phoehne

10 ms·

91.
by phoehne 4y ago
For the youths who don’t get the joke, Windows NT 4.0 limited their TCP stack to 10 inbound connections on the Workstation edition. 3.5.1 didn’t have that restriction. Server edition and workstation edition still had the exact same kernel
92.
by phoehne 4y ago
Brings back fond memories of building CORBA services and imagining a world beyond just DNS, DHCP, and a few vendor specific services.
93.
by phoehne 4y ago
This will probably get lost in the noise. But I think what sucks the fun out of side projects is where your agency is taken away by the ‘right way’ to do things. That’s what makes embedded and a lot of Raspberry Pi projects so fun. There
94.
by phoehne 4y ago
It’s a very sophisticated form of a recurrent neural network. We used to use those for generating a complete image based on a partial image. The recurrent network can’t explain why it chose to reproduce one image instead of another. Nor
95.
by phoehne 4y ago
Only when you’re asking for a memorized response. If you were at ask me to create a driver for a novel hardware device in Ada, there’s no memorized answers. I would have to work it out. I do that by creating mental models, which LLM’s do
96.
by phoehne 4y ago
It doesn’t really have a model for choosing. It’s closer to pattern matching. Essentially the pattern is encoded in the training of the networks. So your query most closely matches the stuff about X, where there’s a lot of good quality t
97.
by phoehne 4y ago
I think you’re right in one sense, and we both agree LLMs are not sufficient. I think they are definitely the death knell for the junior python developer that slaps together common APIs by googling the answers. The same way good, optimizi
98.
by phoehne 4y ago
I use the know it all uncle. Who does know a lot but is sometimes wrong. But either way, the answer comes with the full conviction they’re right. Certainly not good for medical diagnosis, but great for “I’d like a python script that retr
99.
by phoehne 4y ago
FreeNAS kinds of sits in that area, but it suffers from some of the same community contribution problem. It’s definitely hit or miss. I’m thinking the solution has to be like 10 core things that are rock solid rather than dozens of things
100.
by phoehne 4y ago
That's a reasonable plan. And I 100% agree that controlling your data is preferable to having it mined. What frustrates me, and why I don't do it any more, is we don't have a manageable story for most things. In many cases
101.
by phoehne 4y ago
Espressif has compilers you can download. GCC may work fine, but I’d use their compilers.
102.
by phoehne 4y ago
Sorry, I didn’t mean the userland stuff above the kernel, which, along with the boot loader, is the part that is really restricted in any sense. With the standard library, there are large parts that are not kernel safe. That’s because the
103.
by phoehne 4y ago
It's not impossible. Other people have done it. Some have done it with a minimal kernel or micro-kernel in C and then services on top of that in C++. Just doing a quick perusal of their code base, the low level stuff is functions an
104.
by phoehne 4y ago
It's been decades since I worked with Ada, so I could be wrong, but you could build it with an embedded runtime that allows you to do tasking and (I think I remember) exceptions. But it also has a language standard way to create limit
105.
by phoehne 4y ago
Some vendors base their compilers on GCC or Clang. But if you're not using their provided compilers, you either need to 1) be willing to shoulder the expense of making changes to GCC or Clang for whatever you're adding to the sil
106.
by phoehne 4y ago
The ARM cores used in STM32s have great support. I think, at this point, almost any language and toolchain you could ever want. Good debugger support. The RISCV cores are going to get there and that's why I'm glad Espressif is
107.
by phoehne 4y ago
There are 2 flavors of the ESP, depending on model. The Xtensa cores and RISCV, go with the RISCV variant. The Xtensa variant requires you to use their compilers, which are derived from he ones provided by Cadence. There's a better
108.
by phoehne 4y ago
Anyone who says C is not important hasn't worked on embedded systems or operating systems. I work on embedded systems. There are roughly 3 choices for languages[1]. C, Ada, and C++. Except the version of C++ is sometimes often vendo
109.
by phoehne 4y ago
I was actually trying to point out the irony of the idea that artists and programmers should get used to being displaced by AI, while this author won't? But that being said, I could see a world where the language model AI generates new
110.
by phoehne 4y ago
You'd be surprised. PDP-11s and emulated VAXen are not for hobbyists. They're for companies that still have that PDP-11 or critical VAX they can't get rid of. It's usually in an embedded or context - like running a nu
111.
by phoehne 4y ago
I don't miss it. I had several projects where it seemed like a good idea. I pushed it as a good idea. I really wanted it to work. Some developers (usually the sharper bulbs) got it. A lot of developers really struggled with it. E
112.
by phoehne 4y ago
I don't think the things that makes programmers gag is plain old XML. As other comments noted, the markup/content ratio is a matter of your choice. And, as opposed to binary formats from the 80's and 90's, human readab
113.
by phoehne 4y ago
Here’s something I don’t get. If the generative AI gets so good it can diagnose disease or give us perfectly curated experiences, why wouldn’t it just “generate content” at an expert level? Why would we need to compensate “content creator
114.
by phoehne 4y ago
FYI - I worked as a post-sales consultant for a NoSQL vendor that did a lot of search. This is what seemed the sweet spot for data ingest. ETL was part of what we did, along with search and other applications. I would never have called m
115.
by phoehne 4y ago
Math. A surprising number of people in 'data science' lack good understanding of basic statistics, leading to poor data quality or incorrect conclusions from the data.
116.
by phoehne 4y ago
I recently 'demoted' myself so I could focus on software development instead of talking about software development and managing software developers. I'm much happier.
117.
by phoehne 4y ago
I program in C professionally. It's a good language to learn how processors actually work. It's on the embedded side, where there are other options but either they're not mature (Rust) or were never really taken up (Ada).
118.
by phoehne 4y ago
I did, but then I realize it isn't technology that fulfills me. It's my sense of accomplishment because I build (at least a few) things and I like to learn new things. Even if those things have no day-to-day value in my work.
119.
by phoehne 4y ago
rpilocator.com tracks them. Jeff Greeling has a couple of good YouTube videos on the problem. Sometimes, if you're willing to wait, you can put in an order on a site like digikey and you'll eventually get some at list price. Bu
120.
by phoehne 4y ago
When you add performance as a consideration, for just straight computing tasks, you might also want to consider a refurbished or used 1 liter PC. They'll have a 6th or 8th generation intel CPU and power draws usually below 60 watts.

More ›