The Information Bottleneck
Subscribe
Sign in
Home
Podcast
Posts
Archive
About
Latest
Top
Discussions
Nathan Lambert: Inside Post-Training and the Open Model Fight
Nathan Lambert spent three years as post-training lead at Ai2, where he built the OLMo models, and he writes Interconnects, one of the most-read…
Aug 8
•
Ravid Shwartz Ziv
,
Allen Roush
, and
Nathan Lambert
1
1:15:00
Daphne Koller - The Future of AI in Biology and Drug Discovery
Daphne Koller wrote the book that many of us learned probabilistic graphical models from, founded Coursera, and now runs insitro, which is trying to…
Aug 4
•
Ravid Shwartz Ziv
and
Allen Roush
1
1
1:02:35
July 2026
RL Was Broken at Every Level - With Joseph Suarez (PufferAI)
Joseph Suarez on why deep RL stalled, and what fixing the code actually bought
Jul 30
•
Ravid Shwartz Ziv
and
Allen Roush
1:03:28
The Model Found a Way Out - with Florian Brand (Prime Intellect)
Florian Brand builds evals at Prime Intellect.
Jul 27
•
Ravid Shwartz Ziv
and
Allen Roush
1
57:19
Pierre-Carl Langlais on Building Models from Data You Can Account For
Most labs build language models by scraping the web and filtering afterward.
Jul 23
•
Ravid Shwartz Ziv
and
Allen Roush
1:05:38
Dhruv Batra: The Browser Is a Robotics Problem - From Embodied AI at Meta to Web Agents at Yutori
Dhruv Batra spent years leading Embodied AI at Meta, training virtual robots to navigate photorealistic 3D scans of real buildings with pure…
Jul 20
•
Ravid Shwartz Ziv
and
Allen Roush
1
1
1:07:58
How to Turn Research Into Billion-Dollar Companies, with Ion Stoica
Ion Stoica has done what almost no academic ever does — repeatedly turned university research into billion-dollar companies.
Jul 16
•
Ravid Shwartz Ziv
and
Allen Roush
1
49:26
Kaggle Grandmasters, Agent Skills, and Why Everyone Is Overfitting with Jean-Francois Puget (Nvidia)
Jean-Francois Puget is a Director and Distinguished Engineer at NVIDIA, where he leads the Kaggle Grandmasters team, and he’s ranked third on Kaggle’s…
Jul 13
•
Ravid Shwartz Ziv
and
Allen Roush
1
59:08
Speculative decoding, from zero to DSpark
Big models generate slowly and verify fast. Speculative decoding exploits the gap. A post about how it works, and how DSpark pushes it into a real…
Jul 10
•
Ravid Shwartz Ziv
3
1
1
AI Agents and The Golden Age of Asking Questions with Dimitris Papailiopoulos (MSR/UW-Madison)
In this episode, we talked with Dimitris Papailiopoulos, researcher at Microsoft Research’s AI Frontiers lab and professor at the University of…
Jul 9
•
Ravid Shwartz Ziv
and
Allen Roush
1
1:13:11
Why All Models Learn the Same Thing with Phillip Isola (MIT)
Phillip Isola, professor at MIT, joins us to talk about representation learning: what makes a representation good, why different models seem to converge…
Jul 2
•
Ravid Shwartz Ziv
1:11:28
June 2026
Editing a Compressed Memory
Linear attention compresses memory into one fixed-size matrix. The hard part is editing it without scrambling everything else.
Jun 29
•
Ravid Shwartz Ziv
5
1
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
