Forget Parameters: How Behavioral AI Beats Raw Scale
Poolside AI's Eiso Kant argues 'behavioral intelligence' in models like Laguna S beats raw parameter count for complex tasks. Rethink your LLM strategy.
The AI Frontier: from open weights to open research — Eiso Kant, Poolside AI
With swyx, Alessio Fanelli, Eiso Kant · Sunday, July 26, 2026
Eiso Kant, CEO of Poolside AI, discusses his company's mission to foster an open and competitive AI landscape, moving away from an oligopoly of intelligence. He highlights Poolside's 'Model Factory' approach to rapidly build and iterate LLMs and shares insights on how 'behavioral intelligence' in smaller models can offer significant capabilities, challenging the exclusive focus on raw scale. Kant also offers strong opinions on the future of LLM training, tool calls, and the ethical implications of open versus closed AI development.
Key takeaways
- Poolside AI's Laguna S, an 118B parameter model, solved complex coding challenges like Erdos 397, proving smaller models can achieve results previously expected only from much larger ones. Read more →
- Poolside AI raised $500M, making a calculated bet on diversified AI at a time when many investors doubted the strategy, committing to an open competitive AI landscape. Read more →
- For LLM builders, Reinforcement Learning (RL) time, not pre-training, is now the critical bottleneck. Poolside AI CEO Eiso Kant points to batch size constraints as the limit for scaling compute. Read more →
3 articles from this episode
Eiso Kant: Open AI Is a Geopolitical Fight for Your Future
Poolside AI CEO Eiso Kant warns against closed AI "oligopolies," likening restrictions to a sci-fi dystopia. Open AI isn't just code, it's a geopolitical imperative.
Poolside AI's Kant: RL is LLM's 'Biggest Wall Clock Bottleneck'
Eiso Kant of Poolside AI says LLM reinforcement learning is bottlenecked by batch size. Disaggregate hardware and use low-precision for major speed gains.
More Latent Space episodes
- Inside the Team That Killed Mandatory Code Review — Quinn Slack, AMPSep 13 · 6 articles
- The Inference Frontier: from 100 to 10,000 tokens per second — Sean Lie, Cerebras CTOSep 6 · 6 articles
- Trillion Token Context. No, Really — Anima Anandkumar & Benedikt Jenik, Accelerated UnderstandingSep 6 · 5 articles
- 🔬 Why Transformers Hit a Wall the Moment Physics Shows Up — Anima Anandkumar, CaltechAug 30 · 5 articles