Pilotgpt?

Their website homepage show 2 examples of use.
1. Giving approach speed in a vision jet. Not useful 0/10
2. Transcribing atc communication, 5/10, I find this intriguing, not $42/month intriguing, but intriguing nonetheless.
 
There needs to be some demo videos or something.
 
I couldn’t resist looking at it. Even went through some of the initial stuff and corresponded with the developer briefly. I didn’t see anything particularly useful and told them parroting a bunch of lost comm reg mnemonics was particularly useless in the case of real lost com (see below).

OTOH, color me cynical, but I suspect an app that has no trial, asks for $500 in advance, claims use by 1000 pilots, and has next to no information about what it does. I’d expect the website and the ads to be pointing to at least one video showing it being used in flight.

1772628226172.png
 
Like all new bleeding edge technologies, it is there to benefit the designers sense of purpose, and wallets. Oh, and the VC folks stand to get a nice cut of the action too.

Time will tell if it benefits the market, or turns into another brief blip on the RADAR.
 
The ads I saw suggested its use in conjunction with ForeFlight, couldn’t figure that out
 
Let'sgoflying! said:
The ads I saw suggested its use in conjunction with ForeFlight, couldn’t figure that out
Click to expand...
It's not about ForeFlight. It would do the same with Garmin Pilot or any other app what will operate in an iPadOS split screen. They are just using FF as an example because it's popular and familiar. That's all they're doing. I do it with other aps. I even have iOS shortcuts that open FF or GP with another app in exactly the same configuarion they are showing.
 
1772638979988.png
This is progress!!
1998 brought us the Children of the Magenta Line
2026 brings us Children of AI
Are we now going to let Claude fly instead of Otto?
 
Remember the dot com bubble? Well, I'd say this is just part of the AI bubble. Or just the overall concept that adding computers to things doesn't usually make them better. Just different.
 
An astrophysicist on YouTube made a rather long video in which she opined that "There is no AI. It's machine learning." She did state that it is a useful tool in her line of work, but that it should not be used to make decisions.
 
Albany Tom said:
Remember the dot com bubble? Well, I'd say this is just part of the AI bubble. Or just the overall concept that adding computers to things doesn't usually make them better. Just different.
Click to expand...

dot com bubble (which some call the telecom bubble) was a realignment of the stock market's perception with ground reality. There was no change in the growth of the technology - it has actually continued to grow exponentially.

AI stocks may collapse, but that doesn't mean the adoption of AI will stop. It's here to stay. A single pilot cockpit is coming, whether we like it or not, with the first officer replaced by an AI.
 
I will not be riding on those airplanes.
 
sarangan said:
dot com bubble (which some call the telecom bubble) was a realignment of the stock market's perception with ground reality. There was no change in the growth of the technology - it has actually continued to grow exponentially.

AI stocks may collapse, but that doesn't mean the adoption of AI will stop. It's here to stay.
Click to expand...
“AI” is just the marketing term. Which is really “ML”, or brute force statistics.

The value in “AI” isn’t the generative stuff, that’s just how it’s sold to the public. The value is in analyzing the data they’ve collected on you so they can decrease expenses and increasing their sales. Every purchase you make with a “loyalty” card is used to build a picture of who you are, what you like, what you don’t like, what you’re doing, and when you're going to come back.

sarangan said:
A single pilot cockpit is coming, whether we like it or not, with the first officer replaced by an AI.
Click to expand...
From a technology background, I think we’ll skip single-pilot cockpits and go straight to zero-pilot cockpits. Planes will still be monitored by someone, be it a pilot or some other newly created career, but it will be behind a screen or in a simulator environment where more than 1 plane is managed.

My guess is something similar to an ATC style system, mixed with harbor pilot.

You have a “Ground” team that handles navigating from gate to runway. The plane is then passed off to a “Local” team that handles takeoff and initial departure procedures. An “En Route” team watches over the plane until it's time to pass it back to the destination's “Local” team who handles the arrival and landing. Then it’s returned to the “Ground” team to handle the runway to gate.

