How to Build Long-Horizon AI Agents — Mitch Troyanovsky, Basis – The MAD Podcast with Matt Turck

How to Build Long-Horizon AI Agents — Mitch Troyanovsky, Basis – The MAD Podcast with Matt Turck — Transcript

Market views

Bullish

AIMODELS— Long horizon agent capabilities are improving rapidly, with the speaker believing capabilities roughly double every 7 months. Multiple paradigm-shifting model releases (Opus 3, O1, O3) have enabled increasingly autonomous agents that can work for hours or days.
AILABS— AI labs are making significant progress on training models specifically for long-horizon agent work, including theory of mind capabilities and agent orchestration. Models are being trained on high-quality data for specific domains like coding, leading to better practical performance.

Not investment advice.