leoedin

> I think where intuition fails a lot of us (myself included) is that in the software world if you have the code/documentation for something then building it is trivial. But in the hardware world, even if you have all the instructions and knowledge, actually applying that know how to create a working project is still very far away, through certainly nowhere near as far away as if you have started from scratch.

It's certainly true that you can build software if you have the code (well, mostly true. Modern frameworks suffer from quite severe bitrot). But I'm not sure that you can extend it.

Large software projects rely on people understanding them. If everyone who worked on the project is gone, it's very hard for someone to regain the architectural understanding required to extend or refactor things. Without any overlap, there's nobody to answer questions. As long as there's been a continuously overlapping stream of developers, that knowledge is more or less retained - but if all the developers leave, and someone picks up the project again later it's almost impossible for them to make meaningful improvements.

AI probably changes this completely.

This is basically how the rest of the world feels with US tech giants.

It's not necessarily a bad thing if there's balance to it. Maybe it's OK if Germans are great at industrial automation and Americans are great at writing word processors. But I'm not sure there's a nice balance to it. Especially if you're in a developing country which is good at neither of those things.

I think it's been a long time since any motor racing was just about engineering the best possible car. The problem is that engineering got too good - the best possible car we could make would simply be far too fast to be safe. So F1 (and all motor racing) are always playing the rules to try and keep an even field (otherwise the best funded teams would be leagues ahead) and to stop the engineers from killing the drivers.

There's no way western militaries would just add troops to the front. If we're fighting a war with Russia - regardless of where it is - it would be no holds barred. And just adding more troops to the front is not the way an advanced tech-heavy loss-averse western military would do it.

The only way western troops are deployed en-mass inside Ukraine is as a peace guarantee with some level of Russian agreement. The moment there's footage of British/French/German troops getting hunted down by drones, the public will be crying for direct strikes into Russia and it'll be WW3.

The maps the operators use in the UK still look like digital versions of these control panels. Sometimes they're visible on screens in stations.

In fact, there's a live online version: https://www.opentraintimes.com/maps/signalling/tl-lbg-1

Yeah, I assumed 50% to account for various losses and other less efficient generators. I don't know what the actual number is - probably in the 40-50% range.

Interestingly thanks to renewables the "carbon intensity" of a kWh of electricity is already better than a kWh of fossil fuels in some countries. A kWh of petrol releases about 250g of CO2, and quite a few countries (predominantly developed ones) have lower carbon intensities per kWh of electricity. Considering round trip efficiencies the carbon released by an electric car in those countries is easily 1/3 to 1/4 of the carbon released by a petrol car.

In the USA it's not as good, but electric cars are still releasing about half the carbon.

It just seems like everyone is re-inventing the tech of the 50s and 60s, but at a fraction of the cost by using a completely different set of parts, supply chains and cost structures. And massive eventual volumes.

My guess is the Ukrainians will put a rocket on their interceptor drones, and develop some sort of radar or heat seeking guidance system. Perhaps we'll end up with an interesting hybrid interceptor missile that uses electric propulsion for part of the journey?

The book says 50km/30 miles per day in a petrol car is 40kWh. For a Tesla that would be 7.5kWh of electricity. It also assumes 12 km/litre - which in UK terms is 33 mpg. A Tesla has regenerative braking, so it would be better to compare the fuel efficiency to an aerodynamic hybrid. A Toyota Prius gets 61 mpg (UK) - which would mean 21.6 kWh per day.

So really the comparison is 7.5kWh of electricity compared to 21.6kWh of petrol. The Prius gets 34% of the miles compared to the Tesla for a given energy input. In the olden days when renewables didn't do much, the chemical energy input for 7.5kW of electricity was probably 15 kWh. So the electric car is a little bit more efficient (taking advantage of the efficiency of large power plants).

These days, thanks to renewables the fossil fuel input into the electricity is lower.

See also - mega rail projects that start from nothing (cough HS2).

Politicians think doing engineering projects is like buying something off the shelf. Every engineer will tell you the only way to do stuff is iteration - starting small, learning from your mistakes and growing things. If you start a megaproject without a team of people experienced in doing more or less the same thing recently, it will cost vast amounts. Particularly if you push too fast and commit large amounts of money before you've had a few iteration cycles.

