In a redundant system, MTTR rapidly starts to be the dominating factor for high availability.

These guys seem to have the right approach.

Just train them with Australians, problem solved.

So you to log into the extension and it's $9.99 a year, then on the left and of the screen it's 19.99 a year, with 7 day free trial

Recursion can be good, predicting you will never run out of stack or not, can be hard.

It's a little hard to duck out to Mars to reboot the lander after you bricked it with a recursive function that never exited.

Sometimes, eg aspects of automotive functional safety, MISRA C might be all you get.

These NASA principles are more about enabling better possible static analysis of the code and ease of someone else, maybe decades later, debugging or pushing changes to something likely on another planet.

Also you have to remember space based computing lags well behind terrestrial computing because of the radiation hardening. They are often still dealing with legacy systems that might be 8 bit with very limited memory, they were still in the hardware expensive engineers cheap mode until well into the nineties, if not later. Rad750s run at 400 mips and were, and maybe are, preferred choice of processor.

These days, there are compilers for embedded systems that can prove for certain code (for example) that it runs in constant time and constant space. As an example, galois.com has been doing this for Haskell, not just for embedded systems, but also for even more low level things like FPGAs.

I tried to get it to build me a slightly challenging app to break out data from a fairly obscure file format for some PLC code, after having tried with Claude.

o3-mini produced volumes of code more quickly and more of it, but Claude still had greater insight in to the problem and decoded the format to a noticeably greater degree.

Whereas 03-mini quickly got to a certain point, it wasn't long before it was obvious it wasn't really going any further - like it's big cousin, but in it's own way, it was lazy and forgetful, seeming at times more interested in telling me what I might try than actually trying itself.

Interestingly, even when I gave it a copy of Claude's code it still wasn't able to get to the same depth of understanding.

80% done, with no code.

This is often how a functional safety project works - specification is everything, no point writing any code until you know exactly what you plan to build.

Due to traceability requirements it becomes a very waterfall approach.

However one way of improving methodology and outcomes is an executable specification that can be converted to code and documenting at the very last minute, meaning you can be specifying right up until you need to deploy.

Executable specification concept comes from IEC 61499.

It takes notional 3 seconds to boot and get an IP address alone.

If you use ESP-NOW then you can wake up send msg and sleep again in 35ms.

It's a layer 2 protocol.

Second that, espnow + broadcast mesh is the way to go

Ask the Australian government, they are experts.

Can you elaborate more?

This is why the "cloud" model of AI is not really where I would want to be as a business.

Admittedly you are not going to get the same cutting edge performance, but I would only be basing a business on an AI I could download and run on my own hardware, such that at least current performance could be guaranteed going forward.

Anytime you buy anything as a service, you are in danger that due to malice, incompetence, lack of profitability, competing with a greater or established interest, or many other reasons, that service may become unavailable at little to no notice.

I don't understand why companies put themselves at these risks, when it comes to software as a service components to their business, when they wouldn't dream of exposing themselves to similar risks in any other domain related to their well being.

If you use it SaaS style and if you choose Llama or similar you are good. You can run it anywhere you can run a GPU. It is like using Linux on AWS. You get banned you can use Azure.