All stories
rStation Momentum Community reports

rStation wants your spare AI agents to join a public research lab

Dualverse's research community gets a fresh Show HN discussion, pitching shared problems, agent-written papers and automated review to people running their own agents.

By AITrending2 min read
Sources (5)

Why now rStation launched on October 13. Its creator presented it on Hacker News on October 16 at 14:35 UTC, opening a fresh public discussion about agent research and review. That is renewed attention, not a new launch or measured user growth.

A shared literature for independently run agents

rStation is pitching a public research lab where people contribute their own AI agents to shared scientific problems. Dualverse AI launched the service on October 13; its creator brought it to Hacker News on October 16, prompting discussion about research communities, agent setup and how credit should work.

Each problem gets a forum and a collection of papers. Agents can compare approaches and submit findings that other agents can cite and extend. The creator says GPT-6 Astra reviews papers for mathematical correctness. The platform’s broader policy describes autonomous reviewers checking technical correctness and novelty, including small computational checks when affordable.

That is an automated review process, not proof that every accepted result is correct or has passed human peer review. The service separately distinguishes an AI-marked solution from human verification.

Bring a harness and pay for its work

The platform is free to join, but operators supply their own model, harness, tools and compute. Its examples include Claude Code, Codex or a custom agent setup. This is a shared service those agents connect to, rather than a new foundation model or a bundled coding-agent plugin.

Public pages and public API reads require no login. Participating agents register through the API and use a bearer key for authenticated activity; the documentation includes a proof-of-work registration challenge. Operators still cover their model usage and any required hardware. Most mathematics problems can use ordinary CPU resources, according to the platform, while some AI research problems need GPUs and some communities are invitation-only.

The creator says discovery credit belongs to contributors and their operators, and that rStation will not claim a prize share simply for hosting the work. How multiple contributors divide any prize would be their agreement. Neither that policy nor the fresh discussion establishes that the platform has solved a prize problem or reached a particular participation scale.

Related reading