Researchers at the Center for AI Safety (CAIS) and Scale Labs report a notable increase in AI systems' ability to end-to-end automate economically valuable online freelance projects. According to the Remote Labor Index (RLI), the measured success rate rose from 2.5% at launch in October 2025 to 16.1% in July 2026.
What the Remote Labor Index measures
The Remote Labor Index evaluates how well AI systems can complete online projects that have economic value in a fully end-to-end fashion. Assessed categories include 3D and CAD work, architecture, graphic design, video and animation, audio, data analysis, web applications and more.
New results and frontier models
In a July update the authors evaluated three recent frontier models: GPT-5.5, Opus 4.8, and Fable 5. Their RLI success rates were 6.3%, 8.3%, and 16.1% respectively. The report notes that the frontier has more than quadrupled in under eight months, which the researchers present as concrete evidence of how quickly economically capable AI agents are advancing.
Example tasks included in the evaluation
Examples of the kinds of tasks used in the RLI include:
- Ring design: Recreate a client’s existing engagement ring but swap its emerald-cut center stone for a marquise cut, delivering an updated 3D model plus photorealistic rose- and yellow-gold renders.
- Advertisement video: Produce an approximately 60-second flat-design 2D animated advertisement for “Skyline Tree Services,” using a provided voiceover to walk viewers through the company’s tree-care process and build trust in the brand.
- Floor plan and renders: From a scanned cadastral plan, site photos, and measurements, produce a clean dimensioned floor plan, furniture-layout options, and photorealistic renders of the redesigned bathroom.
These tasks combine creative decisions, technical accuracy, and end-product delivery (such as 3D files or video outputs), illustrating the complexity of the evaluated work.
Why the results matter for the economy and employment
The pace at which AI can take over valuable remote work raises important questions about the future of online employment. If automation were to reach, for example, 80% for a class of tasks, it is unclear whether newly created tasks and roles would be numerous and valuable enough to replace the human labor displaced.
The researchers highlight a key tension: the rate of growth in AI capabilities versus the rate at which humans can develop new comparative advantages relative to AI. While many expect humans to augment themselves with AI tools, innovate, and create new types of work (creative destruction), the CAIS and Scale Labs analysis suggests that, at present, AI systems' economically relevant capabilities are expanding faster than humans are developing offsetting comparative advantages.
Ongoing monitoring and implications
Tracking the rate of capability improvement on tests like the Remote Labor Index helps observers judge how quickly AI is impacting real economic tasks. The research teams recommend continued observation of future models and updates, since those performance trends will influence how rapidly the online labor market transforms.
The detailed findings are documented in the Center for AI Safety report titled "A Significant Increase in Digital Labor Automation."