Are the models "doing the calculation" or are they calling a calculator tool? There's a lot of talk about how models can do maths now, but I'm struggling to understand if that just means they just need to recognise that it's a maths problem and pass it to a tool, or if they're truly doing the numerical manipulation themselves.

I was thinking that one of the great ironies of artificial intelligence is that it will make good teaching even more labour intensive. Those kids that can afford the one to one tuition required to push past our natural inclination to be lazy will learn things, while everyone else will succumb to outsourcing all their thinking.

There’s no way I’d have learnt everything I know now about software in today’s environment. So much of software development - debugging techniques, architectural decisions, structuring data etc is learned through trial and error. As the models get better is becoming easier and easier to just not look too closely at their output. I suspect it’s human nature.

I don’t enjoy clothes shopping, because I hate trawling through stuff and spending hours not finding anything. I’d love to have a personal shopper who could judge my style, my size and my needs to curate a list of things to buy.

Right now AI can probably do the buying part, but seems to be absolutely no where near the curation part. Maybe that’s a good thing for humans, so I’m not too fussed.

ARM is a fragmented mess compared to x86. Switching to ARM (right now at least) means losing the ease of x86 booting. In the x86 world, trying another Linux distro is as simple as downloading it and putting it on a USB stick. You can be 95% certain it will work. In the ARM world, nothing is standardised, nothing will ever work first time.

There are projects trying to tackle it, but compared to x86 it's a fragmented mess. That's what we're losing with ARM laptops - universal computing.

I've been working on embedded systems using iMX8s, and it's equally awful. You have to maintain your own fork of UBoot. You have to spend hours tweaking device trees. The "NXP" kernel is out of date and never updated, the mainline kernel is missing loads of drivers.

Meanwhile in x86 land you can just download a distro and boot it. Secure boot works out of the box. It's night and day.

Suddenly the ban on computer networks in Battlestar Galactica doesn't seem so far fetched! We're not that far off having enough distributed compute for LLM viruses to install corrupted LLMs everywhere and then start attacking everything within reach.

What an empty article. Testing an OTA update can survive a power interruption during the downloading is such a basic test. It doesn't need an article that long. I guess AI wrote it from a few sentence prompt?

They didn't do any interesting tests. What about brownouts, voltage ripples, timing the reset signal to see if there's any critical moments in the update process, corrupted update files, high EMC environment etc etc. That would be interesting.

I think the article is an advert for a testing platform that enables automating this sort of test. But weirdly they don't show how their platform does the automating, so it just looks like they're really pleased with themselves for doing really basic engineering.

I had very few issues using Pop OS on a Dell for all sorts of corporate stuff. It wasn't perfect - the sleep behaviour was particularly terrible - but it worked well with Teams and Google Meet. I later installed Fedora and that more or less fixed the sleep issue too.

I bounced around between Windows and Linux on my personal computers for the last 20 years. Generally I'd run Linux until something died or I needed Windows, and I'd return for a bit. I've been exclusively using Linux since 2019 or so. I don't think I'm going back.

I was wondering how much financial value they put in one truck. Presumably the value density of this stuff is incredibly high (although not really to anyone else, so security probably isn't a huge issue). At what point do you split the shipment between multiple trucks to avoid huge losses if there's a crash?

When I worked on satellites we had some electronics which used a radiation hardened FPGA which cost something like $500k per chip. Carrying that around was nerve wracking.

Thermostatic mixers make sense for continuous flow applications - showers, filling a large bathtub for example. But I don't think they make sense for a kitchen tap - where you might be filling a small basin.

The goal is that the water in the basin is the correct temperature, not the stream filling it - which means you need to compensate every time based on how much cold is already there before the hot water arrives. And that varies seasonally.

The time it takes for hot water to arrive tends to be a big proportion of the total time the tap is running, so you actually don't want controlled 40C water once hot water arrives. Perhaps if you had instant hot water, this would be useful.

Yeah, this isn't "here is a complete solution", it's "here is an interesting package for optimisation, applied in a specific way that you probably haven't heard of".

I found it quite interesting. The guy isn't going to start a company around it. United aren't going to drop what they're doing. But now more people know about scikit-decide and that is a good outcome.

