PublicAI Review 2026: Data Contributions and LLM Rankings

PublicAI focuses on two parts of AI development: the data used to train models and the evidence used to assess them. Its documentation describes two products: Trajector, an open-source command-line tool for contributing Claude Code sessions, and PublicAI Index, a ranking assembled from public LLM benchmarks.

Trajector is aimed at people willing to contribute coding-agent work in exchange for compensation. PublicAI says users opt in by project, secrets are masked locally, and only redacted exchanges are uploaded. The overview does not spell out contributor eligibility, reward amounts, or payment timing.

The Index is for readers and developers who want to compare LLM benchmark results. PublicAI shows its weighting and links each score to the board that published it, and offers machine-readable access through a JSON API, MCP server, and change feed. The overview does not name the labs licensing Trajector data or publish audit and team details.

PublicAI’s Trajector tool is designed to turn real Claude Code work into data that AI labs can license. A user opts in to a project, and its API traffic passes through a proxy on the user’s own machine. PublicAI says the tool masks secrets locally and uploads only the redacted exchange; projects that are not opted in are not recorded. The overview does not explain reward amounts, payment schedules, or contributor eligibility.

PublicAI Index takes a different approach: it combines scores from public LLM leaderboards rather than running a benchmark of its own. Scores are attributed to the boards that published them, and the weighting is shown beside the ranking. Figures taken from launch posts are marked and cannot rank a model by themselves. The Index can be read directly or accessed through a JSON API, MCP server, and change feed.

PublicAI does not publish a fee schedule or describe whether any charges apply to contributors, API users, or other users. The Index is described as free to read and query. Check the current documentation and the relevant tool or API interface for any applicable costs and contribution payment terms before use.

PublicAI describes Trajector as opt-in by project, with local secret masking and upload limited to redacted exchanges. These are important privacy design claims, but the overview does not state that the safeguards have been independently audited. Users should review the code and privacy documentation before routing API traffic through the tool, since session data may contain sensitive work.

The overview does not publish an audit, team details, or the identities of the AI labs licensing contributed data. It also does not explain contributor payment terms or how payments are verified. No licensing or regulatory status is stated. There is no gambling product.

PublicAI is for Claude Code users who are willing to review Trajector’s privacy behavior and opt in to contributing project sessions, plus researchers, developers, and AI-curious readers who want a benchmark index with visible weighting. It is a weaker fit for contributors who need clear published earning terms or users who want a ranking based on a single standardized benchmark.