when running models locally, I presume the models are trained and the weights and stuff are exported to a “model.” For example Meta’s LLama model.

Do these models get updated, new versions released? I don’t quite understand

wow 10 months flew by since this was posted and since then the United States had a surprise privacy bill that is bipartisan that sort of addresses the issues you and I mentioned. https://www.washingtonpost.com/technology/2024/04/07/congress-privacy-deal-cantwell-rodgers/

This bill was proposed around the same time the TikTok ban was announced. I speculate that law makers had a difficult time framing the arguments against TikTok when “the data of citizens have no protections so there was no easy legal grounds to forbit the likes of TikTok to harvest it”

From what I’ve heard, this bill is pretty good. I need to educate myself more on it, however.

For backup and sync I use Syncthing. I can specify which folder on which devices I want to sync to which folder on the server.

I use a folder based gallery on my phone so when I move stuff around on my phone (or on my server) it gets replicated on all my devices.

I also have a policy to sync specified folders (and subfolder) with my family’s devices. No more " hey can you send me all the pics from the XYZ trip"

We take a trip. Make a subolder for that trip in a shared folder dump all our pictures there, get home and open the folder on the computer and prune together.

simply put, programming is glorified automation. There are jobs where the process that needs automating makes money.

Debian has the advantage of not using snapd like Ubuntu does. You have to not only remove snaps but also instruct the package manager not you pull in snaps as dependencies and not to favor snap packages.

I have fond memories of Ubuntu being my first distro many years ago but pushing snaps onto users to compete with flatpak is a nuisance.

I don’t think I am well positioned to answer that question given my experience. Ill give it my best.

I believe the advantage of more abstraction of gRPC was desireable because we can point it at a socket (Unix domain or internet sockets) and communicate across different domains. I think we are shooting for a “microserves” architecture but running it on one machine. FFI (IIRC) is more low level and more about language interoperability. gRPC would allow us to prototype stuff faster in other languages (like Python or go) and optimize to rust if it became a bottleneck.

Short answer is, we are able to deliver more value, quicker, to customers (I guess). But I don’t know much about FFI. Perhaps you can offer some reasons and use cases for it?

At work, we started the c++ migration to rust doing the following:

1. Identify “subsystems” in the c++ code base
2. Identify the ingress/egress data flows into this subsystem
3. Replace those ingress/engress interfaces with grpc for data/event sharing (we have yet to profile the performance impact of passing an object over grpc, do work on it, then pass it back)
4. Start a rewrite of the subsystem. from c++ to rust
5. Swap out the two subsystems and reattach at the grpc interfaces
6. Profit in that now our code is memory safe AND decoupled

The challenge here is identifying the subsystems. If the codebase didn’t have distinct boundaries for subsystems, rewrite becomes much more difficult

hey, that’s what the internet is for; information sharing :)

for the dummies (like me) that can’t read the room, especially online, a sarcasm tag /s goes a long way 🙃

you sound like a Microsoft engineer ;)

hahaha good point.

That colleague, keep in mind is a bit older, also has Vim navigation burned into his head. I think where he was coming from, all these new technologies and syntax for them, he much rather prefers right clicking in the IDE and it’ll show him options instead of doing it all from command line. For example docker container management, Go’s devle debugger syntax, GDB. He has a hybrid workflow tho.

After having spent countless hours on my Vim config only to restart everything using Lua with nvim, I can relate to time sink that is vim.

Had a distinguished collegue (from the Bell Lab days) say to me recently:

“IDEs take up a lot of RAM on my machine. Vim takes up a lot of squishy RAM in my head. I need squishy RAM to hold info relevant to problem solving, not options available in my tool chain.”

As a former Vim user myself, I have to say I really dislike screensharing with coworkers who use Vim. They are walking me through code and shit pops up left and right and I don’t know where it comes from or what it is I’m looking at. Code reviews are painful when they walk me through a large-ish PR.

These days, I tend to bring my vim navigation/key bindings to my IDE instead of IDE funcs to Vim. Hard to beat JetBrains IDEs, especially when you pay them to maintain the IDE functionality.

code is just text, so code editors are text editors.

What sets IDEs apart are their features, like debugger integrations, refactoring assists, etc.

I love command line ± Vim and used solely it for a large portion of my career but that was back when you had a few big enterprise languages (C/C++, Java).

With micro services being language agnostic, I find I use a larger variety of languages. And configuring and remembering an environment for rust, go, c, python etc. is just too much mental overhead. Hard to beat JetBrain’s IDEs; now-a-days I bring my Vim navigation key bindings to my IDE instead of my IDE features to Vim. And I pay a company to work out the IDE features.

for the record, I am in the boat of, use whatever brings you the greatest joy/productivity.

How exactly does this make apple look bad?

sincere question, I am layer 3&4 network stack developer so I am quite out of the loop for mobile apss/web tech

don’t insult children like that.

Yeah I was not a fan of paying for Spotify and them cramming ads of podcasts down my throat when I wanted to listen to music. Plus their shuffle is a joke. Music discovery was pretty sweet though

Hard Fork: for keeping up with the biggest tech news. they do dissecting of potential impact if stuff.

Lex Fridman: He interviews really interesting subjects. I’ll listen to subjects I’m interested in based on who they are or the subject matter they are an expert in. Lot’s interesting tech folks. My favorite episode so far is with John Carmack: Doom, Quake, VR, AGI, Programming, Video Games, and Rockets. Epsidoe is 5 f***king hours but broke it up into several sessions and Carmack is so good in articulating, it flew by.

Huberman Lab: before software I liked biology and medicine. I like these occasionally because I get to learn how systems outside of software/hardware work. These I will watch/listen in a sitting as one would to a movie. It demands your attention to follow along. (I don’t like when doctors have podcasts with all the “alternative medice” BS. But Huberman is an active researcher at Stanford and in charge of a lab that cranks out sweet research. Def credible dude and very methodic and tries to rule out bias).

You can still buy a lifetime licenses of office but you have to buy it from 3rd party sellers and then validate the license with M$. Example Deal..

I bought 2 of them and also saved the install binary to have office suite.

I use libreoffice personally but I have family members that get frustrated when they cannot find the same formatting options

newsblur