65 Targets and 77 Hours: Cataloguing My Seestar S50 Night Sky

I’ve never written about astronomy here, which is odd, because for the last year and a half one of my favourite gadgets has been a smart telescope. Since March 2025 I’ve been pointing a ZWO Seestar S50 at whatever is up and letting it do its thing, and every file it has produced has been piling up on my NAS ever since.

This week I finally asked Claude Code to make sense of that pile. The result is a note in my Obsidian vault for every target I’ve captured, and some numbers that surprised me.

What the Seestar is

The Seestar S50 is an all-in-one telescope: a 50 mm triplet apochromatic lens with a 250 mm focal length (f/5), a small 2 MP Sony sensor, a motorised mount and a computer, all in one box. You choose a target in the phone app, it finds it, tracks it and takes 10-second exposures, stacking them live as it goes. There is no eyepiece. What you get is a growing image on your phone, and a folder full of files.

It saves each individual frame as a .fit file (the FITS format astronomers use, with all the capture details in a header) along with JPEG previews and its own stacked results. That’s where the problem starts.

The pile

Before this week I couldn’t have told you how much I’d captured. Claude Code on my Windows PC looked at the share on the Synology NAS and came back with:

• 24,857 .fit files, taking up 98.2 GB
• 49,802 JPEGs, another 20 GB
• 44 videos and a handful of other files, for about 119 GB in total
• 118 folders, all at one level, because the Seestar makes separate folders for each target’s stacks, sub-frames, mosaics, photos and videos (M 31, M 31_sub, M 31_mosaic and so on)

Once you fold those variants together, that’s about 65 distinct targets. Nobody is going to look through that by hand, least of all me.

One note per target

The vault already has a note per camera, so the Seestar got one, and then a new Astronomy folder with a note for each of the 65 targets. Claude read the file names and FITS headers and filled in each note with:

• What it is. A short description, its catalogue numbers and common name, and a type: galaxy, emission nebula, planetary nebula, cluster, comet, planet or star.
• Capture stats. Number of sub-frames, exposure, filter, gain, total integration time, number of nights, first and last dates, the biggest stack and how much space it takes on the NAS.
• A session table. One row per night, with the frames and integration for that night. Nights run noon to noon, so frames taken after midnight count towards the evening they started.
• Space for me. Empty fields for a rating, the processed image, where it ends up published, and my own notes.

Because the stats are in each note’s properties, an Obsidian Base turns the whole folder into a sortable table, which is where the fun numbers come from.

What it adds up to

• 65 targets: 20 galaxies, 16 emission nebulae, five planetary nebulae, five planets, seven stars, a few clusters, two comets, the Moon and the Sun (through a solar filter).
• About 24,600 sub-frames, almost all of them 10-second exposures.
• About 77 hours of integration across 82 nights, from the first session in March 2025.
• Nearly half the targets were a one-night stand. 31 of the 65 were only ever captured on a single night.

The clear favourite is the Whirlpool Galaxy, M 51: 17.4 hours over 21 nights, with the biggest single stack at 1,362 frames (nearly four hours). That one target is almost a quarter of everything I’ve captured. After that it drops off quickly:

• the Pacman Nebula (NGC 281), about 7.1 hours
• the Horsehead Nebula region (IC 434), about 6.2 hours
• Bode’s Galaxy (M 81), about 3.7 hours
• the Andromeda Galaxy (M 31), about 3.4 hours

The session dates tell their own story too. April 2025, the first full month, had 16 nights; June and July together have just one. In Scotland it barely gets dark in midsummer, and the catalogue shows it.

What the catalogue doesn’t do yet

This was a one-off run, so it’s already slightly out of date. The newest session in it is from August, and I’ve been out again this autumn: a comet, 161P/Hartley-IRAS, in October, and Andromeda again this week. Those stacks are on Flickr, but they aren’t in the notes yet.

It also only counts. It says nothing about whether the frames are any good, and some of them certainly won’t be.

What comes next

I’ve written the next steps up as a project in my vault. They are ideas at this stage rather than anything built:

• Quality scoring. Score every sub-frame for sharpness, background and satellite trails so the bad ones can be thrown out before stacking, and keep the per-target stats up to date as new sessions arrive.
• Proper stacking on the PC. Siril can be scripted, and can stack many nights of one target together. It would run on the Raspberry Pi, but the Pi is already busy running everything else, so the Windows PC, with a proper graphics card, is the better place for the heavy work.
• Publishing. Finished images are to go on vividgrey.co.uk. The idea is the same pattern as this blog: Claude creates a draft post per target through the WordPress API, and I review and publish it myself.
• The extras. A gallery, a Messier and Caldwell observing log, and something that notices new uploads and updates the notes.

The 17 hours of the Whirlpool Galaxy are the obvious first test. If 21 nights of data stack into something better than the best single night, the rest of the pipeline is worth building.

If you’ve been following my posts about what I’ve built with Claude Code, this is a slightly different use of it. There’s no service here and nothing running on a timer, just a big pile of data turned into something I can actually browse. Seventy-seven hours under the stars, and I didn’t know until this week.

This post was drafted by Claude Code from my vault notes on the Seestar and the 65 Astronomy target notes, and written into WordPress as a draft through the API. Claude chose the photo from my Flickr account. I reviewed and edited it before publishing.