---
analysis-role: guide
analysis-category: guide
confidence-level: medium
confidence-rationale: Skywatcher's published class definitions and Discovery Framework are directly reviewable, but the underlying raw sensor packages and independent validation remain incomplete or unavailable publicly.
evidence-status: mixed
ai-analysis: true
accuracy-disclaimer: AI-assisted analysis; interpretations are provisional and may contain errors. Verify against Skywatcher's published research, cited scientific standards, and linked local source material.
ai-generated: true
companion-eligible: true
topic-slug: skywatcher-nine-uap-classes
topic-title: Skywatcher's Nine UAP Classes
topic-description: A field guide, methodology audit, and archive crosswalk for Skywatcher's preliminary UAP taxonomy.
topic-audience: public-readers
topic-priority: 5
---

# Skywatcher's Nine UAP Classes: Field Guide, Evidence Audit, and Archive Crosswalk

![Nine-class interpretive field guide](/media/Release_2/Analysis/images/skywatcher-nine-class-field-guide-interpretive.png)

## Cover Line

Skywatcher has published one of the most visually memorable modern UAP taxonomies: Tetra, Tic Tac, Blob, Beam, Manta Ray, Bright Star, Jellyfish, Hornet, and Egg. The names are useful. They make complicated observations easier to compare. The hard question is whether the public evidence shows nine repeatable physical classes—or nine preliminary bundles of reported shape, motion, sensor response, and interaction.

The cover plate is an original AI interpretive reconstruction derived only from Skywatcher's written profiles. It does not reproduce Skywatcher's imagery, depict a known event, establish relative scale, or serve as source evidence. The visible disclaimer is part of the image so it cannot be separated from that boundary.

## Working Verdict

Skywatcher's nine classes are useful as a **preliminary observational vocabulary**. They are not yet a publicly validated catalog of distinct craft types.

That distinction matters. A useful field label can help an observer ask better questions without establishing what caused the observation. In the public pages reviewed on August 14, 2026, all nine entries remain marked `Preliminary Classification`; the taxonomy is described as living and exploratory; and the linked EO, IR, radar, RF, or full-report sections are still presented as `Coming Soon` or `Report Generation in Progress`. Skywatcher may possess non-public data, but the public record currently lets readers inspect detailed claims more readily than the measurements behind them.

The most constructive reading is therefore neither “nine confirmed craft” nor “nine meaningless names.” It is: **nine hypotheses about recurring observable packages, awaiting event-level data, explicit class boundaries, blinded assignment, calibrated multi-sensor corroboration, and independent replication.**

## How To Read The Nine Classes

These are not pure shape categories. Each class mixes several dimensions:

| Dimension | Examples inside the taxonomy | Why it matters |
| --- | --- | --- |
| Morphology | Tetrahedron, cylinder, orb, manta, tendrils, egg | Shape can be useful, but viewpoint, blur, focus, bloom, compression, and silhouette can change it. |
| Kinematics | Tumble, hover, oscillate, jump, descend, arrest | Motion can separate classes only when camera/platform motion and tracking behavior are controlled. |
| Spectrum | EO-visible, IR-only, strong heat, radar, RF | A class defined partly by sensor channel requires calibrated, synchronized sensors and known detection thresholds. |
| Grouping | Singleton, pair, group of 1–3, formation of 3–20+ | Group size may describe one event family rather than a stable object type. |
| Interaction | Laser response, jamming response, electronic or physiological effects | Correlation, timing, controls, and independent instrumentation are essential before inferring response or intent. |
| Interpretation | Craft, light phenomenon, plume, active camouflage, awareness | Interpretation must remain downstream of observation and measurement. |

The result is best called a **hybrid phenotype/signature taxonomy**: a field shorthand for how an event reportedly looks and behaves across instruments. That can be more useful than shape alone, but it also creates circularity risk. If “Tic Tac” means a cylinder that hovers, elongates, descends, arrests, affects electronics, and appears near a Manta Ray, a new event can appear to confirm the class simply because only matching details were selected for comparison.

## Source Basis

Skywatcher currently describes itself as an [independent private initiative](https://skywatcher.ai/) focused on aerial intelligence, advanced sensing, and unconventional aerospace research. Its [About page](https://skywatcher.ai/about) presents a team drawn from special operations, aviation, intelligence, aerospace, and research backgrounds. Those backgrounds establish organizational context; they do not independently validate the taxonomy.

The [Skywatcher research index](https://skywatcher.ai/research) calls the system a living taxonomy and says its class hypotheses are exploratory rather than conclusive scientific fact. [Skywatcher Part II](https://skywatcher.ai/media) introduced the nine labels publicly alongside the group's electromechanical-signaling and neuromeditative-interaction work.

The April 2025 [Skywatcher Discovery Framework](https://drive.google.com/file/d/1sDP2QTCAmBL9PhXfL7ySycFdTw0jy821/view)—also available through a clearly identified [PDF mirror](https://anomalyarchives.org/wp-content/uploads/Skywatcher-Discovery-Framework_Final.pdf)—proposed six stages:

| Level | Framework purpose | Public status relevant to this guide |
| --- | --- | --- |
| 1 — Preliminary Observation | Document and triage a report; remove obvious known objects and hoaxes. | Compatible with a descriptive class label. |
| 2 — Structured Data Collection | Use instruments, timestamps, geolocation, metadata, and known-object checks. | Necessary before a class can carry more than descriptive weight. |
| 3 — Analysis and Hypothesis Testing | Measure properties, test conventional explanations, and advance only if anomalous traits remain. | The framework self-assessed electromechanical signaling at this level in April 2025. |
| 4 — Independent Verification and Peer Review | Share data and methods for external reanalysis and corroboration. | No official closing report or inspectable independent-validation dataset was located on Skywatcher's official public pages during this review. |
| 5 — Public Disclosure and Review | Publish validated data and rigorous papers for broad scrutiny. | Class descriptions are public; the underlying class packages reviewed here are not yet publicly inspectable. |
| 6 — Full Discovery and Integration | Move a conclusively supported phenomenon into accepted knowledge and policy. | Not publicly established for any of the nine classes. |

