What is METR?
METR is a nonprofit AI evaluation organization that assesses the capabilities and risks of advanced AI systems. As of Apr 2026, Nathan Labenz noted that METR’s estimated doubling time for task length was under four months, implying a greater than 8x increase over a year, possibly 10-12x. In May 2026, Ben Todd said METR has 20 useful projects they would like to do but only capacity for two or three.
Where it appears in the record
Every line below is attributed to a named speaker.
METR chart shows AI task-length doubling time of under 4 months, implying an 8-12x increase over a single year.
“It seems like everybody has kind of estimated a higher number than the trend so far suggests, which is like kind of a little under four months doubling time for task length, which means we'll be greater than 8x, maybe somewhere in the like 10 to 12x over the course of just a year.”Host (Nathan Labenz) · 26 Apr 2026
METR has 20 high-priority engineering projects it wants to staff but capacity for only 2 to 3, illustrating a severe talent bottleneck at a leading AI evaluation org.
“They have 20 really useful projects they'd love to do but they only have capacity to do two or three of them.”Ben Todd · 26 May 2026
METR (AI evaluation organization) is worth watching as a talent-constrained chokepoint in AI safety infrastructure, with far more high-value work than current capacity allows.
“They have 20 really useful projects they'd love to do but they only have capacity to do two or three of them.”Ben Todd · 26 May 2026
Most forecasters have estimated a longer doubling time for AI task length than the sub-4-month trend the METR chart actually shows, consensus is systematically too conservative.
“It seems like everybody has kind of estimated a higher number than the trend so far suggests, which is like kind of a little under four months doubling time for task length, which means we'll be greater than 8x, maybe somewhere in the like 10 to 12x over the course of just a year.”Host (Nathan Labenz) · 26 Apr 2026