fxtentacle

„Fitting“ is still too nice of a word choice, because it implies that it’s easy to identify the best solution.

I suggest „randomly adjusting parameters while trying to make things better“ as that accurately reflects the „precision“ that goes into stuffing LLMs with more data.

Related to your answer, I would say the reason is that it works good enough for now and it can always be patched later. Back in the good old days that we remember, software was frozen on a gold master disc, which was then tested for weeks or months before its public release. The fact that bugs could not easily be fixed in the field meant they would incur support costs or lost revenue with people returning their purchased software box.

In my opinion that is the true reason why the old native software was developed to such a high standard. But then once online stores and shrink wrap agreements made it impossible to return buggy software, then the financial incentives shifted towards shipping a partially broken product.

Who cares about pleasing with good performance when you can instead keep customers hostage?

1993 was before the west entered the last stage of capitalism. It was a time when companies still competed on products rather than using monopolistic force to squeeze ever more revenue out of the same people by turning every life necessity into a subscription. Similarly, it was a time when you could mail-order a house and build it yourself. Rental prices were low because there was no regulatory capture on housing construction yet.

Where I disagree with you is video streaming. In my opinion, YouTube and the commercialisation of holiday memories (which later became Instagram influencers) were the beginning of widespread depression. Seemingly regular people sharing their exceptional life somehow forces everyone else to compare themselves to the dreams presented on YouTube and most people will come up short and then most people will feel insufficient. I believe that’s why early YouTube ads were so powerful. Your ad for exotic goods would play immediately after the viewer became painfully aware of how boring they are, when measured against the top 0.1% on a global scale.

If all the AI revenue projections were correct, then 1% of worldwide GDP would end up at AI companies. Or said differently: you buy a sandwich for $5 and somehow AI gets $0.05 out of that transaction.

I think TSMC laughed them out of the room when they announced the original numbers. So maybe there’s no reaction now because everyone already knew not to trust OpenAI’s promises.

Unbelievable! Next you’ll tell us that Elon‘s self driving car promises were all just hype for the cult…

I once lived in Singapore for a while and we were all sure that nobody would steal anything anyway, so we just never bothered to lock the doors. (That was also very helpful if you wanted to stop for a quick coffee with a date in the middle of the night.) You could see the MacBooks from the street, but nothing ever went missing. I don’t know what exactly it was, but Singapore felt incredibly safe and crime-free.

Are we sure it's not failing anymore after the advent of LLMs?

The Airbus A320neo can already takeoff, ascend, cruise, descend, and land all by autopilot. It can even download your flight plan from the airline's servers.

But you still need the pilots because the system can only handle the happy path. As soon as there's any blockade or strong weather change, the autopilot will just turn off. And then you need the pilots.

I would say software engineering with AI is similar: The AI can handle CRUD just fine. But once things get messy, you need someone who can actually think.

With a bit of AI sprinkled in, Rust code can surely also waste gigabytes of RAM on "Hello World" ;)

Scathophagidae are flies that really like eating shit. We know how to cheaply produce massive amounts of shit.

But that doesn't mean we solved world hunger. In the same way, AIs churning out millions of lines of code doesn't mean we have solved software engineering.

Actually, I would argue that high LOCs are a liability, not an asset. We have found a very fast way of turning money into slop, which will then need maintenance and delay every future release. Unless, of course, you have an expert code reviewer who checks the AI output. But in that case, the productivity gains will be max 10%. Because thoroughly reviewing code is almost the same amount of work as writing it.

This website offers a very venture capital take on ideas. They present it as if the single determining factor if an idea was great is the valuation at which it was sold. I don’t really understand why they think that should nullify valid criticism about the early beta versions of software that turned out popular later.

Also, just because a company has a lot of revenue that doesn’t automatically make it a successful company. Economics crashed and burned while still growing revenue day over day. And the jury is still out if OpenAI will ever be able to pay back the billions it borrowed.

It’s what everyone used back in the Quake 3 days.

Adult clothing no. But for kids clothing you kind of need it, because without the cutting function attaching rubber bands to neck, wrist, and belly seams is very difficult.

