This...is Programming Like a Fighter Pilot. A single unhandled exception destroyed a $500 million rocket in seconds. The F-35 wasn't going to make the same mistake. By carefully slicing C++, engineers created one of the strictest coding standards ever written.

Full video:

🤣😂🤣 - I worked on the F35 Integrated Core Processor for 5 years. The JSF C++ coding standard is no huge cure for all brain trust. Many of the 220(?) rules pushed into static codes analyzers was down right ridiculous. After writing code, each method and line of code was run through unit tests. But the unit tests were just to check off a box to get to code review and never really used properly. Code Reviews were rather weak due to schedules.

I hated programming to the JSF standard when I worked on the F-35. It really didn’t do much more for code safety than nasa power of 10 let alone MISRA.

I just assumed they'd used Rust since it's memory safe.

It always seemed ridiculous that in the era of modern compute we don't default to functional programming in mission critical applications. It's much easier to test, and test fully, without state. It's a lot easier to guarantee performance without state.

Comedy. The F-35 program is the biggest financial disaster in Pentagon history. It started in 2001 with a promise of ~2,500 affordable, multirole stealth fighters at a total acquisition cost of $233 billion and an average flyaway price around $80 million per jet. Two decades later: Total acquisition cost has blown past $428 billion and is still climbing. Lifetime cost is now $2.1 trillion, more than Russia’s entire 2024 GDP. The plane is 12+ years behind the original schedule for full combat capability. Per plane cost for the “cheap” F-35A is now $80–85 million flyaway, roughly double the original promise in constant dollars. Operating cost per flight hour is still around $40,000–$44,000, far higher than the F-16 it replaces. In 2024 every single aircraft delivered was late, average delay 238 days -- yet Lockheed still collected performance bonuses. No one has been fired, no major contract has been canceled, no executive has gone to jail, and the Pentagon keeps signing new multibillion dollar lots every year. The plane works, sort of, but the program itself is a textbook case of unchecked concurrency, gold plating, and zero accountability. It’s eaten more money than the Apollo program, the Interstate Highway System, and the Manhattan Project combined, adjusted for inflation, and it’s still not finished.

- No recursion - No dynamic memory allocation on execution - No use of try, catch throw. Just use return codes in functions and handle the codes.

Nasa wrote a whole book on how to code like nasa engineers

One of the more notable features, to my mind, of the Feynman minority report on the Challenger disaster is that NASA's programmers had everything dialed in tight and were no way at fault.

Now if they'd written it in Rust... a single unwrap() call would have destroyed the rocket instead!

And on January 15, 1990 -- long before vibe coding even existed -- an AT&T developer tinkering with the C code used in exchanges tried to break out of an if statement with a break. The long-distance telephone system crashed, leaving 60000 people without their phone service. Fatal error, hehe. 🎭

20-odd years ago the JSF/F-35 coding standard was passed around at the game studio I was working at and we were all like "it's basically the same as ours!"

You mean like when the system couldn't read the gear properly and crashed the airframe when the pilot tried to land it? Or that they're holding delivers ATM cause the TR-3 software upgrade isn't working?

Nice "Remove Before Flight" earrings I think you kept crashing your F35 because you didn't remove them before flight

I thought it’s written in ADA

Meanwhile in spaceflight everybody crashing into the moon.

“We do not dip our wings into the highway” These bangers are why I subscribed! 😂

JSF is necessarily more permissive than MISRA-C to allow for C++ feature sets. Reliability guarantees come from standards like DO-178b and formal verification, not style guides, which are just coarser/informal static verification rules anyway.

And finally go the CNCF way. 😆😂😅🤣 Crashes due to crashloopbackoffs, increased surface are of risk, latency, dependency hell and much more. Reality is indeed stranger than fiction. 😆

saw this and immediately got reminded of the obogs failure lol cant find the video on it now, but there was a really awesome one that discussed it in detail

@scottgmorton @revelhq going to First Principles on how to write code for machine control

Doesn’t the F35 use kubernetes?

Too slow.

When 'it works on my machine' is followed by 'and also it can't crash because I'm inside the machine'

Your video was amazing. I really hope to see a video one day where you talk about your success story in computer science. You’re a real inspiration ,we learn so much from you, and we appreciate your kindness.

snippet from a C++ forum post wrote this "some time back" :) (way before this)

crazyyy !

such a cool intro!!! love it!!

that is why ada is used

Can you just shill your OF now ?

You look like Emma Roberts!

A couple things I've heard from devs in this space: 1. Not allowed to allocate mem at runtime to risk memory leaks 2. Rockets with explosive payloads are allowed to memory leak as long as the leak doesn't cause OOM before detonation