None of this will be done with the current planes, outside of retrofitting a few in order for live testing. And none of this will be announced until the first manufacturer has their new plane ready. At which point all other manufacturers will quickly follow suit and make their own announcements. The lobbying of governments will occur, and the public will be sold on zero-cockpits planes being the safer future of aviation.

It’s probably not going to happen tomorrow. But not because of any technological limitations. And when it does, it will happen a lot quicker than you will expect.
 
sarangan said:
dot com bubble (which some call the telecom bubble) was a realignment of the stock market's perception with ground reality. There was no change in the growth of the technology - it has actually continued to grow exponentially.

AI stocks may collapse, but that doesn't mean the adoption of AI will stop. It's here to stay. A single pilot cockpit is coming, whether we like it or not, with the first officer replaced by an AI.
Click to expand...
Oh, I don't think AI is going away, any more than websites are going away. But as you point out, 20+ years ago some people thought the growth of web stuff was infinite, and it wasn't; and today people think that AI is going to change the world like the invention of the computer itself, and to me, not so much.

As to AI and aircraft, I don't think it's about single pilot cockpit. I read it as some folks thinking that now your pilot doesn't really have to know how to different modes of any particular airplane work, because the AI thing will tell them. From talking to some folks over the years, we have had periods where some US airline pilots really weren't great at hand flying their aircraft, and we have had some groups in the US that don't necessarily think that type rating is a thing. I can only imagine what lower budget, lower safety countries are planning to do.
 
I will say, Claude Code is a revolution in software development. It’s pretty amazing.

AI will be as big (or bigger) of a revolution than “the internet” was.
 
dans2992 said:
I will say, Claude Code is a revolution in software development. It’s pretty amazing.

AI will be as big (or bigger) of a revolution than “the internet” was.
Click to expand...
I can’t disagree. At its most basic, AI is a super fast, super efficient researcher/analyst. Even when public ChatGPT is wrong, it’s because it’s pulling in data from sources that are not necessarily reliable. Basic GIGO. Plus, the quality of results also reflects the quality of the prompts (the questions and follow-up questions we ask).

Take those abilities - a super researcher and a good questioner. Now place them in a professional environment where the source data is reliable and the human prompter is trained…

In that sense Ai may well be evolutionary rather than revolutionary.
 
"where the source data is reliable"

there's always a catch...
 
Bob Noel said:
"where the source data is reliable"

there's always a catch...
Click to expand...
I'm not sure how much of a "catch" it is. After all, even outside the AI sphere, I expect there to be a difference in the quality of the source for, say, medical information on a condition, between a physician who specializes in that condition and SGOTI.
 
midlifeflyer said:
I'm not sure how much of a "catch" it is. After all, even outside the AI sphere, I expect there to be a difference in the quality of the source for, say, medical information on a condition, between a physician who specializes in that condition and SGOTI.
Click to expand...

not questioning that. however, there have been multiple times in history when the experts have been wrong.
 
“Hey, pilotGPT, I just reached MDA, and the approach chart says there are obstacles in the visual segment. Should I continue to follow the advisory VNAV?”
 
MauleSkinner said:
“Hey, pilotGPT, I just reached MDA, and the approach chart says there are obstacles in the visual segment. Should I continue to follow the advisory VNAV?”
Click to expand...
If only they had a free trial where you can ask that question. I'd love to see their answer. Here's Claude's. I'm including my phrasing of the question and Claude's answer. What do you think?

