Beam: The Great American Open Model with ReflectionAI Co-Founder and CEO Misha Laskin
Misha Laskin discusses the training of ReflectionAI's open-weight Beam model and how enterprises may shift from renting to owning AI capability.
Listening guide
Sarah Guo and Elad Gil talk with AI founders, engineers and researchers on No Priors. Here are six of their episodes, each with a Chain of Thought conversation to play next.
I host Chain of Thought. We have no affiliation with No Priors: no partnership, no arrangement, nothing either of us gets from this page.
No Priors began in February 2023. Sarah Guo and Elad Gil interview people building AI systems and companies, and sometimes talk through a subject together without a guest.
Recent episodes ask how open models are trained and paid for, what new chip designs mean for inference, how established companies can use AI, and how to supervise agents with access to enterprise systems.
| Detail | No Priors | Chain of Thought |
|---|---|---|
| Hosts | Sarah Guo and Elad Gil | Conor Bronsdon |
| First episode | February 2023 | November 2024 |
| Format | Interviews with builders and researchers, plus host-only conversations | Interviews with engineers, founders and operators shipping AI in production |
| Where to listen | Apple Podcasts | Episode archive and podcast apps |
Details for No Priors come from the sources below, checked 14 October 2026.
Chain of Thought follows AI infrastructure through the decisions of the people building and operating it. Conor Bronsdon asks guests what they shipped, what constrained the work and what they learned. No Priors also asks founders and researchers how their work may change AI products and markets.
Start with No Priors for conversations about the technical and business bets behind AI, from model training to chips and company strategy.
Start with Chain of Thought to hear how a team put one of those bets to work in a product or production system.
You do not have to pick one. The pairings below are where the two shows meet.
Misha Laskin discusses the training of ReflectionAI's open-weight Beam model and how enterprises may shift from renting to owning AI capability.
Then play
How Intercom Cut $250K/Month by Ditching GPT for Open Models | Fergal Reid Fergal Reid, Intercom Fergal Reid of Intercom takes the open-model question into a customer support workflow, explaining why his team moved one summarization task from GPT to a post-trained Qwen model.Walter Goodwin discusses Fractile's model-architecture approach, its full-stack team, key technical bets and shortening the gap between chip design and payoff.
Then play
Tuning GPU Performance with AI Agents | AMD’s Anush Elangovan on ROCm 10 Anush Elangovan, AMD Anush Elangovan of AMD follows the compute question down to GPU workloads, where agents tune kernels with Hyperloom.Michael Lee describes holding-company structures and engineering integrations, with an insurance brokerage and a bank as examples.
Then play
Using AI to Modernize Your Legacy Applications | MongoDB’s Rachelle Palmer Rachelle Palmer, MongoDB Rachelle Palmer of MongoDB gives an implementation case for changing an established business, with her team using AI to modernize old application code.Stefano Ermon explains Inception's use of diffusion for text and code generation and the inference speed and serving choices behind its Mercury models.
Then play
Beyond Transformers: How Liquid AI Is Rethinking LLM Architecture | Maxime Labonne Maxime Labonne, Liquid AI Maxime Labonne of Liquid AI picks up the architecture question with a different model design, describing its effect on inference latency and device size.Ofir Ehrlich and Gonen Stein discuss mapping, classifying and controlling access to enterprise data so that AI systems can use it.
Then play
Data Federation, Not Centralization, Is What Enterprise AI Needs | Jitender Aswani, Starburst Jitender Aswani, Starburst Jitender Aswani of Starburst continues the data-access question by explaining how AI systems can query sources across an enterprise without first moving them into one store.Maxim Bar Kogan describes an oversight layer for enterprise agents, including permissions, context, model rollout and monitoring agent intent.
Then play
We Built Agents, Nobody Built HR | Tyler Akidau, Redpanda Tyler Akidau, Redpanda Tyler Akidau of Redpanda makes that oversight concrete through identity, authorization, observability and accountability for agents at work.Facts about No Priors on this page were checked against these pages on 14 October 2026.