تغريدة مثبتة
Rishub Nahar
243 posts
Rishub Nahar أُعيد تغريده
Today, we’re thrilled to share that our friends at @TrajectoryLabs have officially launched from stealth!!!
We first met Ronak a few days before Delta was founded in February 2025, and it’s been an incredible journey seeing him grow from his time at Windsurf, to DeepMind, and now to Trajectory.
The Trajectory team has seen Delta’s full trajectory, and now we’re incredibly excited for the world to see theirs!! 🚀
English
Rishub Nahar أُعيد تغريده
We're releasing early results from training Kos-1 Experimental, a Kimi K2.5 checkpoint post-trained on the same medical RL data we used for Kos-1 Lite.
As clinical workloads become more agentic, we wanted a model that pairs medical domain knowledge with tool-calling knowhow.
English
World model" can mean a video that looks like Sora, a robot brain, or a set of abstract embeddings. These are completely different things.
Two AI research lineages quietly merged in the last two years to produce this confusion. New piece on what happened:
moe-capital.com/blog-home/the-…
English
Super interesting interpretability tool from @kamath_harish
Harish Kamath@kamath_harish
Interpreting language models can feel like stumbling through a dark forest - sometimes you just wish you had a flashlight! In our new post, we introduce HeadVis, our latest flashlight for studying attention heads.
English
Rishub Nahar أُعيد تغريده
After George Floyd was murdered in 2020, a farm near my house lowered their Confederate flag and raised the American
It seems small, but it shows that people change, that movements matter
As Dr King said: The arc of the moral universe is long, but it bends towards justice
English
Rishub Nahar أُعيد تغريده
Dr. Martin Luther King, Jr. dedicated his life fighting for equity and justice. He taught us that even in the face of intimidation and discrimination, we must never stop working towards a better future – a lesson that feels especially relevant today.
Change has never been easy. It takes persistence and determination, and requires all of us to speak out and stand up for what we believe in. As we honor Dr. King today, let’s draw strength from his example, and do our part to build on his legacy.
English
This is really well done @bertgodel @thegavinbains
Daanish Khazi@bertgodel
(1/5) New post: "Mismatch Praxis: Rollout Settings and IS Corrections". We pressure-tested solutions for inference/training mismatch. Inference/training mismatch in modern RL frameworks creates a hidden off-policy problem. To resolve the mismatch, various engineering (e.g., FP16 unification, deterministic kernels) and algorithmic (e.g., importance sampling) fixes have been proposed. In this work, we examine how rollout settings (temp, top-p, and top-k) affect mismatch, and how importance sampling corrections bear out in practice. We find that while Sequence-TIS is theoretically optimal, it can succumb to catastrophic variance in long-horizon contexts. Additionally, non-standard rollout settings create subtle mismatch patterns that require careful engineering fixes. Token-TIS with default rollout settings proved to be the most robust setting for long-horizon training.
English
@SushanthRaman @generalcatalyst Absolutely incredible, huge congrats to you and the whole team
English
We raised our $27M Series B led by @generalcatalyst.
Grateful to our customers and team members who made a bet on us before it was obvious.
English
Rishub Nahar أُعيد تغريده
Interpretability + evals are orthogonal bases of AI assurance: evals flag issues, interp traces & fixes them. That loop is the new moat for builders.
Harish Kamath@kamath_harish
Interpretability is simultaneously becoming critical and achievable. Its a green field, has a very low barrier to entry, and there’s not enough people thinking about it. If you’re working on interpretability or are willing to be convinced, I’d love to talk to you!
English
the biggest market dislocation right now is that there is still not a Cava in San Francisco
@cava
English
Rishub Nahar أُعيد تغريده
This is one of the best investor focused shows I’ve done in a long time.
@emergencecap is one of those brands where if you know VC, you know how good they are.
Their track is fricking insane:
$2BN Invested : $8BN Returned in Cash
Salesforce $275.61BN
Veeva Systems $33.04BN
Zoom $25.37BN
Bill.com $20.5BN
Doximity $9.8BN
Gusto $9.6BN
Chainalysis $8.6BN
Box $4.5BN
$387 billion (total market cap of Emergence Companies)
My 7 key takeaways with @jakesaper 👇
English
for someone who has never eaten meat, can you please describe the taste of fried calamari relative to an onion ring - asking for a friend
English
English
Drake Kendrick is the closest I've been to tasting real beef.
Beyond Burger, y'all best watch out
English
