Guest blog ElbaSatGuy - AI & LLM

thebaldgeek here. I've had in mind to do some podcasts and YouTube's videos on ACARS and once getting the format a bit settled, I've wanted to open it up and do some 'interview' style media and have guests on the 'show' to talk all things ACARS. A bit like the old X (Twitter) Spaces 'Fireside chats' (Or as was often the case, fireside rants), but slightly calmer and more structured. But the end goal was the same. I want to connect users and osint folks that use ACARS to really dig into and take flight tracking to the next level via ACARS. Since this blog has been running for a few months, the idea occurred to me to open it up and feature some guest blogs from time to time. And so here we are. Mike, aka @ElbaSatGuy and I have been DM'ing for what must be years (cant scroll back that far to find the first one!). Mike drives a big rig / truck / semi (in Australian) / 18 wheeler. He affectionately calls it his 'mobile sigint platform'. He operates multiple ACARS modes from the truck, ranging from Iridium to VDL, and even some HFDL. Unlike me, Mike is a real programmer! With decades of software engineering, including start-up dot coms , helping Fortune 100s to mom and pop solutions. Mike has focused his rest breaks on the subject of using LLM's and AI to tap into the firehose of ACARS. You can find Mike on X: https://x.com/ElbaSatGuy and BlueSky: https://bsky.app/profile/elbasatguy.bsky.social With that as an intro; Guest blog #1. Take it away Mike.... If you’ve ever looked up and seen a contrail slicing through the sky, chances are that aircraft is sending messages to the ground — automatically. These messages, often transmitted over radio or satellite, are part of a system called ACARS (Aircraft Communications Addressing and Reporting System). Think of it like a text messaging system for airplanes. But unlike your group chat, ACARS messages aren’t always easy to read or understand. That’s where artificial intelligence (AI) steps in — and why this project is all about using modern AI tools to make sense of these cryptic airborne messages. What’s in an ACARS Message? ACARS messages are sent between airplanes and ground stations, carrying all sorts of information — positions, weather updates, system diagnostics, and sometimes even requests for gate changes or maintenance. The catch? These messages can be wildly inconsistent. Some are neatly structured; others are like shorthand notes scribbled in turbulence. They’re designed for efficiency, not clarity — at least not to humans outside the aviation industry. Here’s an example of what an ACARS message might look like: Q0 D-ABCD 1234Z FL370 POS N5012.3 E00321.5 Translation? Something like: aircraft D-ABCD, at 12:34 UTC, flying at 37,000 feet near a certain location. But many messages are even more jumbled — sometimes missing punctuation, mixing formats, or containing abbreviations that vary between airlines. The Problem: Too Much Data, Not Enough Structure With thousands of aircraft sending ACARS messages 24/7, the result is a huge pile of unstructured data. If you wanted to answer a simple question like “Where was this plane when it sent this message?”, you’d have to wade through all that inconsistency. Traditional software methods — like regular expressions or keyword searches — don’t cut it. It's the classic "needle in a haystack" problem, made worse because the needle might not look like a needle. The Solution: Teaching AI to Understand ACARS The idea is to take advantage of recent advances in AI language models — the same kinds of models behind ChatGPT — to process and interpret these messages. Here’s how the approach works, in simple terms: Collect the Messages: A local system stores a large number of ACARS messages in a searchable database. Teach AI What to Look For: Instead of trying to write exact rules, the system uses semantic embeddings — a way of converting messages into numbers that represent their meaning. Look for Similarity: By comparing these embeddings, the system can find messages that are like a known example, even if the format is different. Extract Key Info: From there, AI can identify patterns — for instance, extracting positions like latitude and longitude, or recognizing messages that mention altitude, weather, or specific aircraft behaviors. What Makes This Cool? AI brings something new to the table — the ability to make intuitive leaps in messy data, which human analysts or rule-based software would struggle with. For example: You give the AI a known position-reporting message. The AI finds 200 others that look nothing like it but carry a similar message intent. Now you can map the results, cluster them, or track patterns over time, using visualizations to make things clear. What Could This Be Used For? This kind of AI-enhanced ACARS analysis could enable: Flight behavior tracking: Identify how specific aircraft operate — where, when, and under what conditions. High-level insights: See which airlines or aircraft types send the most maintenance messages, or which routes have the most turbulence reports. Anomaly detection: Spot weird or unusual messages that might indicate spoofing, miscommunication, or system issues. From the Ground Up: Tools and Workflow The project runs locally — no cloud required — using: A lightweight SQLite database to store messages and analysis. A software embedding model that converts ACARS messages into AI-friendly vectors. A visual interface for exploring, comparing, and searching the messages interactively. The entire system is designed to be fast, flexible, and friendly to future extensions — like adding maps or timelines of message activity. Final Thoughts: Why It Matters In a world of satellites and GPS, much of aviation communication still happens through old-fashioned radio. But there’s nothing old-fashioned about using AI to decode and understand those signals at scale. By combining aviation data with modern AI tools, we can uncover hidden patterns that yield unexpected insight. ========================================================================== tbg back.. Wow. Mike, thanks a ton for that AI / LLM vs ACARS intro. Exciting times ahead. Mike and I are working together to host some of these apps that the screenshots came from. Keep in mind, this is all about stoking the ACARS dumpster fire. There will be glitches and rough edges, but that's the fun of it. We will definitely have Mike return and expand on some of the sections in this blog in one or more future posts, so keep an eye out for that. If you want to take a shot at writing an ACARS-centric blog, DMs are open, drop your email in the website chat, or just find me online, and I look forward to sharing this space with others as the flight progresses!