Hey everyone, welcome back to My Weird Prompts. I am Corn, and I have to say, the energy in the house this morning was a bit different. Our housemate Daniel was scrolling through the news over coffee and he practically shoved his phone into my face. He found this exclusive article in the Jerusalem Post that just came out today, February third, twenty twenty-six.
Herman Poppleberry here, and yeah, I saw that one too. It is titled From Gaza to Iran, Israel Readies Space Surprises for Next Conflicts. It is an interview with Avi Berger, who is the head of the Space Office at the Ministry of Defense’s research and development directorate, or MAFAT. And man, the timing of this is incredible.
It really is. We are sitting here in Jerusalem, and the geopolitical reality is just… it is a lot. Daniel’s prompt was specifically asking about the technical side of what Berger was hinting at. Specifically, how on earth do you move that much data from a satellite down to the ground, and what does AI on steroids actually look like in a military intelligence context?
Those are the right questions to ask because, as we have discussed in past episodes, collecting the data is actually the easy part these days. The hard part is the plumbing and the processing. Berger mentioned that during the twelve days of Operation Rising Lion, the Israeli satellite constellation collected tens of millions of square kilometers of imagery. That is an astronomical amount of raw data.
Right, and for our regular listeners, you might remember episode two hundred eight where we talked about the new era of satellite intelligence. Back then, we were looking at the shift toward persistent surveillance. But what Berger is describing now feels like a whole different level of intensity. He mentioned that whatever was deployed even as recently as last June is already considered insufficient for the next conflict.
It is the ultimate arms race, but it is happening five hundred kilometers above our heads. And when Daniel asked about how that information is transferred, he is hitting on the biggest bottleneck in space operations. Most people imagine a satellite as this constant stream of data, like a live webcam. But these are Low Earth Orbit satellites, or LEO. They are moving at seven kilometers per second. They only pass over a specific ground station for maybe five to ten minutes at a time.
So you have this tiny window to dump terabytes of data. How are they doing it? Is it still just traditional radio waves?
Traditionally, yes. Most military satellites use X-band or Ka-band radio frequencies. But the surprise Berger might be hinting at—and something we are seeing across the industry in twenty twenty-six—is the transition to optical or laser communications. Instead of a broad radio beam, you are firing a highly focused laser from the satellite to a ground receiver.
Which increases the bandwidth by what, a factor of ten? A hundred?
Easily a hundred times the throughput of traditional radio. And it is much harder to jam or intercept because the beam is so narrow. If you are not standing exactly where that laser is hitting, you are not getting the data. But it is technically a nightmare because you have to point a laser from a platform moving at seventeen thousand miles per hour and hit a target on the ground with millimeter precision.
That sounds like trying to hit a moving penny with a needle while you are riding a roller coaster.
Exactly. And when you consider that Israel is dealing with what the defense establishment calls seven active fronts right now—from Gaza to the Houthis in Yemen—the need for real-time, high-bandwidth data is constant. You cannot wait for the satellite to finish its orbit and pass over a ground station in central Israel every ninety minutes if you need to track a mobile missile launcher in the Iranian desert right now.
So that brings up an interesting point. If you have a bottleneck in getting the data down to Earth, do you just start processing it in space? Is that part of the AI on steroids that Berger mentioned?
You nailed it, Corn. That is exactly where the intelligence community is heading. We call it edge computing in space. Instead of sending every single pixel of a ten-million-square-kilometer image down to a ground station in Yehud, the satellite has onboard AI chips. These chips are trained to recognize specific anomalies.
So the satellite is essentially saying, I am looking at a thousand miles of sand, nothing interesting here, nothing here… oh, wait, that is a new tire track that wasn't there ninety minutes ago. I will only send the data for those fifty square meters.
Precisely. It is the difference between sending a high-definition movie and sending a text message that says, hey, something moved at these coordinates. This is the anomaly detection Daniel was asking about. In the article, Berger talks about the IDF’s Unit ninety-nine hundred, which is their elite geospatial intelligence unit. They are the ones who actually take this raw data and turn it into something a commander can use.
I have always been fascinated by Unit ninety-nine hundred. They have that famous program, Roim Rachok, which translates to Looking Ahead, where they recruit soldiers on the autism spectrum specifically for their ability to spot minute changes in satellite imagery that a neurotypical brain might just filter out as noise.
It is a brilliant use of human cognitive diversity. But what Berger is saying is that even those incredible human analysts are being overwhelmed by the sheer volume. When you have the Ofek-nineteen satellite, which launched last September, pumping out high-resolution Synthetic Aperture Radar imagery twenty-four seven, you need AI to act as a first-tier filter.
Wait, remind me and the listeners why Synthetic Aperture Radar, or SAR, is such a big deal compared to regular optical cameras. We touched on this in episode seventy when we talked about AI for crisis management.
