layoric

I discovered this project recently and used it for Himawari Standard Data format and it made it so much easier. Definitely recommend using this if you need to create binary readers for uncommon formats.

I must be confused, my understanding was m7a was 4th generation Epyc (Genoa, Bergamo and Siena) which I believe all have 2 threads per core no?

Exactly, and the performance of consumer tech is wildly faster. Eg, a Ryzen 5825U mini pc with 16GB memory is ~$250USD with 512GB nvme. That thing will outperform of 14 core Xeon from ~2016 on multicore workloads and absolutely thrash it in single thread. Yes lack of ECC is not good for any serious workload, but great for lower environments/testing/prototyping, and it sips power at ~50W full tilt.

> $50 for a dyno with 1 GB of ram in 2025 is robbery

AWS isn't much better honestly.. $50/month gets you an m7a.medium which is 1 vCPU (not core) and 4GB of RAM. Yes that's more memory but any wonder why AWS is making money hand-over-fist..

Decoder for people reading:

PM - Product Manager

FS - Fullstack developer

FE - Frontend developer

BE - Backend developer

Interestingly the French is completely different.

https://github.com/microsoft/edgeai-for-beginners/blob/main/...

> Unlike railroads and fibre, all the best compute in 2025 will be lacklustre in 2027.

I definitely don't think compute is anything like railroads and fibre, but I'm not so sure compute will continue it's efficiency gains of the past. Power consumption for these chips is climbing fast, lots of gains are from better hardware support for 8bit/4bit precision, I believe yields are getting harder to achieve as things get much smaller.

Betting against compute getting better/cheaper/faster is probably a bad idea, but fundamental improvements I think will be a lot slower over the next decade as shrinking gets a lot harder.

For clarity, all the parent said was John Jumper is the current director at Google DeepMind. Which I think is useful context, not sure why flagged..

Not having snow accumulate on the panels definitely will be contributing to that gain since a bunch is lost on more horizontal panels in those parts of the world due to a layer of snow sitting on top for quite some time after the event.

One big trade off/risk is a large vertical panel essentially becomes a sail in high winds.

Thank you for this context, should be the title IMO..

The on going costs via power consumption are on a completely different scale

And who is on the other side of that debt? And how did it get there?

> I think people also like the idea that there can be these systems in place for hundreds of years, and an individual can come along and intelligence and hard work, can turn the systems upside down or develop something better.

My interest over the years of Bruce Lee was much more from this perspective. Many stories talk about how hard he trained, and other aspects of essentially an underdog story. Combined with his communication[0], he comes across very thoughtful, and very grounded in many ways. Putting anyone on a “legend” status pedestal is always fraught with issues, but definitely a figure that inspired a lot of people.

https://youtu.be/uk1lzkH-e4U?si=Uu44M-UC1tKYv894

Right, the last 10% will be expensive or you accept a potential 10% failure rate.

Same with every other aspect of their offerings. Look at EC2 even with instances like m7a.medium, 1 vCPU (not core) and 4GB memory for ~$50 USD/month on demand or ~$35/month reserve 1 year. It isn't even close to be competitive outside other big cloud providers.

EDIT: clarity on monthly pricing.

Just by the nature of the product, the manipulation will be a lot more subtle, and likely more successful than traditional ads..

Yup. If you want something to plumb some pretty low volume events, sure serverless like Lambda can be useful. Anything which would be considered high levels of compute, you are just wayyy over paying. Hell, even EC2 on spot instance is expensive compute. I do like some AWS services, but yeah they come at a premium that is just getting more and more expensive.

The VPS is like renting office space. You don't own the space, but for the most part get to use it how you want, and all the responsibilities that come with that.

"Serverless" is like paying for a hot desk by the minute, with little control of your surroundings, but it is convenient and cheap if you only need it for an hour.

> Creative destruction is brutal math. The capital? Gone. Completely vaporized. But infrastructure isn’t stock certificates. Those fiber optic cables didn’t vanish when Pets.com did.

I read quotes like this and reminded that it is common that people forget money is just a competitive resource we use to outbid each other for _real_ things. Money moves around, it isn't lost or "Completely vaporized", someone receives it at the other side of the transaction. It is still in circulation, it can still be used to outbid people for real things, just by different people.

Also, pets.com still exists, it just forwards to petsmart.com.

It’s almost like they are cashing in on their good name…

Software updates and data collection. Eg, my mums Toyota Corolla 2018 already has a disabled infotainment button because of dropped support. If it was a 2019 it would have been eligible for an update, but not for 2018.

Lots of cars from the same period are collecting and sharing data to various different companies from weather to insurance.

Personally I don’t want monitoring or software updates, and definitely don’t want any cloud dependencies.

Same, early 2010s IMO seems to be the point where the industry really started to shift. There are some good cars from before this time, but keep them running past the 2030s will be a challenge.

In Australia, AWS dev position in 2021 had a take home. Microsoft contract position before that didn’t but they were desperate to fill seats on a poorly executed gov contract.

As a sounding board for things you are already well familiar with, I agree, and have experienced the same, and that can be useful. It's also a much better experience than say using Google to do the same, or just a rubber ducky.

The NLP these models can do is definitely impressive, but they aren't 'thinking'. I find myself easily falling into the habit of filtering a lot of what the model returns and picking out the good parts which is useful and relatively easy for subjects I know well. But for a topic that I am not as familiar with, that filtering (identifying and dismissing) I do is much less finessed, and a lot of care needs to be taken to not just accept what is being presented. You can still interrogate each idea presented by the LLM to ensure you aren't being led astray, and that is still useful for discovering things, like traditional search, but once you mix agents into this, things can go off the rails far too quickly than I am comfortable with.

Claude is decent for sure, but if you are using these models for 'smarts', that is a whole separate problem. I also think honestly people are sleeping on Mistral's medium 3 and devstral medium. I know it isn't 'smart' either (none of them are), but for mundane tasks need valid code output, it is extremely good for the price.

I agree, feels very LinkedIn sometimes.. that's not a good thing..

> it's that the system is intended to be general, and you can't even enumerate ways it can be made to do something you don't want it to do, much less restrict it without

Very true, and worse the act of prompting gives the illusion of control, to restrict/reduce the scope of functionality, even empirically showing the functional changes you wanted in limited test cases. The sooner this can be widely accepted and understood well the better for the industry.

Appreciate your well thought out descriptions!

Spot on. The issue I think a lot of devs are grappling with is the non deterministic nature of LLMs. We can protect against SQL injection and prove that it will block those attacks. With LLMs, you just can’t do that.

I’m not a fan of k8s, but “serverless” imo is not a good trade either, especially AWS lambda. ECS with ALB is probably an ok middle, but still expensive, even with heavy use of Spot instances. Cloud isn’t going to be getting cheaper..

No ECC is the biggest trade off for me, but the C236 express chipset has very little choice for CPUs, they are all 4 core 8 thread. Ive got multiple x99 platform systems and for a long time they were the king of cost efficiency, but lately the ryzen laptop chips are becoming too good to pass up, even without ECC. Eg Ryzen 5825u minis