Skill Trees & Task Scoring

One of Orn’s defining capabilities is its ability to not only process and label submissions, but also to evaluate performance. This is achieved through a dual system of task scoring and skill trees, which together measure how well a user performs specific actions and how those actions accumulate into long-term ability profiles.

Task Scoring begins at the individual submission level. Every task is assessed against a reference library of expert demonstrations that serve as gold standards for performance. For example, in a sushi-making task, Orn compares the user’s video against a curated set of professional chefs preparing sushi. The system evaluates whether the steps are followed in the correct order, whether timing and execution align with expert patterns, and whether the final presentation meets quality expectations. This approach ensures that scoring is not subjective but anchored in objective benchmarks of excellence. Similar reference models exist across domains — from folding clothes to playing piano chords to guiding a wheelchair — ensuring consistency across all types of activities.

The outcome of this process is a numerical performance score that reflects accuracy, completeness, and execution quality. A high score indicates strong alignment with expert baselines, while a low score signals errors, missed steps, or poor execution. Rejected submissions — such as those that fail task requirements, recycle old content, or attempt to bypass liveness checks — not only receive no credit but also negatively impact the user’s scoring profile.

These task-level scores then feed into Skill Trees, which represent a contributor’s evolving capabilities across domains. Unlike flat reputation systems, Skill Trees provide a multi-dimensional view of ability. They are organized by activity type — such as kitchen, sports, outdoor tasks, or dexterity-based challenges — while also incorporating cross-cutting attributes like precision, endurance, and creativity. This structure allows Orn to capture both specialization and versatility: a contributor might be highly skilled in kitchen tasks with strong precision, while another excels in outdoor activities requiring endurance.

Skill Trees play a crucial role in task allocation and progression. Certain high-value, specialized, or safety-sensitive challenges are only made available to users whose Skill Trees demonstrate sufficient capability in the relevant domain. For instance, a user who has built a strong record in basic culinary tasks may unlock access to more advanced cooking challenges, while a consistent performer in dexterity-based puzzles may be invited into higher-stakes precision tasks. This gating mechanism ensures that tasks are completed by contributors who are both qualified and motivated, maintaining quality across the ecosystem.

At the same time, Skill Trees incentivize users to improve and diversify their abilities. By striving for higher scores and expanding their profiles across multiple domains, contributors unlock access to more tasks and higher VP rewards. This creates a merit-based system where progress is tied directly to demonstrated ability, not just participation volume. The result is a self-reinforcing loop: better performance leads to higher scores, higher scores unlock more opportunities, and more opportunities drive greater engagement and contribution quality.

Performance data from users continuously feeds back into Orn’s models, refining scoring criteria, strengthening manipulation detection, and expanding the expert reference libraries that serve as gold standards. Through this system of Skill Trees and Task Scoring, Orn transforms raw submissions into a living record of user ability. Every task becomes both a piece of labeled data and a performance datapoint, building a robust reputation system that aligns individual incentives with ecosystem-wide quality.

All thresholds, parameters, and detection methods described are subject to continuous refinement as technology advances and as the requirements of the ecosystem evolve.