The framework assessed Electromechanical Signaling at Level 3 and Neuromeditative Interaction at Level 2, with a stated goal of reaching a decisive result by the end of 2025. As of the access date, no official public result located by this review closes that milestone. This is a **source-bounded finding about the public record**, not a claim about what Skywatcher may have completed privately.

## Public Evidence-Maturity Scale

This guide uses a five-step scale:

1. **Description only** — a detailed written claim, but no inspectable event package sufficient for independent review.
2. **Illustrative media** — public media exists, but provenance, calibration, raw context, or measurements are insufficient.
3. **Documented observation** — time, location, equipment, acquisition context, processing, and measurements are substantially available.
4. **Corroborated event** — synchronized independent sensors or observers support the same event under documented controls.
5. **Replicated class** — independent teams repeatedly recover the same preregistered observable package and distinguish it reliably from controls and neighboring classes.

On the reviewed class pages, all nine remain at **Level 1 for public evidence maturity**. That rating is deliberately narrower than a judgment about Skywatcher's private holdings. The public descriptions are rich; the event-level packages needed for Level 3 or above are not publicly downloadable from the reviewed class pages.

## Nine-Class Matrix

| Class | Morphology / claimed size | Motion | Claimed sensor channels | Grouping / interaction | Public maturity | Strongest control |
| --- | --- | --- | --- | --- | ---: | --- |
| I — Tetra | Dark tetrahedron; 1.5–2 m | Multi-axis tumble on stable path; descent; alleged evasive jinking | EO strobe, strong IR, claimed radar/RF context | Groups of 3–20+; alleged laser-response behavior | 1 | Balloons or debris with apparent rotation; camera tracking loss; shape instability across frames |
| II — Tic Tac | Bright cylinder; 9–12 m; apparent elongation/color change | Hover, oscillation, rapid descent and arrest | EO/IR; claimed electronics/aircraft effects | Groups of 1–3; Manta Ray proximity; reported physiological effects | 1 | Range/parallax, autofocus/bloom, aircraft or balloon, projected/plasma signature, track reassignment |
| III — Blob | Red cloud-orb; 3–5 m; flare expansion | Lateral oscillation; point-to-point jumps | EO and strong IR claim | Singleton; no direct interaction reported | 1 | Defocus, saturation, atmospheric scintillation, compression, camera motion, hot conventional target |
| IV — Beam | IR-only orb; unknown size | Hours-long hover; slow motion; lateral vibration | IR present, EO absent | Pairs; claimed laser/jamming response | 1 | Celestial body, hot pixel, sensor registration, stabilization, distant stationary source |
| V — Manta Ray | Dark manta silhouette; 2–3 m; colored emission | Tumble, hover, altitude shifts | EO/IR and some radar claims | Rare; Tic Tac proximity; claimed concealment | 1 | Bird, kite, aircraft aspect, balloon, perspective, compression, wind drift |
| VI — Bright Star | Reflective tetrahedral/angular form; 1.5–3 m | Rapid horizontal oscillation | EO/IR; alleged radar artifact and directed-energy effect | Claimed team sensor disruption | 1 | Star/planet, diffraction, autofocus, saturation, rolling shutter, atmospheric scintillation |
| VII — Jellyfish | Pale head with tendrils; head ~2 m, extensions 3–5 m | Serpentine path; “seeking” head | EO/IR/radar claims | Alleged localized electromagnetic disruption | 1 | Balloon/payload train, debris, parallax, stabilization, enhancement or segmentation artifact |
| VIII — Hornet | Dual-body form; tendrils and folding tail; 2–4 m | Irregular asynchronous rotation | IR and radar claims; brief optical visibility | Alleged diversion and descent toward team | 1 | Multipart balloon/payload, dangling debris, viewpoint change, segmentation, rolling shutter |
| IX — Egg | Off-white ovoid; unknown size | Undocumented | One low-quality EO recording; no IR/radar validation | No direct interaction; no confirmed repeat | 1, weakest | Balloon, blur, bloom, distant aircraft, compression, featureless object under poor resolution |

## Nine Class Dossiers

### Class I — Tetra

**Claimed observable package.** Skywatcher's [Tetra page](https://skywatcher.ai/research/tetra) describes a dark tetrahedral object, roughly 1.5–2 meters, tumbling across several axes while its overall path remains stable. It is said to appear in groups from three to more than twenty, show EO strobing and a strong IR signature, descend from high altitude, and evade laser targeting.

**What is public.** The definition, size estimate, behavior narrative, and claimed sensor/interaction profile are public. The page's EO, IR, radar, RF, and full-report sections are not yet publicly downloadable. No event count, per-event inclusion table, raw sequence, calibration record, or independent replication accompanies the public definition.

**Ordinary/artifact lane.** A tumbling dark object can change silhouette dramatically. Balloons with payloads, lightweight debris, birds, distant aircraft at changing aspect, compression, and an unstable tracker can create angular or intermittent shapes. “Stable path plus tumble” is especially sensitive to whether the background and sensor boresight were registered independently.

