Small businesses may soon rely on AI agents to autonomously complete specialized freelance work, significantly reducing project costs and turnaround times.
What’s the Remote Labor Index and what changed?
The Remote Labor Index (RLI) tests how well AI systems perform economically valuable projects online in a fully end-to-end way. It measures autonomous capability on complex freelance tasks.
According to research from the Center for AI Safety and Scale Labs, the frontier success rate on the RLI jumped from 2.5% at launch in October 2025 to 16.1% in July 2026. The index evaluates models on tasks including 3D CAD, architecture, graphic design, video and animation, audio, data analysis, and web applications. The assessed projects require models to process complex instructions, like generating photorealistic floor plans from scanned site photos or producing 60-second animated advertisement videos.
The frontier of economically capable AI agents has more than quadrupled in under 8 months.
What’s the evidence behind the Remote Labor Index?
Evaluators tested 3 recent frontier models on the RLI to measure autonomous task completion. Fable 5 achieved the highest score, while GPT-5.5 and Opus 4.8 lagged behind.
Fable 5 scored 16.1% on the index, establishing the new frontier capability. Opus 4.8 achieved an 8.3% success rate on the same suite of freelance tasks. GPT-5.5 scored 6.3% on the benchmark. The index forces agents to complete self-contained workflows, such as recreating a client’s emerald-cut engagement ring as a marquise cut and delivering updated 3D models with photorealistic rose and yellow gold renders.
Fable 5 currently operates at more than double the autonomy rate of Opus 4.8 on specialized freelance work.
How does the Remote Labor Index compare to the alternatives, and what background do founders need?
The RLI evaluates multi-step, specialized project completion rather than isolated coding or text generation.
While the RLI measures 16.1% success on specialized tasks, the OSWORLD 2.0 benchmark shows agents struggle with long-horizon computer use, with Claude Opus 4.8 reaching only 20.6% binary accuracy. OSWORLD 2.0 tasks take a skilled human user approximately 1.6 hours to complete, which is 48x longer than the 2-minute median in OSWORLD 1.0. As agents become more fluent with professional software like Slack, AWS, and GitLab, their ability to handle complex economic workflows will compound, creating a proxy for how quickly they can replace human labor.
The RLI measures specialized output generation, while OSWORLD 2.0 measures multi-hour software orchestration.
How does the Remote Labor Index affect day-to-day operations for small businesses?
The RLI provides a concrete signal that AI capability expansion is outpacing human comparative advantage in specialized freelance work.
The jump from 2.5% to 16.1% in under 8 months demonstrates that person-light, AI-heavy organizations will soon out-compete un-augmented human labor. Small business owners should monitor this benchmark and prepare to integrate autonomous agents for 3D design, architecture, and video editing as the reliability curve approaches 80%.
Small business owners must prepare their operations to integrate autonomous freelance agents within the next 24 months.
A stack of copper coil assemblies sits on the warehouse floor, waiting for a junior technician to wire them into the new rooftop units. The lead installer assigned the job to an autonomous diagnostic tool instead, expecting a perfect wiring schematic delivered directly to the tablet. The tool reported a 100% completion success rate, but when the technician mounted the units, the capacitor routing was entirely wrong, requiring 3 hours of rework on a job that should have taken 20 minutes. This silent failure is exactly what a 16.1% success rate looks like in the field. The AI can execute the complex, multi-step task of reading a mechanical schematic and generating a parts list, but it leaves a hidden defect that causes an operational failure later. You outsource a complex video animation job to an agent, it renders the file, and you hand it to the client without checking the frame transitions. The RLI proves the capability exists, but an 83.9% failure rate means you still need a human to verify the final deliverable before it touches your customer.
What’s the final verdict on the Remote Labor Index?
The RLI provides a concrete signal that AI capability expansion is outpacing human comparative advantage in specialized freelance work. A 16.1% success rate proves agents are functional but not yet autonomous.
The jump from 2.5% to 16.1% in under 8 months demonstrates that person-light, AI-heavy organizations will soon out-compete un-augmented human labor. Small business owners should monitor this benchmark and prepare to integrate autonomous agents for 3D design, architecture, and video editing as the reliability curve approaches 80%.
Small business owners must prepare their operations to integrate autonomous freelance agents within the next 24 months.
Source: Import AI
