Say Intentions / 2026

Why I'm building Flightpad

Hi there! I’m Jay, pleased to meet you. I’m the solo developer building Flightpad, and when something around Flightpad says “we” it mostly means just me. Flightpad is an ambitious project, but I’m lucky to be supported by a pretty great family and a set of talented friends. My mom, in particular, is one of the best QA people I’ve ever worked with and she’s been a great help.

Growing up I always loved flying and wanted to be a pilot. My mom, for many good reasons, worried about aviation safety, so I put flying out of my mind. My son is studying mechanical and aerospace engineering. He thought it would be helpful and fun to learn to fly. The timing was good, because GA has gotten much safer and planes even have ballistic parachutes now! Like a good son and dad, I sent him to talk to his grandma and get her blessing. I was expecting her to talk him out of it and offer other suggestions, but instead we both went and got our PPLs.

For me, flying is everything I dreamed it would be. I absolutely love it, yet I’m always aware of the risks. In 2026, in modern planes, the greatest risk is myself and my decisions.

I had a lot of time to learn through my PPL training. I got held up on my medical and took advantage of the delay to fly with a few different instructors and collect many stories and second-hand experience. Every instructor I’ve flown with has stories of near-misses due to missing or bad information. As a software engineer who has worked with very, very large datasets of very, very disorganized data, this really bothered me.

I finally got my medical, did my solo flights, and passed my checkride in November of 2025. Taking my wife up as my first passenger was a dream come true, and flying around the desert is incredible (especially in the winter).

January 2nd, 2026 in Telegraph Canyon

Telegraph Canyon is a little over 60 miles east of Phoenix, and one of the many places in the southwest desert that is filled with natural beauty. Four people were on a scenic helicopter flight when it struck a slackline suspended between two bluffs. Nobody in the helicopter made it. This tragedy will forever stay with me.

The group raising the slackline had filed a NOTAM and marked the line with LEDs and windsocks. They followed directions and best practices. The NOTAM wasn’t filed where a pilot departing from a different airport would see it, though. About an hour after the helicopter crashed, first responders reported another helicopter that narrowly missed the signal line, flying around 10 feet below it. Even with a NOTAM and markings, the line was that easy to miss.

When I first read about this, it brought back a familiar frustration: I rarely felt I’d briefed enough for local flights. I always try to be diligent and overprepare for any cross-country flight, but local flights often felt a bit too ad hoc. I could definitely see myself missing that type of NOTAM, and that was a scary thought.

Starting Flightpad

Pilots have a lot of information, sometimes too much, but nothing gives a good enough overview, so there are always gaps. I check aviationweather.gov and ForeFlight, look for TFRs and NOTAMs, and still feel I could be missing something (like the Vision Jet that struck construction equipment on a runway). The information just isn’t organized well. Through January I kept thinking about how much better it could be.

During my flight training, I developed a reasonable process that helped me feel better. Before each flight, I would start a new note on my phone from a simple template. At the top, each airport had a section with key facts like frequencies, TPA, and little stars if there were NOTAMs. I would copy down the METARs available to know what to expect, even though I’d get up-to-date weather from the ATIS.

So on February 10th, I decided to start building something. I reached out to a product designer I enjoy working with, and we began designing what the app could look and feel like. I started laying the groundwork for the data, learning about the underlying data products. It wasn’t long before I had the shape and a data provider, and like all good software engineers I thought I’d have the app ready for testing in a couple of weeks.

What Flightpad actually does

Flightpad isn’t an EFB, but an application to help organize information about frequently used airports. Flightpad highlights any adverse conditions in the area to help improve decision making, and can assist in flight planning. The most important feature is surfacing NOTAMs that really matter, because there are a few big gaps with the way NOTAMs are issued and commonly displayed.

One of the most significant problems is when NOTAMs are issued against an airport, they can be invisible to pilots using another nearby airport. It’s possible that someone departing Pegasus Airpark (5AZ3) wouldn’t see the NOTAM filed for the slackline, as it was filed against Superior Municipal Airport (E81) – the two airports are around 25nm apart.

Flightpad annotates each NOTAM through a decoding process that determines its location (and in the case of shapes, the boundaries of the NOTAM). From there, it’s a fairly easy and fast query to pull NOTAMs in geographical areas. Nothing in the decoding or annotation process uses AI/LLMs at all. Annotation is entirely deterministic and based on a very large set of rules.

The next gap is the needle in a haystack that happens when an airport has dozens of NOTAMs. I regularly fly out of KHND, which only has 5 NOTAMs as I write this, but 44 potentially relevant NOTAMs in the area. The annotation process also ranks NOTAMs based on notability, but it’s context-aware. What is notable for a single airport may be different when looking at a group of airports.

Together, these two features allow a pilot to see notable NOTAMs at a glance. The approach Flightpad takes also allows a degree of configuration, so the pilot can tune Flightpad to their environment. Maybe 50nm is too big or too small, or a 12-hour upcoming window is too short? Flightpad will let you configure the thresholds to get the view that makes the most sense for that list of airports.

The general shape is the same when we look at significant weather (SIGMETs and AIRMETs), which are quite a bit easier because those come with locations. NOTAM parsing is pretty messy, but that’s a different story.

This first version of Flightpad doesn’t have all the features I want, but I’ve learned the importance of working with and learning from users. I’m only one user, so I may not get it right. I want to work with all pilots and help them get more accurate, helpful information with fewer steps.

ForeFlight and Garmin Pilot are great, but until they are able to prioritize cumulative views of important information that is relevant to lists of airports, I feel very strongly that Flightpad needs to exist and be in the hands of pilots.

I really did believe I’d have a version of Flightpad out by April or May, and it’s October and I’m only now ready for external testers. Getting the NOTAMs right was hard. I have built dozens of tools to audit and validate my approach. I’m proud of what I’ve built and I believe it increases safety in aviation. It’s not perfect yet, but I would love your help getting closer.

When I officially launch Flightpad on the App Store, it will be a paid application, but all beta testers will get the final version for free. The core app will never be subscription-based.

A quick note on AI tools and software development

AI is a polarizing topic in some circles, and I have nuanced views. I loathe AI slop, but it is undeniable that the current AI tools are a fundamental shift in how software can be built and tested. To me, this feels like the moment when power tools became available. I imagine power saws and drills brought about similar increases in individual productivity.

I could build Flightpad without using any AI tools, but the fact is I wouldn’t. It’s been a lot of work using them, especially for just one person, but without them it would be too much. Being able to use these tools to organize work, generate code that eventually matches exactly how I would write it (and often better!) has made me profoundly more effective.

As mentioned above, there are no LLM/AI workflows in the application itself. Decoding, annotating, and determining notability are done entirely through deterministic processes. For now I do not trust any predictive, non-deterministic algorithm for this. It’s possible that decision models like Jev could match the performance of a deterministic system, but a deterministic algorithm feels better.

I don’t feel AI tools have made software much better in general, and every time I encounter Alexa for Shopping or half-baked AI support agents I want to turn off my computer for good. For individual developers, though, it’s such a massive help, and, at least for me, typing the code was never the fun part of the job. If you hate AI tools and don’t want to use Flightpad, I totally get it and respect your decision, but I think you’re missing out!

Further reading