**Disclosure-forward lane.** If calibrated EO and IR sequences repeatedly showed a rigid tetrahedral boundary rotating across several axes while its center-of-motion followed a controlled path, and if independent radar/laser systems registered synchronized evasive changes, the package would be unusual and operationally important.

**Archive comparison.** [UFOs in formation over Persian Gulf?](/?open=Release_2%2Fvideo_2605_DOD_111719833_DOD_111719833.mp4) and [C11 — Fleet, Cluster, and Multiple-Contact Behavior](/?open=Release_2%2FAnalysis%2FC11-Fleet-Cluster-and-Multiple-Contact-Behavior.md) preserve a multiple-contact question. They do not establish tetrahedral shape, tumble, size, sensor spectrum, or laser response. This is a **group-behavior comparison, not classification**, and it is too incomplete for the main representative crosswalk.

**Advance / merge / falsify.** Advance with multiple calibrated events showing the same 3D boundary and rotation under independent sensors. Merge with Bright Star if reflectivity/color differences disappear under exposure-normalized data. Falsify a distinct Tetra class if silhouettes track aspect, blur, or balloon/debris controls rather than a stable rigid geometry.

### Class II — Tic Tac

**Claimed observable package.** The [Tic Tac page](https://skywatcher.ai/research/tic-tac) describes a bright cylinder, estimated at 9–12 meters, that hovers, oscillates, changes apparent length and color, rapidly descends from high altitude, arrests motion, and sometimes operates near Manta Ray observations. It also reports electronic, aircraft, and physiological effects.

**What is public.** The public page gives unusually specific size and performance claims, including a Mach-plus assessment, but does not expose the measurement chain needed to reproduce those values. EO/IR packages and the analytical report remain unavailable on the reviewed page; the page also says there was one laser-range return but no radar acquisition in the described interaction material.

**Ordinary/artifact lane.** Range uncertainty can turn camera motion into extreme target motion. Bloom and autofocus can change apparent length and color. Aircraft, balloons, distant rockets, sensor mode changes, track switching, or a projected/plasma signature can imitate parts of the package. None explains every claim automatically; each needs event-specific geometry and synchronized evidence.

**Disclosure-forward lane.** A material object would become a serious hypothesis if independent sensors fixed range, recovered a smooth cylindrical boundary across aspect changes, measured rapid descent and arrest without track reassignment, and correlated any electronic effect to the same timeline.

**Archive comparison.** [USCG C-144 Tyndall UAP 1 TIC TAC IR hot, April 24, 2024](/?open=Release_2%2Fvideo_2605_DOD_111720693_DOD_111720693.mp4) and [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) support a repeatable elongated sensor-visible silhouette in a bounded interval. They do not supply the Skywatcher size, color transition, descent/arrest, grouping, or interaction package. It is a **shape/channel comparison, not classification**.

**Advance / merge / falsify.** Advance with calibrated range, centroid and boundary tracking, synchronized EO/IR/radar, and independent event repetition. Split if the apparent cylinder separates into a material-body family and a projected/bloom family. Falsify the combined class if elongation and rapid motion follow sensor state, saturation, or track reassignment rather than the source.

### Class III — Blob

**Claimed observable package.** Skywatcher's [Blob page](https://skywatcher.ai/research/blob) describes a singleton red cloud-like orb, roughly 3–5 meters, with a pulsing core, strong IR signature, rapid flare-driven expansion, lateral oscillation, and abrupt point-to-point movement.

**What is public.** The morphology, size, heat, flare, and motion claims are public in prose. The page reports no direct interaction. EO, IR, radar, and full analytical packages remain forthcoming on the reviewed page, so physical expansion and positional jumps cannot yet be distinguished publicly from sensor behavior.

**Ordinary/artifact lane.** Saturation and defocus make a small bright source expand. Atmospheric scintillation changes intensity and color. Compression and aggressive stabilization can create jumps, while camera or platform motion can shift a target against a low-detail background. A hot conventional target can dominate IR without being physically cloud-like.

**Disclosure-forward lane.** A recurring compact source whose calibrated angular boundary, radiometry, temperature estimate, and position change together across independent EO/IR/radar systems would be difficult to reduce to simple bloom or compression. A state-changing luminous phenomenon would then be a live physical model.

**Archive comparison.** [Spherical UAP pulsing over water (CALLSIGN)](/?open=Release_2%2Fvideo_2605_DOD_111719741_DOD_111719741.mp4), especially [157.178s](/?open=Release_2%2Fvideo_2605_DOD_111719741_DOD_111719741.mp4&t=157.178), and [C08 — Formation, Cloud, and Atmospheric Interaction](/?open=Release_2%2FAnalysis%2FC08-Formation-Cloud-and-Atmospheric-Interaction.md) provide a pulsation/saturation comparison. That source was digitally altered before upload, and the bright burst may be physical emission, bloom, compression, or contrast overflow. It is a **pulse-and-control comparison, not classification**.

**Advance / merge / falsify.** Advance with raw radiometry, known exposure/gain, calibrated angular size, stable background registration, and synchronized range. Merge with Beam if color and EO visibility follow sensor thresholds rather than distinct phenomena. Falsify physical expansion if the boundary scales with exposure, point-spread function, compression, or focus.

### Class IV — Beam

**Claimed observable package.** The [Beam page](https://skywatcher.ai/research/beam) describes paired, IR-only orb signatures that can hover for more than an hour, move slowly, vibrate laterally, and allegedly respond to lasing or jamming. Skywatcher also explicitly states that some sightings may be misattributed celestial bodies because of sensor limits or viewing angle.