I have never tried that because it finishes a single piece of clothing so quickly that it never seemed like it would make any sense to take risks of lower fabric layers being uneven/wrinkled in exchange for increasing throughput. Catching fire never seemed to be an issue with the CO2 laser that I used. Most fabrics will kind of melt before they start to burn. And the vacuum table sucking air through the fabric will also cool it down.

I also fixed clothing sizes for my family using a hacker mindset, but in a different way:

Did you know that most professional sewing charts are just DXF files?

And did you know that DXF is the most common file format for laser cutters?

;)

=> just let the machine cut out precisely the clothing shapes that you need

After a few tries, I also started to add small sideways cuts to the outlines as alignment markers. And then you just need to connect the pieces where you marked them while always leaving roughly 0.5 cm of gap to the laser cut line. I went with 0.5 because my sewing machine has a hardware alignment guide with that offset. And at that point, it takes a skilled tailor only mere minutes to finish a shirt, which means in exchange for their 1x hourly rate they will be willing to finish off 5x proto t-shirts for you.

Maybe we just need to go all the way: How about a WASM core with a React GUI that runs inside a custom Electron renderer which outputs the TUI? 100% CPU guaranteed. And you'll never find that important piece of information in an all-monochrome wall of text with no icons. Why use a low-level print() when you could improve your productivity with a high-level framework? /s

I've also learned that it is a good idea to physically put my hand around the payment thingy to make sure I am the one pressing the buttons.

You could hire a reasonably skilled dev in India for a week for $1k —- or you could pay $20k in LLM tokens, spend 2 hours writing essays to explain what you want, and then get a buggy mess.

1. there's no official mention of any CE / FCC certification

2. I asked shortly after the author offered to answer questions, but there was no response

I'm starting to wonder if this project maybe just shipped without CE / FCC certification because they didn't know it would be required.

"I had forgotten that there could ever be domestic conflicts where law enforcement were not themselves belligerents."

ouch. Must be weird living where you live.

For colours to look natural you need your white light to contain lots of different wave lengths. It’s usually measured as Ra. Artificially looking LEDs are easily 10x cheaper than photography grade LEDs. Also, this guy is probably paying taxes and handling stuff the proper legal way. If you order from Alibaba, chances are you’ll not be paying taxes. Plus if they offer a 5 year warranty, they probably need to keep some money around for repairs.

Do you have any recommendations for FCC/CE testing providers?

This turns point clouds into meshes.

That means it doesn’t need depth. Depth is helpful for getting good point locations, but SLAM on multiple frames should also work.

I’m guessing that they are researching this for AR or robot navigation. Otherwise, the focus on accurately dividing the scene into objects wouldn’t make sense for me.

Yes, incremental improvements is how most science gets done.

The US has become a world leader in suing curious people into submission. As soon as you touch any commercially available tech and do anything that the manufacturer dislikes, you're at risk, thanks to § 1201:

"No person shall circumvent a technological measure that effectively controls access to a work protected under this title."

Extending your digital camera with new firmware? Illegal.

Inventing a custom ink or add-on for your printer? Illegal.

Repairing a tractor? or a ventilator? Illegal.

How do you expect anyone to get world-leading science done in this environment?

Let's call it "form over function."

That landing page example is devastatingly bad. You start with a page that has usage numbers, uptime, support 24/7 and a customer rating above the fold. You end up with a page that lacks all of these advantage and instead looks bland and has horrible typography and even less text contrast.

In line with that, the Dashboard looks more organized in the "after" picture, but that's because it lost most of its useful information.

Another good idea is to check a trademark DB, because "Mail Pilot" is actually a registered trademark of Mindsense LLC.

To me, this sounds like:

AI is helping me solve all the issues that using AI has caused.

Wordpress has a pretty good export and Markdown is widely supported. If you estimate 1 month of work to get that into NextJS, then maybe the latter is not a suitable choice.

I strongly disagree. I’ve yet to find an AI that can reliably summarise emails, let alone understand nuance or sarcasm. And I just asked ChatGPT 5.2 to describe an Instagram image. It didn’t even get the easily OCR-able text correct. Plus it completely failed to mention anything sports or stadium related. But it was looking at a cliche baseball photo taken by an fan inside the stadium.