Me:​
Hypothetical: A pilot is flying a "2D" instrument approach - no official vertical guidance. But the GPS navigator produces a "+V" glidepath. The pilot reached MDA and the approach chart says there are obstacles in the visual segment. Should the continue to follow the advisory VNAV? Why or why not?​
​
Claude:​
This is a great question that gets at a real safety issue in modern avionics. The short answer is: no, the pilot should not continue following the advisory VNAV below MDA — and here's why.​
​
What "+V" Advisory VNAV Actually Is​
The "+V" (or "LNAV+V") glidepath is not an approved vertical guidance path — it's a convenience feature generated by the GPS navigator using barometric altitude and a computed descent angle. It has no regulatory status and is not validated against obstacle clearance below MDA.​
​
Why This Is Dangerous on an Approach With Obstacles in the Visual Segment​
When an approach chart notes obstacles in the visual segment (below MDA), it means the designer could not guarantee obstacle clearance in that area using normal circling/straight-in criteria. The pilot is expected to:​
1. Acquire the runway environment visually before descending below MDA​
2. Maneuver visually to avoid those obstacles — using their eyes, not instruments​
The advisory VNAV knows none of this. It simply projects a constant-angle path to the runway threshold. That computed glidepath may fly directly through a charted obstacle in the visual segment.​
​
The Regulatory and Procedural Framework​
• MDA is a floor, not a target. On a non-precision approach, the pilot must level off at MDA unless they have the required visual references (14 CFR 91.175).​
• Advisory VNAV does not change the approach type. LNAV+V is still an LNAV approach. The minimums, missed approach point, and obstacle clearance guarantees are all LNAV-based — which provide no obstacle clearance below MDA.​
• The visual segment belongs to the pilot, not the avionics. Below MDA, obstacle avoidance is the pilot's responsibility using visual means.​
​
What the Pilot Should Do​
• Fly to MDA and level off​
• If visual references are acquired and obstacles can be visually avoided → continue visually, disconnect or ignore the +V guidance​
• If visual references are not acquired → execute the missed approach​
​
The Bottom Line​
The "+V" glidepath is a useful tool for establishing a stabilized descent above MDA — it reduces the "dive and drive" problem that causes CFIT accidents on non-precision approaches. But it was never designed to guarantee obstacle clearance below MDA, and the avionics have no awareness of charted visual segment obstacles. Trusting it below MDA on such an approach means delegating an obstacle-avoidance responsibility to a system that is blind to those obstacles.​
​
This is exactly the scenario the FAA and Jeppesen warn about when they note visual segment obstacles on approach charts — it's a flag to the pilot that eyes, not algorithms, must take over at MDA.​

​

 
Last edited:
AV8R_87 said:
You'll hit the obstacle by the time Claude stops rambling.
Click to expand...
:D Remember that I also asked for an explanation. If you don't want understanding, you could stop at

The short answer is: no, the pilot should not continue following the advisory VNAV below MDA​
 
midlifeflyer said:
I can’t disagree. At its most basic, AI is a super fast, super efficient researcher/analyst. Even when public ChatGPT is wrong, it’s because it’s pulling in data from sources that are not necessarily reliable. Basic GIGO. Plus, the quality of results also reflects the quality of the prompts (the questions and follow-up questions we ask).

Take those abilities - a super researcher and a good questioner. Now place them in a professional environment where the source data is reliable and the human prompter is trained…

In that sense Ai may well be evolutionary rather than revolutionary.
Click to expand...
When the 486 was new, an Intel or Microsoft rep sent me a poster of one the brand new x86 chips, basically a picture of it. The poster title was "Evolutionary". And I found that hilarious, because it was absolutely true there was nothing really new and outstanding, but the marketing people liked words that sounded great.

In terms of "data from sources that are not necessarily reliable", I completely agree, and I don't think I'm the only one that's predicting that as AI starts listening to AI more, the entropy levels are going to go way up.
 
The non-determinism of LLMs and ML in general makes this an ill-considered application. Aviation needs precision at all times, not just statistically "most of the time"
 
schmookeeg said:
The non-determinism of LLMs and ML in general makes this an ill-considered application. Aviation needs precision at all times, not just statistically "most of the time"
Click to expand...

non-determinism is tough for the safety-critical domain
 
I wonder if the answer would be the same before the Steamboat crash and the lengthy threads here and on other forums?

midlifeflyer said:
If only they had a free trial where you can ask that question. I'd love to see their answer. Here's Claude's. I'm including my phrasing of the question and Claude's answer. What do you think?

