What We Learned from Letting AI Posttrain AI

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 doubling time of around 7 months. Extrapolating this trend predicts that, in under a decade, we will see AI agents that can independently complete a large fraction of software tasks that currently take humans days or weeks.

Hyperfast AI: Rethinking Design for 1000 tokens/s

I recently spoke at AI Tinkerers Raleigh about hyperfast inference systems and how they’re fundamentally changing AI application design. If you haven’t heard of Cerebras (or however they pronounce it), you’re in for a treat—this is one of the most exciting areas of research in AI right now.

What AI coding costs you | Tom Wojcik

What's the effect of the prolonged AI usage among coders and is it tracked correctly, if it all?

Using AI for Just 10 Minutes Might Make You Lazy and Dumb, Study Shows

New research suggests that reliance on AI assistants can have a negative impact on people’s ability to think and problem solve.

Where’s my ten minute AGI?

Why don’t AIs automate more real-world tasks if they can handle 1-hour ones? Here are at least three fundamental reasons.

scaleapi/SWE-bench_Pro-os

SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?

I don't know if I like working at higher levels of abstraction

AI tools push us to higher abstraction. I'm not sure I like what that costs us.

AI Isn’t Lightening Workloads. It’s Making Them More Intense.

The technology is increasing the speed, density and complexity of work rather than reducing it, a new analysis of 164,000 people’s work activity shows.

Meet Foundry: An AI Startup that Builds, Evaluates, and Improves AI Agents

The Ma of a New Machine – Scott Jenson

The current Silicon Valley flex is trading notes on your favorite new AI tools over lunch. Each week brings another one to explore. Some of these tools are very impressive; I’ve been able to reply to someone with an alternative UI design in less than a minute, and they were dumbfounded: “How did you do that so fast?”

Scott Jenson – Exploring the world beyond mobile

Fast AI requires slow thinking The current Silicon Valley flex is trading notes on your favorite new AI tools over lunch. Each week brings another one to explore. Some of these tools are very impressive; I’ve been able to reply to someone with an alternative UI design in less than a minute, and they were […]

Arvind Narayanan on Twitter / X

To understand and empathize with how workers in many or most fields outside software experience advances in AI capabilities, I propose a little thought experiment. https://t.co/QZG24aRiBe— Arvind Narayanan (@random_walker) July 28, 2026

Alex Cheema on Twitter / X

This is why we need open benchmarks for local AI.Otherwise it turns into tribalism and name calling.We will be publishing the largest database of open benchmarks for local AI, tested on 1,000+ real hardware setups. Every device, every interconnect, different… https://t.co/ZsU3PCdSsZ— Alex Cheema (@alexocheema) March 9, 2026

Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity

Despite widespread adoption, the impact of AI tools on software development in the wild remains understudied. We conduct a randomized controlled trial (RCT) to understand how AI tools at the February-June 2025 frontier affect the productivity of experienced open-source developers. 16 developers with moderate AI experience complete 246 tasks in mature projects on which they have an average of 5 years of prior experience. Each task is randomly assigned to allow or disallow usage of early 2025 AI tools. When AI tools are allowed, developers primarily use Cursor Pro, a popular code editor, and Claude 3.5/3.7 Sonnet. Before starting tasks, developers forecast that allowing AI will reduce completion time by 24%. After completing the study, developers estimate that allowing AI reduced completion time by 20%. Surprisingly, we find that allowing AI actually increases completion time by 19%--AI tooling slowed developers down. This slowdown also contradicts predictions from experts in economics (39% shorter) and ML (38% shorter). To understand this result, we collect and evaluate evidence for 20 properties of our setting that a priori could contribute to the observed slowdown effect--for example, the size and quality standards of projects, or prior developer experience with AI tooling. Although the influence of experimental artifacts cannot be entirely ruled out, the robustness of the slowdown effect across our analyses suggests it is unlikely to primarily be a function of our experimental design.

CEO-Bench

CEO-Bench evaluates whether AI agents can steer a simulated AI startup for 500 days, testing long-term planning, adaptation, and coordination under uncertainty.