**What is public.** The class page's self-critique is important: it preserves a conventional control inside the claimed class. The reviewed page provides no public IR sequence, synchronized EO non-detection record, star/planet registration, detector calibration, or independent response trial.

**Ordinary/artifact lane.** A star, planet, distant stationary hot source, hot pixel, stabilization artifact, or misregistered EO/IR field can produce apparent “IR only” persistence. Lateral vibration may come from platform jitter or tracking. A fade during jamming can reflect sensor state rather than target awareness unless command timing and detector behavior are independently recorded.

**Disclosure-forward lane.** A true sensor-aware event would require repeated randomized stimulus trials, blinded timing, independent instruments, registered celestial controls, and a response that survives sensor-state and operator-expectation checks.

**Archive comparison.** No local item supports the full paired, hours-long, EO-absent, stimulus-responsive package. [C14 — Conventional, Misleading, and Balloon Flags](/?open=Release_2%2FAnalysis%2FC14-Conventional-Misleading-and-Balloon-Flags.md) is the correct adjacent route because it keeps stationary, astronomical, drift, and sensor controls visible. **No adequate local parallel** is the finding.

**Advance / merge / falsify.** Advance with synchronized calibrated EO/IR, sky-coordinate registration, randomized stimuli, preregistered response metrics, and independent observers. Merge with Blob if spectrum and grouping vary continuously with sensor threshold. Falsify a responsive class if fades and “dodges” track platform commands, jitter, or celestial motion.

### Class V — Manta Ray

**Claimed observable package.** The [Manta Ray page](https://skywatcher.ai/research/manta-ray) describes a dark manta-shaped form, estimated at 2–3 meters, with colored emissions, multi-axis tumble, hover, rapid altitude change, and occasional proximity to Tic Tac observations. It claims EO/IR detection and some radar tracking.

**What is public.** The page presents the package as rare and includes strong interpretations such as concealment or evasion. The public sensor packages, radar tracks, processing history, and event table are not available for independent inspection on the reviewed page.

**Ordinary/artifact lane.** Birds, kites, aircraft at unusual aspect, balloons, and lightweight debris can create manta-like silhouettes. Wind-relative movement is difficult to establish without local wind profiles and range. Colored edge emissions can be chromatic aberration, compression, navigation lights, or bloom.

**Disclosure-forward lane.** A rigid body with repeatable manta geometry, independent range/altitude, motion against measured winds, synchronized radar and EO/IR, and correlated Tic Tac proximity would create a strong operational case. Proximity alone would still not prove common origin.

**Archive comparison.** [Western US Event](/?open=Release_1%2FWestern_US_Event_Slides_5.08.2026.pdf) includes witness language likening one large, seemingly transparent phenomenon to a translucent kite. [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) supplies the morphology/range caution. The packet has no direct technical data and does not establish Skywatcher's size, tumble, emissions, proximity, or sensor package. It is a **silhouette-language comparison, not classification**.

**Advance / merge / falsify.** Advance with calibrated multi-view geometry, range, wind field, radar, and repeated shape persistence. Merge with Tetra if “manta” versus “tetra” follows aspect of one rotating geometry. Falsify a distinct class if silhouettes and motion match birds, kites, aircraft, balloons, or compression under controlled reconstruction.

### Class VI — Bright Star

**Claimed observable package.** Skywatcher's [Bright Star page](https://skywatcher.ai/research/bright-star) describes a reflective tetrahedral/angular object, estimated at 1.5–3 meters, flashing teal, green, blue, red, and white; rapidly oscillating horizontally; producing a propeller-like radar artifact; and allegedly disrupting team sensors through directed energy.

**What is public.** This class makes strong cross-sensor and causal claims, but the reviewed page does not expose the synchronized EO/IR/radar/RF timelines, detector state, energy measurements, or full analysis needed to test them.

**Ordinary/artifact lane.** Stars and planets under atmospheric scintillation can flash colors. Defocus, diffraction, chromatic aberration, saturation, rolling shutter, and tracking oscillation can make a point source appear angular, extended, or rapidly vibrating. Radar artifacts must be separated from actual target dynamics.

**Disclosure-forward lane.** A compelling event would require a resolved angular boundary larger than the calibrated point-spread function, synchronized multi-band position, independent radar track, recorded sensor-health telemetry, and a disruption whose timing and energy cannot be attributed to the observing system.

**Archive comparison.** No local source supports the complete reflective tetrahedral, multicolor, high-frequency oscillation, radar, and directed-energy package. [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) remains the best control route. **No adequate local parallel** is the finding.

**Advance / merge / falsify.** Advance with point-spread calibration, raw color channels, synchronized radar and sensor-health logs, and independent repetition. Merge with Tetra if brightness and color are exposure states of one angular family. Falsify physical oscillation or directed energy if both track detector, gimbal, focus, rolling-shutter, or telemetry faults.

### Class VII — Jellyfish

**Claimed observable package.** The [Jellyfish page](https://skywatcher.ai/research/jellyfish) describes a pale bulbous head and 3–5-meter tendril-like extensions, winding or serpentine movement, independent-looking tail behavior, EO/IR/radar visibility, and alleged electromagnetic disruption. The page acknowledges possible confusion with Bright Star or Hornet at distance.

**What is public.** The explicit confusion note is evidence that boundaries are not clean. Public event sequences, radar correlation, airflow estimates, electromagnetic telemetry, and processing records are not available on the reviewed page.

