But if you are using an AI agent, does the language matter? Plus it's an excellent way to learn Rust.

Yeah, it is a little grating that software tends to feel like it's running "at the same speed" as software from the 80s or 90s. As hardware gets faster and faster, software seems to just get more bloated to compensate. Frameworks and libraries can be too blame for a lot if it. But also the lack of need to optimise anything. I'd love to see a modern (young-ish) developer trying to write something for the 32k machines of the 80 (C=64, say), or even the 512k machines like the Amiga and Atari.

That would be a fun Hackathon to watch!

You're writing this new in 2026. You're using AI Agents to do a lot of the work, it seems, based on the files in the repo.

And you are writing it in C ??!!?

Now, I love C -- I love that you can do anything at the hardware level, and do crazy stuff with pointers, and all the low level goodness. It's a fun language, seriously powerful, fast.

But C makes it notorious easy to introduce a plethora of security issues! Not matter how hard you (or your agent) try, they WILL get in there. Buffer overruns being the simplest to exploit, and the simplest to accidentally include (and can be hard to find too).

So why not Rust? Why not just take the easier option to safety. Especially with something that you want many other people to be using. Rust is powerful, secure, fast. I really can't see a reason to not use it for a project like this.

I daresay it's not too late to have your agent rewrite everything in Rust. It's 160kloc. It won't take it long.

The more online-meeting software packages, the more simultaneous meetings we can be in!

The reason I make the distinction, and dislike the phrase "reached orbit", is because of the number of people who turn around and say "they haven't reached orbit yet!". Not reaching orbit by choice is very different to not reaching orbit because the rocket failed.

Very very different things to be comparing.

Saturn V wasn't launching a fully reusable 100t ship on a funny reusable booster. It was launching 4 expendable stages, one of which, with a mass of only 5.5t, would reenter and land, but never be reused.

The level of acceptable risk with Saturn V (and the Gemini and Apollo programmes as a whole) was also much higher. Saturn V wouldn't be allowed to fly today -- the FAA would never grant it a licence! Lol.

Plus the design and testing approach was very different. Saturn V (and Blue Origin) had a lot of testing on the ground, with multiple rounds of testing on individual components -- the first time they were all tested together was on its first launch after many years of planning and development.

The Starship programme has done it's design and testing throughout these test campaigns. They built something and flew it, and then knew what worked and what didn't, and changed stuff, and flew it. And charged stuff and flew it. Rinse repeat. They were able to test actual conditions of launch, space flight, and reentry.

Two very different methods. The SpaceX approach is more expensive, but results in a much better and refined product. It allows for design, full integration testing, and build to run along side by side so that changes can be made throughout the programme. The end result isn't what was on the initial design because the tests feed into the design improvements in an ongoing basis.

Saturn V took 5 years to design and build (nothing reusable). New Glenn took 10 years (reusable booster, no reusable upper stage). Starship is at 14 years, has reused a booster, and test landed multiple upper stages. (Hopefully a catch and reuse soon, maybe with the next month).

All different beasts. All impressive in their own right. All different design choices and goals. All amazing engineers.

Typically "reached orbit" means that an orbital rocket achieved orbit because on a launch nothing failed. It's all too common that a first launch of a ticket fails. Falcon 1 failed to achieve orbit twice, iirc.

Starship, because of their iterative testing process, just hadn't achieved orbit because they chose to not enter orbit. They could have, but because the ship is designed to survive reentry, they chose to stay suborbital during the entire test campaign, out of an abundance of caution.

I feel this is different to "not reaching orbit". They could have. They were only ever a few seconds off orbit on all of their suborbital flights.

Today to decided to do an extra burn, 15s long with 1 engine, to raise the perogee out of the atmosphere. Today they chose to go orbital.

And an amazing milestone.

But the wording "reached orbit" just doesn't feel right for this approach and for this achievement.

It's fine -- as the AI is defending itself in court, a compaction will happen, and while it re-researches everything it'll delete the law book it's using, switch tack entirely, and declare itself guilty.

Don't tell him NASA use metric or he might actually do that!

No, no, he knows stuff without having to be told..!

The US will do anything to avoid going metric! /s

But... But... Boomers are bad! You can't go telling them all the many many many good things Boomers have done -- that's completely against "the narrative"...!

which rules Trump himself it, I guess. How does he define "high I.Q." anyway?

Yep, use root for this so that if they hit the replication server first they have full access to your production server. smh.

How many times has an Antropic AI hacked other systems now? 4? 5? That we know of, anyway. A rogue AI hacking the wrong system and "pressing" the wrong button, and up goes a nuke or 5. Could be by misleading an early warning system (and Stanislav Petrov isn't working that day). There are a number of similar scenarios. I doubt an AI will kill us because it has a "kill all humans" vendetta -- it'll be something unplanned and unforeseen and accidental.