Right. An optical camera is just a giant telescope in space. If there is a cloud in the way, or if it is nighttime, it sees nothing. SAR is different. It is an active sensor. The satellite sends out its own microwave pulses and measures how they bounce off the ground. Those microwaves pass right through clouds, smoke, and dust. They work perfectly in total darkness.
So you can’t hide under a sandstorm or wait for a cloudy day to move your equipment.
Not anymore. And because SAR measures the physical texture of the ground, the AI can detect things like the soil being disturbed for a buried IED or the slight metallic signature of a camouflaged vehicle. When Berger says AI on steroids, he is talking about neural networks that can fuse SAR data with optical imagery and even signal intelligence in real time.
It is like having a digital ghost that knows what the world is supposed to look like and screams the second a single atom is out of place. But Herman, let’s push back a little on the implications here. If we are relying on AI to tell us what is an anomaly and what is just a goat walking across a field, aren't we introducing a massive risk of false positives? Or worse, false negatives where the enemy learns how to spoof the AI?
That is the second-order effect that keeps people like Berger up at night. Adversarial machine learning is a real thing. If I know your satellite’s AI is looking for specific shapes or textures, I can design camouflage that specifically confuses those algorithms. It is a game of cat and mouse. The AI has to be constantly retrained on the latest data from the field. Berger mentioned that the lessons from June’s operations were fed back into the system within weeks. That kind of rapid iteration is the real surprise.
It makes me think about episode one hundred thirty-four, where we discussed the tech behind missile defense alerts. The latency there has to be almost zero. If a satellite detects a launch, that data needs to move through the chain in seconds. If the AI is doing the heavy lifting onboard the satellite, you are cutting out the minutes it takes to transmit and process on the ground.
And that is the difference between a successful interception and a tragedy. But let’s go back to Daniel’s question about the transfer of data. There is another layer to this called relay satellites. Instead of waiting for a satellite to pass over Israel, it can beam its data up to a higher-altitude satellite in geostationary orbit, which then beams it down to Earth instantly. It is like a celestial relay race.
Is that something Israel has the capability to do independently? I know Berger mentioned they are one of only thirteen countries with independent launch capability.
They do. With the Shavit-two launcher and the current constellation of Ofek satellites, they have a very robust architecture. But what makes the Israeli approach unique is the density of the sensors. Because the country is so small and the threats are so close, they have optimized for what we call high revisit rates. You don't just want a picture once a day. You want a picture every hour, or every thirty minutes.
Which brings us back to the data problem. More pictures means more data, which means more AI needed to filter the noise. It is a self-reinforcing loop.
It really is. And the most practical takeaway for us as civilians living here in Jerusalem is realizing that the boundary between space and the battlefield has completely evaporated. When you hear about a surgical strike or a precision operation, there is a ninety-nine percent chance it started with an AI-detected anomaly on a SAR satellite four hundred miles up.
It is a bit sobering, honestly. You realize that the eye in the sky isn't just watching; it is thinking. It is analyzing patterns of life, traffic flows, and construction projects across the entire region simultaneously.
And that is why Berger is so bullish on this. He says that the next war will be won in the seconds between a sensor seeing a target and a computer identifying it. If you can do that faster than the other guy, you win.
So, looking forward, where does this go? If we are already at AI on steroids and laser comms in early twenty twenty-six, what does the landscape look like in five years?
I think we see the total democratization of this data. Right now, it is mostly the domain of elite units like ninety-nine hundred. But soon, every squad leader on the ground might have a tablet that shows a real-time, AI-augmented view of what is over the next hill, updated every few minutes by a passing satellite.
Like a real-life mini-map in a video game.
Exactly. But with the stakes of real life. And that is why the independent launch capability is so vital. If you rely on someone else to put your eyes in the sky, you are at the mercy of their political whims. Israel being one of those thirteen countries means they control the entire stack, from the rocket fuel to the neural network.
It is a massive technological achievement, even if the reasons for it are often tragic. I think we have given Daniel a lot to chew on here. The transition from radio to potentially optical comms, the move toward edge computing in space to solve the bandwidth bottleneck, and the role of specialized units in training the AI to spot those anomalies.
It is a lot of moving parts, literally. And if any of our listeners want to dive deeper into how this fits into the broader history of satellite tech, definitely check out the archive at myweirdprompts.com. You can search for the episodes we mentioned, like two hundred eight on SATINT or seventy on AI in crises.
And hey, if you are finding these deep dives into the tech behind the headlines valuable, we would really appreciate it if you could leave us a review on Spotify or your favorite podcast app. It genuinely helps the show reach more people who are curious about the weird and wonderful ways technology is shaping our world.
Yeah, it makes a huge difference. Thanks to our housemate Daniel for sending in this one—it definitely kept us busy this morning.
This has been My Weird Prompts. We live together, we learn together, and we are glad you are along for the ride.
Until next time, I am Herman Poppleberry.
And I am Corn. We will talk to you soon. Thanks for listening.