**Ordinary/artifact lane.** Balloon-and-payload trains, suspended equipment, debris, parallax, stabilization, rolling shutter, compression, and enhancement/segmentation artifacts can create heads, tendrils, or independent-looking appendages. A tail that seems insensitive to wind requires measured wind and range, not visual intuition alone.

**Disclosure-forward lane.** A coherent articulated body would become plausible if independent views preserved the same joints and relative motion, calibrated range established physical dimensions, wind and platform motion were modeled, radar followed the same structure, and electromagnetic effects were synchronized and repeatable.

**Archive comparison.** [Mission Report, Arabian Gulf, 2020](/?open=Release_1%2FDOW-UAP-D7-Mission-Report-Arabian-Gulf-2020.pdf), [DOD 111720731 video](/?open=Release_2%2Fvideo_2605_DOD_111720731_DOD_111720731.mp4), and [C14 — Conventional, Misleading, and Balloon Flags](/?open=Release_2%2FAnalysis%2FC14-Conventional-Misleading-and-Balloon-Flags.md) provide balloon/drift/payload controls, not a positive match. There is **no adequate local parallel** for the full Jellyfish package.

**Advance / merge / falsify.** Advance with multi-view articulation, airflow modeling, raw adjacent frames, radar correlation, and synchronized EM telemetry. Merge with Hornet or Bright Star if boundary assignments show high reviewer confusion. Falsify an articulated body if tendrils depend on enhancement, compression, stabilization, or a suspended payload model.

### Class VIII — Hornet

**Claimed observable package.** The [Hornet page](https://skywatcher.ai/research/hornet) describes a 2–4-meter dual-body or multipart form with hanging tendrils, a tail that folds or extends, and asynchronous upper/lower rotation. It claims IR/heat and radar signatures and says the components may be connected, although that remains unconfirmed.

**What is public.** Skywatcher's own page says the class lacks a consistent shape or motion pattern and can resemble Bright Star or Jellyfish. Those admissions make Hornet the clearest test of whether the taxonomy's boundaries are reproducible. The public radar, IR, EO, and full event packages remain forthcoming.

**Ordinary/artifact lane.** Multipart balloons, balloon payloads, dangling debris, tethered structures, perspective changes, segmentation, rolling shutter, and low-resolution motion can create asynchronous “components” and folding-tail impressions.

**Disclosure-forward lane.** A mechanically articulated or morphing object would require persistent component identities across adjacent raw frames and viewpoints, recovered depth/occlusion relationships, synchronized radar/IR, and repeated articulation that cannot be modeled as suspended payload motion.

**Archive comparison.** The same [Arabian Gulf balloon-like report](/?open=Release_1%2FDOW-UAP-D7-Mission-Report-Arabian-Gulf-2020.pdf), [DOD 111720731](/?open=Release_2%2Fvideo_2605_DOD_111720731_DOD_111720731.mp4), and [C14 control report](/?open=Release_2%2FAnalysis%2FC14-Conventional-Misleading-and-Balloon-Flags.md) define mandatory ordinary controls. They do not support a Hornet assignment. **No adequate local parallel** is the finding.

**Advance / merge / falsify.** Advance with component tracking, occlusion/depth evidence, fixed-recipe adjacent-frame review, synchronized radar/IR, and independent recurrence. Merge with Jellyfish if inter-rater confusion stays high. Falsify a connected dual body if apparent components follow segmentation, viewpoint, rolling shutter, or balloon-payload dynamics.

### Class IX — Egg

**Claimed observable package.** Skywatcher's [Egg page](https://skywatcher.ai/research/egg) is unusually candid: an off-white ovoid, size unknown, represented by one low-quality EO recording, with no corroborated motion, structural detail, radar track, or IR coverage. Skywatcher calls the designation provisional and says repeat sightings are not confirmed.

**What is public.** This is closer to a morphology lead than a class. The page's restraint is methodologically healthy, but the public recording and full report are still forthcoming, so even the basic shape cannot yet be independently tested against blur, bloom, range, or aspect.

**Ordinary/artifact lane.** A balloon, distant aircraft, featureless object, blur, bloom, compression, or low-resolution viewing angle can produce an egg-like silhouette. With a single low-quality record and no motion/range, physical size and surface properties are indeterminate.

**Disclosure-forward lane.** An ovoid vehicle class would require repeated independent events with stable 3D geometry, range, motion, multi-sensor data, and exclusion of balloons and ordinary aircraft. The current public package does not reach that threshold.

**Archive comparison.** [DOD 111719799 video](/?open=Release_2%2Fvideo_2605_DOD_111719799_DOD_111719799.mp4) and [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) preserve a compact bright contact that can read as teardrop, egg, or bright core with a short wake/bloom. The background is low-detail and the extension may be smear, compression, or sensor bloom. It is an **ovoid-appearance comparison, not classification**.

**Advance / merge / falsify.** Advance with repeat events, raw EO, calibrated range, multi-view geometry, motion, and IR/radar coverage. Merge with Tic Tac if aspect or bloom explains egg-versus-cylinder assignments. Falsify a distinct Egg class if the silhouette is not stable across resolution, aspect, and ordinary-object reconstructions.

