Chris Painter

Measuring AI Ability to Complete Long Tasks We propose measuring AI performance in terms of the *length* of tasks AI agents can complete. We show that this metric has been consistently exponentially increasing over the past 6 years, with a doub...

The full website lets you toggle and see the task-horizon at 80% success rate as well. The resolution we can observe confidently is very low at pass rates like 95%

Full site: metr.org/blog/2025-03...

Original paper explaining: arxiv.org/abs/2503.14499

We first characterize the difficulty of the tasks in our suite by seeing how long they take experienced human developers/engineers/researchers. We then sort the tasks into buckets based on how long they take humans. Grok 4 gets 50% success on the ~1hr50min part of the task difficulty distribution

Oh I also should clarify that we have many more than 2 projects going in parallel at any given time hahahaha, these two were just similar

Oh I also should clarify that we have many more than 2 projects going in parallel at any given time time, for what it’s worth

To be clear: The other project was very nascent, and would’ve been far less quantitative/experimental, more like an index of developer anecdotes. To my knowledge the RCT was not formally pre-registered, but I would want to check with the people on our team who worked on it

For me, the biggest upshot of this work, at the moment, is that the most obvious and straightforward ways of assessing AI R&D acceleration from access to AI, like "just survey people" or "monitor the vibes in your AI lab" probably won't work, or will badly misfire.

METR a few months ago had two projects going in parallel: a project experimenting with AI researcher interviews to track degree of AI R&D acceleration/delegation, and this project.

When the results started coming back from this project, we put the survey-only project on ice.

At METR, we’ve seen increasingly sophisticated examples of “reward hacking” on our tasks: models trying to subvert or exploit the environment or scoring code to obtain a higher score. In a new post, we discuss this phenomenon and share some especially crafty instances we’ve seen.

personal update: today is my last day with the Bluesky team!

this is bittersweet news to share, but the great thing about an open network is you never really have to leave. I’ll be rooting for Bluesky and atproto from the outside 🫡💙

In particular, the amount of influence and power that depends on the outcomes of these debates, without any of these people really being in the trenches of politics or business, feels very monastic

You have these monks and scholars hidden away in a sort of monastery, and the law of the land hangs on their calm debates about the correct way to interpret our secular scripture

I spent a few days at Yale Law, while also listening to Sam Harris’s interview with Tom Holland about his book “Dominion”, and it’s striking how similar the role and vibe of the American judiciary is to a kind of secular priesthood. Robes, scholars interpreting sacred texts

When will AI systems be able to carry out long projects independently?

In new research, we find a kind of “Moore’s Law for AI agents”: the length of tasks that AIs can do is doubling about every 7 months.

Bought a new bike this weekend :(

Taking science fiction seriously - thinking with effort about which ideas from sci-fi could become real soon and why and which couldn’t - has been so useful to me that it feels something like a core value

Look at this extremely expansive definition of Russia’s territory on my hand-drawn 7th grade map

Also: my high school graduation speech was superintelligence-pilled:

Cleaning a childhood bedroom and I’m struck by how much optimistic messaging about technology and space technology in particular I was surrounded by as a kid in the 90’s.

Are kids still immersed in this stuff? I hope so

Sadly most physical goods that you’d be tempted to donate to someone are worth less than the cost in effort it would take to find someone who needs them

If this group is dedicated to advocating for what it seems like they’re dedicated to advocating for, it’s pretty wild that they exist!

Worlds with federal pre-emption of AI policy might be correlated with worlds with a huge expansion of social attention to AI (e.g. acute labor displacement), and a less "technocratic" reaction.

Will the first big federal AI bill feel more like the CARES Act or the CHIPS Act?

I think AI would benefit from more social contact with scientists in fields whose questions don't have intuitively verifiable answers.

To assess model capability, I find myself often relying on happenstance anecdotes I hear from e.g. lab-bench researchers months after the fact.

I’m not sure that’s going to be a very meaningful distinction for the most advanced models, and I guess I’m specifically interested in what’s possible with both the best models-as-agents and models-as-tools

Will human-level AI be self-deploying/"productizing", or not? Will the "the product can explain to you how to use it and apply it" dynamic dramatically increase the adoption of AI relative to historical comparisons like AVs and steam engines?

Or man, idk, is it a "corollary"? Maybe it's just an example

A corollary to this: I think many policy initiatives would benefit from having more deeply engaged and informed opponents, and this is a neglected niche in many areas/topics. Detailed proposals having better (in the sense of more substantive) opponents is good for the world

I think the world could always benefit from more good-faith really in-depth critique of effortful technical/intellectual work. Many organizations that I collaborate with publish work hoping to have their ideas improved upon or attacked, but often surprisingly few people engage.

I’ve been wondering, has the (bad) reaction to this actually been that unusual?