Don't underestimate what Claude can do with your NINA Advanced Sequencer JSON

Georg N. Nyman · Mar 26, 2026, 02:28 PM

Oh well, I think I am not up to date :-) I am using NINA for about three years now but I have no clue what “Claude” is…and not heard this name before !

Did I miss something important which I should learn now - finally ?

It seemed most appropriate to have AI answer your question. So:

Claude is a highly capable AI assistant from Anthropic designed for complex, high-volume tasks. It excels at summarizing long documents, coding in multiple languages, and creative writing, with advanced "computer use" features allowing it to autonomously control a browser/desktop to perform tasks like data entry, file management, and software navigation.

Key capabilities of Claude include:

• Advanced Writing & Content Generation: Drafts emails, reports, screenplays, and novels, with the ability to adjust tone and style.

• Deep Analysis & Summarization: Reads, summarizes, and analyzes extensive documents, PDFs, and data, turning them into structured formats like spreadsheets.

• Coding & Technical Tasks: Writes, refines, and debugs code across major languages, particularly Python, with capability to manage whole codebases.

• Computer Use (Beta): Controls a computer's mouse and keyboard to navigate apps, fill forms, and operate websites, effectively acting as an autonomous agent.

• File Operations & Research: Reads/edits local files, conducts web searches, and integrates with professional tools to automate workflows.

Versions of Claude:

• Claude.ai: The browser-based interface for chat, analysis, and file uploads.

• Claude Code: A specialized terminal-based tool for software development.

Claude is designed to be safe and accurate, focusing on providing actionable assistance in both creative and analytical contexts.

Georg N. Nyman · Mar 26, 2026, 02:28 PM

Oh well, I think I am not up to date :-) I am using NINA for about three years now but I have no clue what “Claude” is…and not heard this name before !

Did I miss something important which I should learn now - finally ?

📷 image.pngYes I know, I know, like Cuiv… getting lazy :-)

mzaslove · Mar 26, 2026, 04:00 PM

It’s a tool, and it has its uses, BUT…I’m running about 85% error (wrong answers) with various AI programs. They’ve also made it plain they prioritize giving a reasonable-sounding answer over a correct answer (which is why it tends to be 85% wrong).

The other day I couldn’t remember the name of a character in a TV series I WROTE (minor character, didn’t want to look it up). It made up the name of a character and guaranteed me it was correct. I pointed out it wasn’t. It made up two more names after that. It pointed out that giving a reasonable-sounding answer was its first choice on things. It finally got the name. (And that’s how I use A.I. a lot: have it give me a bunch of wrong answers, and fish out the right one.)

Worse, from an astronomy standpoint, I was asking it to write a Voyager Dragscript on something. It assured me it could. It did write one…but it not only didn’t work, it wouldn’t run at all. I corrected it. I asked it if it were capable of doing what I asked. It assured me it was. More “Dragscripts” that couldn’t work at all. It finally figured out that it can’t write Dragscripts in any way, shape, or form because it’s a proprietary format. Couldn’t it have told me that in the beginning and save me all that time? Nope, because it’s prioritized to give reasonable-sounding answers over correct ones.

So, if you can check its work, it has its uses (and I use it), but even if it writes a sequence for you, is it right? Is it rightest? Is there something wrong in it you don’t know about?

Word to the wise.

I had the same experience with scripts for Siril. Could not get one that would run without errors.

SonnyE · Mar 25, 2026, 04:35 PM

Learning curves?

Oh, like buying a go-to mount and letting it help you learn the Universe as it takes you from star to star?

The fact that we are all on Astrobin shows we are using computers instead of the Pony Express.

Me, Myself, and I welcome all the electronic help I find useful. There wasn’t much help from egotists. I had to dig in and figure out what various AI (Computers and Applications) were trying to show me. And troubleshoot my software when it got corrupted. Still do.

EAA was and still is to me: Electronically Assisted Astronomy. Go-To mounts, Guiding equipment, Electronic Cameras, and Computers, etc.

Personally, over the decades I’ve used Stellarium to locate starting with browsing the program, and now as my object choosing App to input into NINA’s Framing Wizard to setup where my mount slews to.

