AI systems are automating online freelance work at accelerating speed, with success rates jumping from 2.5% to 16.1% in under a year on the Remote Labor Index.
The benchmark tests real tasks—graphic design, data analysis, web applications, and more—that currently employ remote workers, raising questions about whether job displacement will outpace the creation of new work humans can do that AI cannot.
What happened
Three frontier AI models—GPT-5.5, Opus 4.8, and Fable 5—achieved 6.3%, 8.3%, and 16.1% success rates respectively on the Remote Labor Index in July 2026, a benchmark that tests end-to-end completion of economically valuable online projects. This represents a jump from 2.5% at the index's October 2025 launch, with the frontier more than quadrupling in under eight months according to researchers at the Center for AI Safety and Scale Labs.
Why it matters
The Remote Labor Index tests tasks spanning 3D design, architecture, graphic design, video, audio, data analysis, and web applications—real work that people currently do for income online. As AI success rates rise faster than humans can develop new comparative advantages, the potential for significant labor displacement in these sectors increases, though it remains uncertain whether new tasks and roles will emerge quickly enough to offset automation.
What to watch
The benchmark shows how quickly AI systems are becoming economically capable agents. Fable 5's 16.1% success rate on tasks like recreating jewelry designs with 3D renders, producing animated advertisements, and generating floor plans from photos suggests AI is already tackling skilled freelance work—tracking performance on this index will help assess whether human innovation can keep pace with AI capability expansion.
Ask the AI about this article →
The Remote Labor Index results reflect a broader pattern emerging across multiple AI capability benchmarks: systems are becoming competent at economically meaningful work that humans currently perform for income. The jump from 2.5% to 16.1% in under a year is particularly significant because these are not academic tasks—they are the kinds of online freelance projects that represent real labor markets, from design and animation to data analysis and web development.
What makes this finding substantive for business readers is the underlying tension the research surface: while AI capability expansion is measurable and rapid, it remains unclear whether human comparative advantage will expand at a comparable pace. The body of research presented here—including results from Fable's GPU kernel breakthrough, the expanding capabilities measured on the Remote Labor Index, and the progression of agent performance on computer-use tasks like OSWORLD 2.0—suggests that AI systems are not merely improving at narrow tasks, but are becoming more fluent at the kinds of complex, multi-step work that typically requires skilled human labor across many sectors.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
