Field Mac mini
Apple M6 Mac mini with 16 GB memory. This is the capture server: connect it to power and your router’s LAN over Ethernet.
- M6 · 16 GB ↗Listed starting price$899.00
Connect the hardware. Install Codex on your station’s Mac. Give it the prompt below.
01 / Connect your hardware
Field rig $4,778.75 HQ $2,528.97
USD · Price check November 14, 2026 · Before tax and shipping. Includes the used lens and marked budget allowances. Hardware only; excludes subscriptions, internet service, and existing monitors or network equipment. These are replacement estimates, not our purchase receipts.Apple M6 Mac mini with 16 GB memory. This is the capture server: connect it to power and your router’s LAN over Ethernet.
Canon EOS R100 with Kolari full-spectrum conversion and Pro AR/AS coating; Kolari EF–EOS R drop-in adapter with the Kolari Clear dust-protective filter; Canon EF 100–400mm f/4.5–5.6L IS USM lens. Mount the camera on a stable tripod and provide continuous camera power.
Elgato Cam Link 4K. Connect the R100’s micro-HDMI output to the Cam Link’s HDMI input, then connect the Cam Link to the field Mac by USB. Enable clean HDMI output on the camera.
Two Samsung T7 external SSDs: 1 TB for the rolling video buffer and 2 TB for the review backlog. Connect both to the field Mac with data-capable USB-C cables. The buffer’s actual duration depends on the recording bitrate.
Two HackRF One devices, each with a Nooelec Iridium patch antenna: 3 dBi, 1.62 GHz center frequency, SMA connector. Connect the receivers to the Mac through a powered USB hub. These antennas serve our L-band observations; they do not provide uniform sensitivity across the full sweep range. Start in receive-only mode.
A second M6 Mac mini with 24 GB memory handles training and archival work. Our HQ volumes include a 4 TB archive drive, a 1 TB training SSD, and a separate 2 TB backup volume, alongside the HQ work drive. Connect HQ and the field Mac over a private network; captured data stays on the field drives while offline and syncs when connected.
A powered USB hub, continuous camera power, Ethernet and data-capable USB cables, and a tripod for the camera. Keep the Mac, capture hardware, and drives protected from rain with ventilation for heat.
02 / Install Codex on the station
Install Codex for macOS ↗ on the same Mac you connected the hardware to. Sign in, open a project folder on an external SSD, and start a conversation with this prompt. Stay nearby to approve macOS permissions.
Set up this Mac as the field server for the Star People rig below, using: https://github.com/ufo-files/field-station https://github.com/ufo-files/anomaly-vision-model Target hardware: M6 field Mac mini (16 GB); full-spectrum Kolari Canon R100 with Pro AR/AS coating, Kolari EF–EOS R Clear drop-in adapter, Canon EF 100–400mm f/4.5–5.6L IS USM; Cam Link 4K; Samsung T7 1 TB buffer and T7 2 TB backlog; two HackRF Ones with Nooelec 1.62 GHz Iridium patch antennas. HQ uses a separate M6 Mac mini (24 GB) with external archive, training, work, and backup volumes. Confirm the actual devices rather than assuming names or ports. Read the current repository instructions and inspect this Mac, its connected camera, external drives, and any optional HackRF receivers. Use the current Apple Silicon installation path, not legacy Raspberry Pi instructions. Check the hardware and available model artifacts before making changes. Ask only for missing information or permissions that you cannot determine yourself. Install the dependencies and Field Station on an external SSD. Configure the camera for continuous live preview, recording, motion detection, and classification using a compatible published model and its matching decision settings. Verify model compatibility and checksums; do not substitute a different model format or claim setup is complete if the required artifacts are unavailable. Use external storage for video, review data, training frames, models, logs, and backups. Give the two connected SSDs separate roles for the rolling video buffer and review backlog. Identify them by stable volume identity and benchmark their write speed. Do not erase or rename drives without asking. Size retention from measured bitrate and available capacity, with background cleanup of eligible unreviewed data. Preserve manual recordings, human-reviewed anomalous evidence, and training frames and labels. Keep camera preview working if recording must pause. Never silently fall back to the internal disk when an external drive disconnects. Configure the current web interface and required workers as persistent services that recover from worker failures and device reconnection. Keep access private to the local network or a configured tailnet. Explain any macOS camera, removable-volume, login, or administrator prompts I need to approve. Verify what restarts automatically after login or reboot rather than assuming it does. Configure the two HackRF receivers by serial number for receive-only spectrum monitoring, using the repository’s narrow L-band and wide-sweep roles. Account for the patch antennas’ limited useful bandwidth when presenting wide-sweep results. Keep RF transmission disabled. Set up HQ replication once I provide access to my HQ Mac: anomalous evidence packages go to its archive volume, training frames and models to its training volume, with a verified second copy on its backup volume. Keep collection working offline and retry sync on reconnection. Do not delete the field copy before verifying the destination checksum; preserve manual and human-reviewed anomalous recordings. Do not assume a shared folder or transport is already configured. Leave public Internet Archive uploads and YouTube streaming disabled until I supply my own accounts and authorize publication. Do not reuse another station's credentials or publish my location or recordings. Verify a live camera image, a playable test recording, model classification, a saved human review, correct external storage paths, and service restart recovery. Clean up only the disposable test artifacts you created. Give me the local web URL and a concise report of what works and any remaining blockers. Do not claim this enrolls the station in a global network unless the repository actually provides a working enrollment service.
Based on the Field Station and Anomaly Vision Model repositories.