dabockster

The issue then becomes "well why don't they just go back to a Teamspeak server? they can self host it!"

But we're forgetting there that the average person online is not a dev. The most they usually know is how to point and click on something. Which also means they usually don't know how to spin up a Linux machine/VM somewhere and install their own chat server.

Discord is popular because it lets almost anyone on Earth point and click to create a chat "server". If someone can figure out how to do that (eg cPanel), you can absolutely break their moat.

That’s a fair point. But I’ve learned from my own life experience to factor in things like Hanlon’s Razor and the Dunning-Kruger Effect when it comes to technology anymore. Especially post-Covid tech.

In this case, while it’s totally possible for this sort of data to still exist somewhere, I think the chances of it surfacing again in any accessible format are rare - purely because of the overall stupidity of the system. Keeping data that “alive” for decades is a skill in itself that seems to only happen in a heavily subsidized “perfect” economic times (at least to the outside observer). Once the going gets tough, there isn’t really any business value to saving the data and it likely gets deleted.

The largest moat I’ve seen with Amazon and e-commerce in general has been the logistics aspect of it. Because even when 90%+ of the goods sold are junk, the remaining 10% are lower volume niche items or hard to find in retail stores for various reasons. E-commerce can absolutely unleash underserved and redlined areas, for example.

So honestly, I think we haven’t even gotten close to the rock bottom point where people will actually start cancelling things like Prime. Especially while the competition still won’t sell their goods to “dirty” areas. And then people wonder why Amazon and Walmart are so big.

The moat, imo, is mostly the tooling on top of the model. ChatGPT's thinking and deep research modes are still superior to the competition. But as the models themselves get more and more efficient to run, you won't necessarily need to rent them or rent a data center to run them. Alibaba's Qwen mixture of experts models are living proof that you can have GPT levels of raw inference on a gaming computer right now. How are these AI firms going to adapt once someone is able to run about 90% of raw OpenAI capability on a quad core laptop at 250-300 watts max power consumption?

Yeah, how much profit will they make if they're able to go for-profit? Revenue doesn't tell me anything.

> nothing about tech is genuinely inevitable

This reminds me of when everyone was saying that "everything on the internet is written in ink" - especially during the height of social media in the 2010s. So imagine my surprise in the first half of the 2020s when tons of content starts getting effectively deleted from the internet - either through actual deletion or things like link rot. Heck, I literally just said "the height of social media" - even that has pulled back.

So yeah, remember that tech ultimately serves people. And it only happens so long as people are willing to enable it to happen.

If not racism, then a real life Mean Girls at a very least. But in this case, Regina George is your boss.

It's your company's problem that your legally valid hiring opinions are being overridden by other departments to cut costs. Especially when there are tons of American citizen CS grads from non-target schools that would probably take the $20k hit (or even more) to get their foot in the door. Your company is acting absolutely stupid.

> If I can hire someone remotely in San Antonio, why can't I hire someone remotely in São Paulo as a result?

Because the person in Brazil would be able to steal your IP and you'd have no financial, legal, or emotional recourse since they're across an international border. Big tech is continuing to find this out the hard way when countries like China and India somehow clone their products in record time.

By hiring in the same country you reside in, you have access to your legal system to protect you and your business interests. And it can also protect your own ego as well (if you have one).

And for the USA specifically, it is also really about the people here. Tech is a surprisingly human industry, and Americans have the humanity and grit to pull it off.

> pay a lower salary

Nothing is stopping tech from doing this right now to American workers. Remember that those higher paying firms do false positive hiring - they reject way more people than they pull in. If you do the reverse and adjust your operations to account for more workers that "aren't good but aren't bad either", you'd probably perform even better than offshoring.

I "tolerated" GNOME after the GNOME 3 release, and used MATE for years before stuff started breaking on it. But KDE 6 has a nice vibe of something between Windows and Mac OS. Still really rough in a lot of spots, but you should definitely try it!

You're correct in thinking of Fedora that way - that's my take on it too. But Windows has a lot of glue and infrastructure that we all take for granted that Linux hasn't really implemented yet. Like how calling a Win32 function, in some cases, integrates functionality found in both the Windows desktop as well as backend processes - stuff that would be several lines/files long in Linux, and would likely need conditional logic for GNOME and KDE/Qt desktops. Or something that's provided by a page in a Windows Control Panel applet that you'd need to download someone's hobby app to do on Linux (emphasis on hobby app - definitely not "Enterprise" grade).

