To measure an AI model's economic value, survey domain experts on how they allocate their time across various tasks. This time-allocation data serves as a proxy for the economic weight of each task, against which the model's performance can be scored.

Block's CTO quantifies the impact of their internal AI agent, Goose. AI-forward engineering teams save 8-10 hours weekly, a figure he considers the absolute baseline. He notes, "this is the worst it will ever be," suggesting exponential gains are coming.

By training AI on your personal data, arguments, and communication style, you can leverage it as a creative partner. This allows skilled professionals to reduce the time for complex tasks, like creating a new class, from over 16 hours to just four.

A case study building a customer success score demonstrates how AI can act as a senior-level strategist. A project that would typically take 50-100 hours of manual work was completed in just 3-5 hours using a multi-model AI approach.

AI tools provide quantifiable productivity gains in technical fields. Developers using GitHub Copilot, for instance, finish tasks approximately 55% faster. Furthermore, 88% of these developers report feeling more productive, demonstrating that AI augmentation leads to significant and measurable improvements in workflow efficiency and employee satisfaction.

OpenAI's new GDPVal framework evaluates AI on real-world knowledge work. It found frontier models produce work rated equal to or better than human experts nearly 50% of the time, while being 100 times faster and cheaper. This provides a direct measure of impending economic transformation.

Benchmarks like GDPVal show models like GPT-4 consistently outperform human experts on professional tasks, meeting the practical definition of AGI for knowledge work. The public discourse, however, has prematurely shifted the goalposts to sci-fi concepts of Artificial Superintelligence (ASI), obscuring the revolution already underway.

A simple framework to estimate AI's current economic impact multiplies three key metrics: the percentage of workers using AI (~40%), their weekly usage intensity (~2 hours), and the average task efficiency gain (15-30%). This calculation reveals a modest but tangible current productivity increase.

Traditional AI benchmarks are seen as increasingly incremental and less interesting. The new frontier for evaluating a model's true capability lies in applied, complex tasks that mimic real-world interaction, such as building in Minecraft (MC Bench) or managing a simulated business (VendingBench), which are more revealing of raw intelligence.