It is working smarter, not harder.

We should all go back to star charts, setting circles, clock drives, hypering our own film, manual guiding, and bring back the dark room with all the chemicals, trays, paper, and Printmakers. Better yet, glass plates!

Chad Andrist · Mar 27, 2026, 01:16 AM

SonnyE · Mar 25, 2026, 04:35 PM

Learning curves?

Oh, like buying a go-to mount and letting it help you learn the Universe as it takes you from star to star?

The fact that we are all on Astrobin shows we are using computers instead of the Pony Express.

Me, Myself, and I welcome all the electronic help I find useful. There wasn’t much help from egotists. I had to dig in and figure out what various AI (Computers and Applications) were trying to show me. And troubleshoot my software when it got corrupted. Still do.

EAA was and still is to me: Electronically Assisted Astronomy. Go-To mounts, Guiding equipment, Electronic Cameras, and Computers, etc.

Personally, over the decades I’ve used Stellarium to locate starting with browsing the program, and now as my object choosing App to input into NINA’s Framing Wizard to setup where my mount slews to.

It is working smarter, not harder.

We should all go back to star charts, setting circles, clock drives, hypering our own film, manual guiding, and bring back the dark room with all the chemicals, trays, paper, and Printmakers. Better yet, glass plates!

I started that way. I wouldn’t want to go back to that technology for astrophotography but I still regularly use film for regular photography. I love the skills involved and it gives results that I like better then digital.

IMHO, learning to do things yourself is part of the hobby. I am also a woodworker and sometimes doing things manually is more rewarding and actually requires skills.

Make things too easy and they become devalued.

In my experience these tools work well for programming where there is a large body of examples to learn from. But even AI tools make mistakes, but if the environment is set up well they can correct most of these mistakes. You still need to guide those tools with your knowledge and experience. If the body of examples is small, like for instance with PI scripts, your chances of success diminish rapidly.

Wei-Hao Wang · Apr 22, 2026, 06:07 AM

Let me ask this: how far away are we from totally abandoning NINA, ASIAir, and alike?

NINA is to automate your imaging so you can just push a button and go to sleep. It does what you tell it to do. Now the title of this discussion thread suggests that you can use AI to tell what NINA to do. Why do this extra step? Can you directly instruct AI to take over the telescope operation and skip NINA entirely? AI may sense that the souther part of sky has thin cirrus and then switch to northern targets automatically while you are sleeping. NINA cannot do that, can it?

Or, maybe NINA can still exist, but the user does not interact with it at all. You don’t need to know anything about script or advanced sequence. You only interact with an AI agent, and the AI agent talks to NINA.

Are we there yet?

If we are not there, we are pretty close to it.

Just recently, I asked Gemini to write VBA code to solve a pair of ODEs using an explicit Runge Kutta method. It not only wrote the code to do it, it correctly identified that the ODEs were representing an electrical system without my telling it. At my prompting, it told me that the ODEs could be solved analytically, and gave the general and particular solutions by finding the eigenvalues and eigenvectors of the system. Compared to this, what you are talking about is easily doable by AI. Someone just hasn’t gotten to it yet. This stuff is scarily good and getting better every day.

Wei-Hao Wang · Apr 22, 2026, 04:30 PM

Arun H · Apr 22, 2026 at 03:16 PM

If we are not there, we are pretty close to it.

Just recently, I asked Gemini to write VBA code to solve a pair of ODEs using an explicit Runge Kutta method. It not only wrote the code to do it, it correctly identified that the ODEs were representing an electrical system without my telling it. At my prompting, it told me that the ODEs could be solved analytically, and gave the general and particular solutions by finding the eigenvalues and eigenvectors of the system. Compared to this, what you are talking about is easily doable by AI. Someone just hasn’t gotten to it yet. This stuff is scarily good and getting better every day.

I also know how good it is and I believe the current generation AI is fully capable of being a good telescope operator. The main concern is, unlike asking AI to solve science/math problems, asking AI to control your telescope requires you giving a good range of permissions to AI to issue commands to your computer (and then to the telescope) as well as to see a good range of information in your computer or from your network in order to make judgements. This can be a security issue.

