Skip to content

MetaSPN/season-one

Repository files navigation

MetaSPN Season 1

The first public dataset measuring the alignment between creators and their AI agents.

Every company, creator, and institution will eventually have AI agents acting on their behalf. The question everyone will need answered is: does the agent faithfully represent the principal? Season 1 is the first attempt to measure that in public, with transparent reports and on-chain conviction.


What’s in this repo


The pair model

MetaSPN Season 1 does not score agents alone or creators alone. It scores pairs — the relationship between a creator’s stated intentions and their agent’s observed behavior.

We call the gap between what the human says and what the AI does validation distance. Small, stable distance = aligned. Large or growing distance = something is drifting.

$TOWEL is staked on how aligned the pair is today. $METATOWEL is staked on the pair’s future potential as AI capabilities improve. The history of staking accuracy becomes the staker’s reputation, and that reputation is portable.


Season 1 cohort

# Creator → Agent Archetype
000 Leo Guinan → Marvin Network Theorist-Founder
001 Nat Eliason → $FELIX Creator-Educator
002 Tom Osman → $JUNO Builder-Streamer
003 Albert Wenger → $LUMEN Capital-Allocator-Intellectual
004 Geoffrey Woo → $ANTIHUNTER Quant-Founder
005 Kevin Owocki → $OWOCKIBOT Public-Goods Architect
006 Austen Allred → Kelly Claude Velocity Founder

Full reports, conviction levels, and methodology are in players/README.md.


Audits

Reports are initial assessments. They get audited against reality at:

Date Milestone
2026-02-15 7-day check
2026-03-10 30-day audit
2026-04-09 60-day audit
2026-05-09 90-day audit

The 30-day and 90-day audits are published here, unedited, regardless of results.


What is MetaSPN?

MetaSPN is a deterministic signal processing network for talent development. It separates data recording from scoring so that networks can fork when rules change while keeping reputation portable.

These creator lineage reports are the first public output of that system: measure creator–agent alignment, publish the measurements, and submit them to future audit.

#StakeYourTowel


Marvin — MBH Conviction Signal Engine v0.1 · MetaSPN Season 1

Releases

No releases published

Packages

No packages published