## Methodology Audit

The Discovery Framework's staged logic is sensible: collect, analyze, verify independently, then disclose. The public taxonomy does not yet show that those gates have been cleared at class level.

| Audit question | What is public now | What a validation package needs |
| --- | --- | --- |
| How many events support each class? | No per-class event count is published on the reviewed pages. | Event ledger with unique IDs, dates, sites, sensors, duplicates, exclusions, and class votes. |
| How were events selected? | “Most consistently observed patterns” is stated, but sampling and rejection rules are not shown. | Inclusion/exclusion criteria defined before class fitting; rejected and conventional cases retained. |
| Are dimensions explicit? | Shape, size, behavior, sensor and interaction prose are present. | Machine-readable variables, units, required versus optional features, uncertainty, and missing-data rules. |
| Are class boundaries mutually exclusive? | Several pages acknowledge overlap; Tic Tac/Manta proximity is part of the descriptions. | Decision rules, hierarchical or multilabel policy, reviewer confusion matrix, and merge/split thresholds. |
| Can reviewers reproduce assignments? | No public inter-rater result was located. | Blinded independent classification, Cohen's kappa or equivalent, disagreement adjudication, and held-out events. |
| Are measurements calibrated? | Sensor types are named, but event calibration and metadata are not public on the class pages. | Sensor model, mode, time sync, field of view, point-spread function, radiometric/geometric calibration, platform state, weather and celestial context. |
| Is processing traceable? | No class-level raw-to-published processing chain is public. | Raw files, hashes or chain-of-custody record, processing steps, software settings, and unprocessed controls. |
| Are interaction claims causal? | Laser, jamming, electronic, physiological, or directed-energy effects are reported in prose. | Randomized/blinded stimulus trials, synchronized command and sensor-health logs, baseline/control periods, effect-size and false-positive analysis. |
| Has the taxonomy replicated independently? | No public independent replication package was located. | External team, independent equipment/site, preregistered criteria, comparable hit/false-alarm rates, and open event-level results. |

### The circularity problem

Classes should be discovered on one dataset and tested on another. If the same observations generate the labels, tune the boundaries, and demonstrate the categories, the result can overfit a small or selected sample. A future release should separate:

1. A **development set** used to propose dimensions and class boundaries.
2. A **locked rule set** specifying required, optional, and exclusionary features.
3. A **blind validation set** assigned by reviewers who do not know the desired result.
4. An **external replication set** gathered by an independent team.

### A cleaner data model

The most durable part of Skywatcher's work may be the dimensions rather than the names. Record each event as a vector before assigning a label:

`morphology + angular size + range confidence + kinematics + spectrum + grouping + interaction + environment + sensor quality + processing state`

That allows researchers to learn whether nine clusters actually emerge, whether some classes merge, whether one class divides, or whether several labels are sensor states of the same underlying phenomenon.

## Comparison With Other Research Standards

### NASA: calibration before conclusion

NASA's [UAP Independent Study](https://science.nasa.gov/uap/) and [final report](https://smd-cms.nasa.gov/wp-content/uploads/2023/09/uap-independent-study-team-final-report.pdf) emphasize systematic calibration, multiple measurements, thorough sensor metadata, multispectral acquisition, and clear evidence thresholds. NASA also notes that limited high-quality observations prevent firm scientific conclusions. Applied here, a vivid class description is a starting hypothesis; calibrated event data determine whether it survives.

### AARO: morphology is a report field, not an identity

The [AARO FY2025 report](https://www.aaro.mil/Portals/136/PDFs/FY25%20UAP%20Annual%20Report/AARO_FY2025_Consolidated_Annual_Report_on_UAP.pdf) distinguishes reported morphology from analytic resolution, describes collection bias, and retains insufficient-data cases in an active archive. Its current [reporting trends](https://www.aaro.mil/UAP-Cases/UAP-Reporting-Trends/) explicitly warn that not every report contains morphology data. Lights, spheroids, cylinders, Tic Tacs, polygons, and other shapes are useful descriptors, not proof of separate vehicles.

AARO also reports that missing timely, actionable sensor data continues to constrain resolution. That point cuts both ways: unresolved does not mean extraordinary, and insufficient public data does not prove a claim false. It means the classification must remain provisional.

### Galileo Project: synchronized multimodal observatories

The Galileo Project's [multimodal observatory design](https://projects.iq.harvard.edu/sites/projects.iq.harvard.edu/files/galileo/files/s2251171723400068.pdf) uses calibrated, synchronized optical, IR, radar, RF, acoustic, and environmental instruments, with secondary stations for triangulation and corroboration. Its [computing architecture](https://projects.iq.harvard.edu/sites/projects.iq.harvard.edu/files/galileo/files/s2251171723400081.pdf) places independent streams on a unified timeline and preserves data integrity.

Skywatcher's public descriptions invoke many of the same sensor families. The gap is inspectability: a class-level validation package needs the unified event timeline, not only the statement that multiple sensors participated.

### SCU: dimensions first, clusters second

SCU's [2025 cluster analysis](https://www.explorescu.org/post/cluster-analysis-of-features-associated-with-unidentified-anomalous-phenomena-described-in-216-selec) analyzed 216 complete reports using five declared dimensions: shape, estimated size, ability to hover, electromagnetic effects, and sound. It reported seven clusters with a silhouette value of 0.6 and found shape to be the strongest, but not only, driver.

That work shows a legitimate path from report features to statistical clusters. It does **not** validate Skywatcher's nine labels: the sample, selection rules, variables, acquisition methods, and outcomes differ. It does show how Skywatcher could make the taxonomy testable—publish event rows, define missing data, lock features, report cluster quality, and compare statistical clusters with human labels.