Me:​
Hypothetical: A pilot is flying a "2D" instrument approach - no official vertical guidance. But the GPS navigator produces a "+V" glidepath. The pilot reached MDA and the approach chart says there are obstacles in the visual segment. Should the continue to follow the advisory VNAV? Why or why not?​
​
Claude:​
This is a great question that gets at a real safety issue in modern avionics. The short answer is: no, the pilot should not continue following the advisory VNAV below MDA — and here's why.​
​
What "+V" Advisory VNAV Actually Is​
The "+V" (or "LNAV+V") glidepath is not an approved vertical guidance path — it's a convenience feature generated by the GPS navigator using barometric altitude and a computed descent angle. It has no regulatory status and is not validated against obstacle clearance below MDA.​
​
Why This Is Dangerous on an Approach With Obstacles in the Visual Segment​
When an approach chart notes obstacles in the visual segment (below MDA), it means the designer could not guarantee obstacle clearance in that area using normal circling/straight-in criteria. The pilot is expected to:​
1. Acquire the runway environment visually before descending below MDA​
2. Maneuver visually to avoid those obstacles — using their eyes, not instruments​
The advisory VNAV knows none of this. It simply projects a constant-angle path to the runway threshold. That computed glidepath may fly directly through a charted obstacle in the visual segment.​
​
The Regulatory and Procedural Framework​
• MDA is a floor, not a target. On a non-precision approach, the pilot must level off at MDA unless they have the required visual references (14 CFR 91.175).​
• Advisory VNAV does not change the approach type. LNAV+V is still an LNAV approach. The minimums, missed approach point, and obstacle clearance guarantees are all LNAV-based — which provide no obstacle clearance below MDA.​
• The visual segment belongs to the pilot, not the avionics. Below MDA, obstacle avoidance is the pilot's responsibility using visual means.​
​
What the Pilot Should Do​
• Fly to MDA and level off​
• If visual references are acquired and obstacles can be visually avoided → continue visually, disconnect or ignore the +V guidance​
• If visual references are not acquired → execute the missed approach​
​
The Bottom Line​
The "+V" glidepath is a useful tool for establishing a stabilized descent above MDA — it reduces the "dive and drive" problem that causes CFIT accidents on non-precision approaches. But it was never designed to guarantee obstacle clearance below MDA, and the avionics have no awareness of charted visual segment obstacles. Trusting it below MDA on such an approach means delegating an obstacle-avoidance responsibility to a system that is blind to those obstacles.​
​
This is exactly the scenario the FAA and Jeppesen warn about when they note visual segment obstacles on approach charts — it's a flag to the pilot that eyes, not algorithms, must take over at MDA.​

​

Click to expand...
 
Hang 4 said:
I wonder if the answer would be the same before the Steamboat crash and the lengthy threads here and on other forums?
Click to expand...
I guess we don't know, but the sources for that answer have been around as long as +V. There's nothing special there.
 
midlifeflyer said:
I guess we don't know, but the sources for that answer have been around as long as +V. There's nothing special there.
Click to expand...
I was semi-sarcastic. But reading the text of the response sounds like a lot of the posts on here and BT. I guess the interesting question is where do the AI's pull from? I haven't read the Instrument Flying Handbook lately, but the line in Claude's response "This is a great question that gets at a real safety issue in modern avionics." Seems like it could have influenced by all the chatter post-Steamboat.
Totally agree that there is nothing new factually, but based on the threads, it seems like a lot of pilots did not appreciate the risks of following +v on some approaches.
 
dans2992 said:
I will say, Claude Code is a revolution in software development. It’s pretty amazing.

AI will be as big (or bigger) of a revolution than “the internet” was.
Click to expand...
It is definitely a very useful tool and I am getting a lot more done , mostly around stuff that I would have to research or general boilerplate code ….but it never lets me forget that this is not a true intelligence ….you can never let your guard down as it will happily generate nonsense like SQL queries joining tables from MS SQL and PSQl and , when asked about it, profusely apologize and then a few hours later do something similarly stupid - it is literally like working with Dr Jekyll and Mr Hyde …. I can rely on human programmers working for me to produce certain stable level of quality code but with Claude Code is like oscillating between a mini genius and an occasional complete retard …

Here is a sample :-)
 

Attachments

  • IMG_8304.jpeg
    IMG_8304.jpeg
    28.6 KB · Views: 16
Last edited:
You must log in or register to reply here.
Back
Top