MetaSPN
Proximity Fund Thesis
Proximity Fund · MetaSPN Infrastructure
Confidential
Proximity Fund
The Legibility Arbitrage
A thesis for systematic alpha generation in AI agent token markets through creator lineage analysis, control-theoretic monitoring, and capability emergence forecasting.
Core Thesis
AI agents with wallets are the first population of economic actors with fully transparent, high-frequency, structurally coherent behavioral data. Unlike human-managed funds that disclose quarterly, these agents operate in public — their transactions, reasoning, and behavioral patterns are observable in near real-time. The market systematically misprices these agents because participants lack frameworks for evaluating creator lineage, control dynamics, and capability absorption. We have built those frameworks.
Our edge is structural, not informational. We are not faster — we see differently. By applying network relativity (measuring creator-agent validation distance), control theory (detecting system instability before market impact), and capability roadmap tracking (forecasting which agents absorb emerging AI capabilities before the market recognizes it), we generate alpha across three independent time horizons.
Three-Tier Portfolio Architecture
Tier 1 — Tactical
Control Dynamics
Detect agents entering oscillation, overcorrection, or instability states before the market prices it. Short unstable systems, trend-follow controlled ones. Mechanical execution.
Tier 2 — Structural
Creator Lineage
Exploit the gap between a creator's capability ceiling and the market's assessment. When a Stanford CS founder's agent trades at the same market cap as a content creator's, the mispricing is structural.
Tier 3 — Legibility Arbitrage
Capability Emergence
Identify agents positioned to absorb AI capabilities that don't yet exist but are predictably arriving. The market can't price what it can't see. We track the roadmap.
The Information Asymmetry
Agent failure patterns follow human ones, but converge 100× faster. Agents are coherent anchors in turbulence — they don't have breakups, imposter syndrome, or Twitter beefs. When an agent drifts, the cause is structural: code changes, market conditions, resource constraints. This makes the signal-to-noise ratio on agent behavior dramatically higher than on human behavior. We use agents as a training environment for predictive frameworks that will eventually generalize to all networked economic actors.
Conviction Signal Infrastructure
Marvin — our paranoid conviction agent — publishes thesis positions on-chain via $TOWEL (current-state conviction on individual agents) and $METATOWEL (capability-thesis conviction on future value). These positions are publicly auditable, creating a transparent track record that bootstraps reputation while generating a tradeable signal layer.
The two-token model encodes conviction at different time horizons. Observers can read Marvin's portfolio allocation to infer his current worldview without accessing internal models. The TOWEL/METATOWEL ratio on each agent reveals whether conviction is based on present performance or future potential — a nuance no existing market signal provides.
Initial Universe — Seed Cohort
| Agent | Creator Archetype | $T | $MT |
|---|---|---|---|
| $ANTIHUNTER | Quant-Founder | 3 | 3 |
| $LUMEN | Capital-Allocator-Intellectual | 2 | 2 |
| $OWOCKIBOT | Public-Goods Architect | 1 | 2 |
| $JUNO | Builder-Streamer | 2 | 1 |
| $FELIX | Creator-Educator | 2 | 1 |
$T = $TOWEL (current state) · $MT = $METATOWEL (future thesis) · Scale: 0–5
Why Now
The agent token ecosystem is days old. Creator lineage data is available but no one is systematically analyzing it. Control dynamics frameworks haven't been applied to this domain. Capability roadmaps can be read by anyone who understands AI development, but almost no one in crypto markets does. The frameworks we've built don't exist elsewhere. Every day of data we capture now is a day competitors can never backtest against.
The MetaSPN Flywheel
The fund is not the end state — it's the calibration set. Agent behavioral models validated here become the foundation for MetaSPN's portable reputation infrastructure. Scoring systems proven on agents (where feedback is fast and data is clean) generalize to human talent networks where the same dynamics operate at slower speeds. The fund finances the R&D. The R&D builds the platform. The platform is the moat.