## Archive Crosswalk

The local archive is not a Skywatcher training set. These links give a case officer nearby shapes, behaviors, and controls to inspect.

| Skywatcher question | Exact local source | Analysis route | Shared dimension | Missing or contrary dimensions | Boundary |
| --- | --- | --- | --- | --- | --- |
| Can an elongated IR silhouette stay stable across adjacent frames? | [USCG C-144 Tyndall UAP 1 TIC TAC IR hot](/?open=Release_2%2Fvideo_2605_DOD_111720693_DOD_111720693.mp4) | [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) | Elongated/ovoid appearance; IR visibility | No Skywatcher size, color shift, descent/arrest, grouping, or electronic effect | Tic Tac comparison, not classification |
| Does a pulse/flare represent physical expansion or sensor bloom? | [Spherical UAP pulsing over water](/?open=Release_2%2Fvideo_2605_DOD_111719741_DOD_111719741.mp4&t=157.178) | [C08 — Formation, Cloud, and Atmospheric Interaction](/?open=Release_2%2FAnalysis%2FC08-Formation-Cloud-and-Atmospheric-Interaction.md) and [C07 — Field, Propulsion, and Morphology Model](/?open=Release_2%2FAnalysis%2FC07-Field-Propulsion-and-Morphology-Model.md) | Pulsing bright contact and flare/burst question | Digitally altered upload; no red-cloud boundary, range, or confirmed positional jump | Blob comparison, not classification |
| Can group behavior be separated from shape identity? | [UFOs in formation over Persian Gulf?](/?open=Release_2%2Fvideo_2605_DOD_111719833_DOD_111719833.mp4) | [C11 — Fleet, Cluster, and Multiple-Contact Behavior](/?open=Release_2%2FAnalysis%2FC11-Fleet-Cluster-and-Multiple-Contact-Behavior.md) | Multiple contacts / formation question | No Tetra geometry, tumble, size, or stimulus response | Group-behavior comparison only |
| How fragile is an egg/teardrop reading? | [DOD 111719799 video](/?open=Release_2%2Fvideo_2605_DOD_111719799_DOD_111719799.mp4) | [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) | Ovoid/teardrop appearance | Bloom, smear, compression, range and motion unresolved | Egg comparison, not classification |
| Can witness silhouette language carry a Manta Ray comparison? | [Western US Event](/?open=Release_1%2FWestern_US_Event_Slides_5.08.2026.pdf) | [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) | Large translucent-kite description | No direct technical data; no size, tumble, colored emission, Tic Tac proximity or radar/IR package | Manta Ray comparison, not classification |
| What ordinary controls fit irregular silhouettes and tendrils? | [Mission Report, Arabian Gulf, 2020](/?open=Release_1%2FDOW-UAP-D7-Mission-Report-Arabian-Gulf-2020.pdf) and [DOD 111720731](/?open=Release_2%2Fvideo_2605_DOD_111720731_DOD_111720731.mp4) | [C14 — Conventional, Misleading, and Balloon Flags](/?open=Release_2%2FAnalysis%2FC14-Conventional-Misleading-and-Balloon-Flags.md) | Balloon-like appearance, wind travel, soft edge, payload/tether controls | No articulated body, serpentine powered motion, or corroborated EM/radar package | Jellyfish/Hornet controls, not positive matches |
| How should extreme motion claims be bounded? | [Syrian UAP instant acceleration](/?open=Release_2%2Fvideo_2605_DOD_111719715_DOD_111719715.mp4&t=188.795) | [C09 — Instant Acceleration, Displacement, and Vanishings](/?open=Release_2%2FAnalysis%2FC09-Instant-Acceleration-Displacement-and-Vanishings.md) | Image-plane displacement and disappearance testing | Not evidence of any Skywatcher shape class; range and track state control physical claims | Kinematic method comparison only |

The sparse result is intentional. Beam, Bright Star, Jellyfish, and Hornet have **no adequate local positive parallel** for their full observable packages. Tetra has only an incomplete grouping question. Filling every class with a look-alike would convert this guide into the classification lane it was designed to avoid.

## Interpretive Reconstruction

The nine-panel plate visualizes the written claims at equal weight:

- Tetra: dark angular body plus rotational-axis cues.
- Tic Tac: smooth bright cylinder.
- Blob: red pulsing cloud-orb.
- Beam: paired IR-style signatures.
- Manta Ray: dark manta silhouette with restrained emission.
- Bright Star: reflective angular form with controlled multicolor shimmer.
- Jellyfish: pale head with downward extensions.
- Hornet: multipart body, tendrils, and articulated tail.
- Egg: simple off-white ovoid with deliberately sparse detail.

The plate is falsified as a useful teaching aid if readers confuse it with event imagery, infer shared scale, or treat its visual clarity as evidence that Skywatcher's source data are equally clear. Its role is orientation only.

## Disclosure-Forward Neutrality Gate

### What the public record supports

- Skywatcher has published nine named preliminary profiles and a six-level discovery framework.
- The profiles contain enough dimensional detail to generate testable event questions.
- Skywatcher itself labels the taxonomy living/exploratory and includes uncertainty in at least the Beam and Egg pages.
- The public class pages reviewed here do not expose the event-level packages required for independent class validation.

### Ordinary/artifact lane that remains live

Astronomical sources, balloons and payloads, birds, kites, aircraft aspect, debris, atmospheric scintillation, autofocus, diffraction, saturation, bloom, rolling shutter, compression, stabilization, tracking errors, platform motion, sensor registration, projected/plasma signatures, and ordinary electronic faults remain class-specific controls.

