MIT's Philip Isola on the Era of Robot-Use Agents: General Models Controlling Any Robot
ycombinator · x · 2026-09-27
MIT professor Philip Isola argues in a recent essay that we may be entering the era of robot-use agents: general-purpose models that can control different robots, write policies, and learn new physical tasks with little or no robot-specific training. One of AI's biggest recent surprises has been how well coding agents generalize beyond software.
YC's Decoded podcast hosted founders of WaddleLabs and Robocurve, two startups at this frontier. Key points from the discussion:
- Robot-use agents should control robots via code, VLAs, or direct actions, picking whichever is optimal for the situation.
- There's a serious lack of evals comparing robot-use agents; the guests frequently ask others to benchmark new models.
More from Embodied
- Tesla FSD miles about to hit 15 billion, a scale no rival matches — Vjeux · 2026-09-27
- Amazon delivery drones overwhelm a Texas suburb, sparking NIMBY backlash — Polymarket · 2026-09-27
- Muse Charm is to phones what phones were to computers, argues Meta researcher Matt Deitke — alex_conneau · 2026-09-27
- Open-source VLA + RL pipeline ranks 1st of 62 teams folding garments on a two-armed robot — chris_j_paxton · 2026-09-27
- DeMiAn: dense language annotations boost robot policy learning, cut compute 62% — rajammanabrolu · 2026-09-27
- Insider confirms Muse Charm direction: hardware existed, OS3 completes the agentic stack — SimonBalmain · 2026-09-27