No it's not. It's an interesting article on how to use scikit-decide to optimise a problem. Nothing more.

Rolling your own crypto is considered bad practice because there are countless battle proven open source alternatives. Where are the open source alternatives for aircraft route optimisation that the article could talk about instead? And it would be a totally different article if it was talking about how to use them. This is an article about using optimisation tools to target a specific problem. Aircraft routing is just the example.

I don't bother buying Nike's because every time I see a pair I like online, it turns out they're "collectibles" and sold out instantly.

Yeah, I can go and buy a pair of black Nikes in any shop today. But even though they make lots of variety, they never seem to make it in enough volume that I can buy it. I'm not interested in being a collector - I just want shoes that aren't black, white or grey.

So I shoes from companies that don't have as much weird hype around them.

> One of the best political moves in a company is to have a lot of people under you.

This is why I’m sceptical of “AI will kill corporate jobs”. A big company has no profit sharing for successful departments. Everyone is trying to get budget to hire because it makes them feel important.

Many people are terrible at writing. LLMs write better than them.

However, many more people are "OK" at writing. But what their writing conveys is personality. Every comment here is written by someone who may not have grammatically perfect writing, but their writing conveys how they talk and think. It conveys what they think is important, and what they brush over. It conveys how much they care about the topic being discussed. Behind each comment is a person. Online forums and discussions are, at their heart, a shared experience of humanity.

LLMs write consistently in the same personality. If your writing is filtered through an LLM, your personality is stripped out. Unless the idea being conveyed is particularly novel or interesting, you might as well not have bothered. Imagine a niche forum where everyone discussing things was speaking through an LLM. It would be BORING.

I worked for a startup in 2019 which was really pushing the "Spotify model" of team structure. Apparently it's all about self-organising teams and having no leadership. The fact the technology output of the Spotify model - the Spotify app - seemed to be kind of crap didn't seem to matter. I think the main output of the Spotify team structure was blog posts promoting the Spotify team structure.

I gave up on antigravity when they retroactively changed the limits. I learnt that I'd exceeded the new, lower limits when I sat down to use it one evening - despite not having touched it for a few days. I don't mind limits, but something about the retroactive application and how it unexpectedly forced me to change my plans really angered me - I cancelled my subscription that day and have used other tools since.

A real lesson in how not to treat your customers. I would probably still be paying for it if they'd applied the new limits more softly.

I'm in a similar boat. I don't think "knuckle down and learn something" has ever worked for me. It's more like "set things up so the learning aligns with my interests" and then... woosh! I'm off and deep into the topic. That was something I had to teach myself (nobody at school said anything other than "pay attention"). And quite often I still miss and steam off in the wrong direction.

> Accepting to run random code sent by someone you don't know should be a red flag for them. Getting asked to run random code sent by someone you don't know should be a red flag for you.

This is a different problem. If a candidate said "I don't want to run the code you send me" it might be a good opportunity to talk through what it does. Maybe I'd set up a VM for them in the cloud, or we could find a mutually acceptable solution. If a candidate says "I don't own a general purpose computing device" then I would really be questioning that candidates interest in computers, and therefore ability to develop and test software for them.

> Why do you think the job posting doesn't explicitly mention "owning a PC" as a requirement?

The job description also doesn't say "owns clean clothes", but how the candidate presents themselves is certainly a consideration when hiring.

I'm talking about hiring a software engineer for a software development job. I have not met a good software engineer who wasn't also some level of geek. I'm thinking of all the great colleagues I've had over the years, and I would be really, really surprised if any of them didn't have some sort of computer at home. These are not "I just use my phone for everything" people.

I'm sure there are outliers, but given the low cost of used ex-corporate laptops in the western world, and presumably the candidate's interest in computers and software development, it's a fairly low bar to expect a candidate to have some sort of PC (laptop, desktop, Linux, Windows, Mac - whatever). I wouldn't expect it to be modern or expensive (I do this for a job and barely spend any money on computers these days), but I'd be really surprised if a candidate had nothing available.

Are those the people you want to hire for software development jobs though?

Honestly I would see that as a huge red flag. In western countries a used laptop costs almost nothing.