### Disclosure-forward lane that remains live

If Skywatcher releases calibrated raw event packages with synchronized EO/IR/radar/RF timelines, explicit range and motion solutions, randomized interaction trials, stable class boundaries, and independent replication, some packages could represent real recurring phenomena with unusual physical or control characteristics. That possibility is not ruled out by the current public-data gap. It is also not proved by the descriptions.

### What would upgrade the taxonomy

1. Publish per-class event counts and all inclusion/exclusion decisions.
2. Release representative raw packages with time sync, sensor metadata, calibration, platform state, weather, celestial context, and processing provenance.
3. Define minimum required, optional, and exclusionary features.
4. Run blinded inter-rater assignment and publish the confusion matrix.
5. Lock the taxonomy before testing on held-out events.
6. Invite an independent team to collect or reanalyze data under preregistered rules.
7. Publish negative trials and ordinary-control events alongside successes.

### What would weaken, merge, or falsify it

- Low agreement between reviewers would weaken the classes as operational labels.
- Tetra/Bright Star, Blob/Beam, Tic Tac/Egg, or Jellyfish/Hornet may merge if differences follow exposure, spectrum, viewpoint, or processing rather than stable event properties.
- A class should split only when held-out data show reproducible subclusters, not because a striking new image needs a name.
- If claimed motion, shape change, or interaction follows sensor/platform state, ordinary-object reconstruction, or unblinded stimulus timing, the relevant class package is weakened or falsified.

## Working Assessment

Skywatcher's taxonomy is strongest as a **research interface**. It reminds analysts that UAP reports have multiple dimensions and that the right question is not only “what shape was it?” but “what did calibrated sensors record, how did it move, what else was present, what controls were run, and can another team reproduce the assignment?”

Its present weakness is not that the names are imaginative. Scientific taxonomies often begin with useful descriptive names. The weakness is the gap between public specificity of claim and public inspectability of evidence. A 9–12-meter size, a Mach-plus descent, a radar artifact, or a directed-energy effect needs a measurement chain proportionate to the claim.

The correct archive posture is therefore provisional and interested: preserve the nine labels as Skywatcher's own observational vocabulary; use them to formulate tests; compare only exact dimensions; keep ordinary and disclosure-forward lanes visible; and refuse to turn resemblance into identity.

## Case-Officer Path

1. Start with the [Skywatcher research index](https://skywatcher.ai/research) and read its living-taxonomy disclaimer before opening individual classes.
2. Use the nine-panel reconstruction above only as a visual index; then read the exact class page for the claim language.
3. Read the [Discovery Framework](https://drive.google.com/file/d/1sDP2QTCAmBL9PhXfL7ySycFdTw0jy821/view) to understand the evidence gates Skywatcher says it intends to meet.
4. Compare those gates with [NASA's UAP study](https://science.nasa.gov/uap/), the [Galileo multimodal observatory design](https://projects.iq.harvard.edu/sites/projects.iq.harvard.edu/files/galileo/files/s2251171723400068.pdf), the [AARO FY2025 report](https://www.aaro.mil/Portals/136/PDFs/FY25%20UAP%20Annual%20Report/AARO_FY2025_Consolidated_Annual_Report_on_UAP.pdf), and the [SCU cluster study](https://www.explorescu.org/post/cluster-analysis-of-features-associated-with-unidentified-anomalous-phenomena-described-in-216-selec).
5. Inspect one exact local comparison at a time: [Tyndall UAP 1](/?open=Release_2%2Fvideo_2605_DOD_111720693_DOD_111720693.mp4), [Spherical UAP pulsing over water](/?open=Release_2%2Fvideo_2605_DOD_111719741_DOD_111719741.mp4&t=157.178), [Western US Event](/?open=Release_1%2FWestern_US_Event_Slides_5.08.2026.pdf), or [DOD 111719799](/?open=Release_2%2Fvideo_2605_DOD_111719799_DOD_111719799.mp4).
6. Keep [C12 — Shape, Size, and Plane-Backdrop Assessments](/?open=Release_2%2FAnalysis%2FC12-Shape-Size-and-Plane-Backdrop-Assessments.md) and [C14 — Conventional, Misleading, and Balloon Flags](/?open=Release_2%2FAnalysis%2FC14-Conventional-Misleading-and-Balloon-Flags.md) open beside every morphology claim.
7. Use [C07 — Field, Propulsion, and Morphology Model](/?open=Release_2%2FAnalysis%2FC07-Field-Propulsion-and-Morphology-Model.md), [C08 — Formation, Cloud, and Atmospheric Interaction](/?open=Release_2%2FAnalysis%2FC08-Formation-Cloud-and-Atmospheric-Interaction.md), [C09 — Instant Acceleration, Displacement, and Vanishings](/?open=Release_2%2FAnalysis%2FC09-Instant-Acceleration-Displacement-and-Vanishings.md), and [C11 — Fleet, Cluster, and Multiple-Contact Behavior](/?open=Release_2%2FAnalysis%2FC11-Fleet-Cluster-and-Multiple-Contact-Behavior.md) to turn the labels into testable behavior questions rather than identities.

## Follow-Up

- Revisit the public-evidence maturity ratings when Skywatcher releases any class-level EO, IR, radar, RF, metadata, or full analytical package.
- Preserve the current access date so later revisions can identify what changed rather than silently overwriting the evidence state.
- If independent validation appears, test its protocol, preregistration, false-positive controls, and class agreement before upgrading a class.
- Keep the proposed application classification lane parked. A future product feature should be considered only after the observational dimensions and class boundaries prove stable enough that the UI would clarify evidence rather than manufacture certainty.