How long a task can AI do?
METR tracks the length of tasks frontier models can complete. Start here if you're wondering which skills may be repriced.
Read METR's time horizon research ↗You'll spend the next decade building something. The question is what deserves the time.
For people already a few years into work and wondering what comes next. This is not a pep talk. It's a set of tests you can actually run.Maybe you were in the room. Maybe someone sent you this afterward. Either way: start with the question that won't leave you alone, then use the deck as a map.
All 61 original slides live here. Open Present mode ↗ for a projector; use ← / → or swipe, and press F for fullscreen.
Don't leave with an abstract feeling. Take the two-page Matter Test. Sort the skills AI changes, map what gives you energy, and pick one small bet with a date on it.
Three doors into the evidence. You don't need to agree with the talk. Read the work and decide where the argument holds.
METR tracks the length of tasks frontier models can complete. Start here if you're wondering which skills may be repriced.
Read METR's time horizon research ↗Stanford Digital Economy Lab studies AI exposure and early-career employment. Look at the patterns, then ask what they leave out.
Read the Stanford findings ↗GMAC's 2026 recruiter survey is a second lens on MBA hiring, skills and salaries. Compare the survey with the kind of work you want.
Explore the GMAC survey ↗Pick a person, not a market. Ask about their Tuesday. Then make a tiny offer they can say yes or no to by a real date.
Put it on the field sheet ↓This page holds the presentation and worksheet prepared for Sai Krishna's Mesa School session. The live conversation, Q&A and audience answers are not recorded here. After the room clears, you can still return to a slide, challenge a source, or run the Matter Test again.