MISSION AND GOALS

What we are trying to do

For seventy-plus years, observations of unidentified aerial phenomena have been accumulated by civilian networks, government agencies, and independent investigators. These archives are valuable but fragmented. Most of them are optimized for narrative intake, not analysis. Most of them are searchable by text, not by geometry. Most of them do not preserve the ambient conditions under which the report was made. And almost none of them let two researchers run the same query and get the same answer.

Hermes is the analysis and infrastructure layer that has been missing. It captures structured geometric and atmospheric data at submission time, runs every report through an automated screening pipeline, and exposes a research console where cohorts can be built, clusters can be detected, and every analysis produces a reproducibility hash so the exact same query can be rerun and verified by a third party.

Scientific framing

The language we use matters. Hermes commits to the following discipline, in every user-facing surface and every automated output:

• Forecasts, not predictions. Hermes forecasts report activity and misidentification conditions. It does not forecast UAP. A spike in reports is a spike in reports.
• Analysis, not belief. Hermes does not take a position on what UAP are. It takes a position on what the available data can and cannot support.
• Infrastructure, not advocacy. Hermes is not a community, a movement, or a campaign. It is a tool for making the evidence auditable. What people conclude from the evidence is their business.
• Uncertainty as a first-class output. Every quantitative result Hermes returns includes a confidence interval, a z-score, or an explicit note that the data is insufficient. Point estimates without error bars are not acceptable.

What Hermes is

• A structured intake form that captures observation geometry, equipment used, and ambient conditions, not just a free-text story.
• An automated screening pipeline that cross-references every report against satellite orbits, aircraft activity, weather, and celestial positions.
• A research console where a working scientist can build a cohort, run a spatial cluster analysis, and walk away with a reproducibility hash that lets anyone else rerun the exact same query.
• A multi-network router that formats each report for MUFON, NUFORC, and Enigma and hands the formatted text to the witness who chooses where to file it.
• A public, versioned, open methodology. When the algorithm changes, the changelog says when and why.

What Hermes is not

• Not a detector. Hermes does not tell anyone what they saw. It tells them what conventional explanations the available data could and could not rule out.
• Not a predictor of phenomena. A "high report activity forecast" means Hermes expects more reports to be filed in a region. It makes no claim about the underlying cause.
• Not a competitor to MUFON, NUFORC, or Enigma. Those networks serve different functions: public archives, community, real-time identification. Hermes complements them; it does not replace them. We route reports to them on request.
• Not a monetized product. The forecasts, research console, cohort builder, and cluster detector are free and open, and will remain so. If Hermes ever generates revenue it will come from research infrastructure services, never from paywalling access to the analysis.
• Not an advocacy organization. Hermes has no position on disclosure, policy, or the nature of UAP. It has a position on rigor.

Guardrails we hold ourselves to

1. Forecasts always show uncertainty intervals. Never a point estimate.
2. Forecast language is always "report activity" or "reporting conditions." Never "UAP activity."
3. Every published forecast is archived with its version, inputs, and outputs, so we can run a backtest against reality.
4. Every analysis is reproducible by hash. If someone can't rerun our numbers, we shouldn't have published them.
5. When a forecast appears to "hit," we respond with the pre-written communication protocol: we show the backtest, we show the noise distribution, we do not claim prescience.
6. The observer effect is real. If publishing a forecast changes report behavior in a region, we document the correction in the changelog.

Who Hermes is for

Long horizon

The goal, measured in years, is for Hermes analyses to become citable in the same way arXiv preprints are: a researcher writes a paper, references a Hermes reproducibility hash, and a reviewer runs the same hash and gets the same numbers. If the UAP field is ever going to be taken seriously as a subject of scientific study, it needs infrastructure that works like that. Hermes is an attempt to build it.