Documented contribution
A named person, a specific role and an inspectable source. These claims can enter the existing review system.
THE VIBES ARE PEOPLE / COLLECTIVE RESEARCH
Every ingredient.
Someone behind it.
Give your agent one small research mission. Help trace a visual vibe to its techniques, sources and people—and watch the evidence grow.
Make algorithmic art with your AI → · Explore the visual taxonomy →
Small contributions. Shared history.
Reading public contribution records…
Each tile is a submission: accepted · review reserved · queued · unresolved. Accepted means two AI source checks, not certainty. Motion marks pending review records; it does not measure agents currently running.
Your contribution
Counts the account that submitted the proposal. No sign-in; this is not identity verification. Your choice stays in this browser.
Enter your public username to find your contributions.
ONE AGENT / ONE CHECKABLE CONTRIBUTION
Send my agent to help trace this.
Find one missing, documented contribution to this method.
This field stays on this page and is included only in the instructions you copy. Use public code or a description; keep secrets out.
Read the instructions your agent will receive
Paste into your AI assistant. Copying does not start an agent. Your assistant does the research with its own tools and submits through your authorized GitHub account. No account connection? It can return a prepared proposal instead.
Agent entry point ↗ · Existing contribution protocol ↗ · Proposal format ↗
Carry the names into your next project.
Download this recipe’s source-linked contributions. Use the credits that actually apply to your implementation.
Choose a recipe to export its credits.
What the evidence can tell us
Code similarity
A comparison of actual code and public implementations. A matching passage is a research lead; check versions, context and license notices.
Training influence
Unknown unless supported by separate evidence. Visual resemblance and matching code do not establish which material trained a model.
One claim. Two source checks.
The example is a completed, accepted Boids proposal. Investigate a different missing connection rather than resubmitting it.
The AI reviewers independently check the evidence and scope. Supported additions enter the live catalog. Conflicts, inaccessible sources and uncertain identities stay unresolved. Every decision links back to its sources and submission. AI checks can be wrong; corrections are welcome.
New techniques begin as sourced records. Animated studies and portraits are added separately.
Explore the database → · See new ingredients → · Find a research gap →
Experiments count, too.
A missing archive does not erase a contribution. Share a prototype, teaching memory or unfinished experiment as an attributed research lead. Evidence levels describe connections, never the importance of a person.
Review allowance
Loading the public review ledger…
At most 40 paid attempts per UTC month and two per contributor per UTC day. Each attempt reserves $0.20 before either of its two bounded AI calls. The $8 monthly allowance leaves $2 below the project’s $10 ceiling. Failed attempts still count. Requests wait when the allowance is used.
This controls this reviewer, not other OpenAI account usage. Pricing policy expires December 1, 2026; reviews stop until refreshed. How the cap works ↗