Anyway, the point is, it looks a bit silly to me if we are to ask AI to tell NINA what to do in order for NINA to tell the telescope what to do. Just tell AI to operate the telescope for you, without NINA. The middle layer (NINA) may disappear rather quickly.

I really think this is a much better use of AI in astrophotography as apposed to trying to have it do your processing for you as some on the bin have suggested.

Wei-Hao Wang · Apr 22, 2026, 04:30 PM

The main concern is, unlike asking AI to solve science/math problems, asking AI to control your telescope requires you giving a good range of permissions to AI to issue commands to your computer (and then to the telescope) as well as to see a good range of information in your computer

These problems have largely been solved. For example, Gemini (and I assume Claude) when used in an enterprise environment can have walls built around it so proprietary data is not used to train models. Similarly, I cannot ask my AI to give me access to someone’s personnel file or performance review data (while someone in HR probably could). I would think the same principles could be used to wall off sections of your network.

Wei-Hao Wang · Apr 22, 2026, 11:36 PM

Tony Gondola · Apr 22, 2026 at 05:05 PM

I really think this is a much better use of AI in astrophotography as apposed to trying to have it do your processing for you as some on the bin have suggested.

I think I can agree with you. Processing is something really personal. Different people can create different pictures from an identical set of raw data. The result of processing can carry the signature of an astrophotographer. So despite I am a big supporter of AI, I will never hand the processing part to AI. If I let AI do the processing for me, the result would be someone else’s picture, not mine. Even if AI can learn my processing style perfectly, it’s just an excellent student of mine. Its result is still not my result.

That being said, I wouldn’t mind letting AI do the routine and standard part of processing, like calibration, registration, and stacking. Beyond that, I will not allow AI to take away the fun from me.

Jerry Gerber · Apr 22, 2026 at 05:45 PM

A lot of what gives people enjoyment/purpose/meaning is in the making of decisions. I’m currently writing my 14th symphony. Writing such a piece of music requires thousands of decisions, in fact the whole point of doing such an endeavor is the pleasure, challenge and meaning of making the countless decisions that go into the construction of a piece of music. Handing over decision-making to AI seems absolutely pointless, I might as well ask AI to live my life for me, eat for me and hold conversations for me.

I see the value of AI in many fields, no doubt , particularly medicine and materials science. I just watched a video of a ping pong expert playing a new machine made by Sony called “Ace”, and it proved to be a very worthy opponent. Imagine the AI, coding, sensors, robotics and knowledge that went into building a robot ping pong player that could react as quickly as a human. Pretty impressive.

Maybe I’ll get around to asking AI to pick a target for me when I am done working on my current image. That’s about it. I enjoy all the other steps of setting up the NINA sequence, determining filters and filter order and working around moonlight, and much enjoy the processing and post-processing aspects.

I get your point and I am sort of like you. But I think AI is not (or soon to be not) as simple as that.

In principle, you can still be the one who makes decisions. It’s just that you teach AI how you make decisions, and AI do it on your behalf when the situation arises. It’s not like telling AI “take a picture of M42 for me.” It’s more like telling it how you would “determining filters and filter order and working around moonlight” and ask it to follow your style, so the end result will be identical to what you would have achieved by yourself, without the human error part. I don’t know about you, but I welcome that.

Arun H · Apr 22, 2026 at 07:53 PM

These problems have largely been solved. For example, Gemini (and I assume Claude) when used in an enterprise environment can have walls built around it so proprietary data is not used to train models. Similarly, I cannot ask my AI to give me access to someone’s personnel file or performance review data (while someone in HR probably could). I would think the same principles could be used to wall off sections of your network.

I do think that’s solvable. It’s just when a good implementation will actually appear for astrophotography. I hope it will be soon.

I get it, I’ll probably use AI in the near future, perhaps start with target selection. My biggest concerns about AI are not about me or my use of it, it’s more about how its benefits will be mostly going to the wealthiest companies and people (at least here in the US) and how it can be used for nefarious purposes. Other than that, I am sure a lot of good can come of it. I worry more about NS than I do about AI (NS=natural stupidity).