You're correct in thinking of Fedora that way - that's my take on it too. But Windows has a lot of glue and infrastructure that we all take for granted that Linux hasn't really implemented yet. Like how calling a Win32 function, in some cases, integrates functionality found in both the Windows desktop as well as backend processes - stuff that would be several lines/files long in Linux, and would likely need conditional logic for GNOME and KDE/Qt desktops.

Even Fedora, which is really rock solid and more "neutral" compared to other distros (lookin' at you, Ubuntu Snap Store), there's a LOT that gets left to either terminal commands or config files when compared to Windows. And even as a dev, I don't want to have to RTFM literally every time I want to change a single setting. Windows solved this 30+ years ago, and it's honestly stupid that the Linux community/field can't even seemingly comprehend this in almost 2026 now.

The WA AG's office doesn't typically post that stuff in a searchable format online. You'd have to run a public records request. But I guarantee it's there in some form.

> Take the core technology and just optimize, optimize, optimize for 10x the cost/efficiency.

This is what really grinds my gears about American AI and American technology in general lately, as an American myself. We used to do that! But over the last 10-15 years, it seems like all this country can do is try to throw more and more resources at something instead of optimizing what we already have.

Download more ram for this progressive web app.

Buy a Threadripper CPU to run this game that looks worse than the ones you played on the Nintendo Gamecube in the early 2000s.

Generate more electricity (hello Elon Musk).

Y'all remember your algorithms classes from college, right? Why not apply that here? Because China is doing just that, and frankly making us look stupid by comparison.

Jumping from this for visibility - LM Studio really is the best option out there. Ollama is another runtime that I've used, but I've found it makes too many assumptions about what a computer is capable of and it's almost impossible to override those settings. It often overloads weaker computers and underestimates stronger ones.

LM Studio isn't as "set it and forget it" as Ollama is, and it does have a bit of a learning curve. But if you're doing any kind of AI development and you don't want to mess around with writing llama-cpp scripts all the time, it really can't be beat (for now).

I wouldn't recommend using anything that can transmit data back to the CCP. The model itself is fine since it's open source (and you can run it firewalled if you're really paranoid), but directly using Alibaba's AI chat website should be discouraged.

It's a good habit to build now in case AGI actually happens out of the blue.

LLMs can automate some tasks (especially basic coding), but with some major caveats:

- They have rapidly diminishing returns on higher contexts. They are very easy to overload with user information. You can post-train them, but that’s too much into actual computer science to be useful to the average office worker. They can work well as a co-worker assistant in a few cases, but they really can’t replace humans long term.

- LLMs only really work well when given the ability to call on-device tooling. Which, with a cloud system like Copilot, is going to be super tame and underwhelming because of lawsuits and various business deals. Tool calling is only really agnostic when you’re running the model yourself on your own device.

- On the topic of on-device models, there’s also the fact that AI provider companies have been caught evading things like robots.txt and causing so much crawling activity that it effectively becomes a DDoS attack. On-device AI doesn’t solve this totally, but most people likely won’t be pushing their gaming GPUs to the absolute limit 24/7 to constantly hit websites with crawl requests.

OneDrive is a love/hate relationship for me. Yeah, it’s invasive. But it’s private [1], secure, and cheap (relatively) file storage. And very few of its competitors integrate into Windows Explorer in the way OneDrive does (even though the APIs are all there for everyone to use).

[1] OneDrive privacy confirmed by my own consumer complaint to the Washington State Office of the Attorney General over ambiguities related to AI training and general privacy in Microsoft’s documentation. Both my complaint and Microsoft’s official response are subject to public record and can be legally actionable in the future.

I’ve been playing around with Fedora KDE more lately and it’s made me realize how vertically integrated Windows is. Like there’s all these edge cases (too many to list) that Windows has taken care of years ago, whereas on Linux you’re usually downloading some 3rd party app that feels rough when installed. Or it’s something the kernel expects a driver to implement, but Windows handles it natively in Microsoft’s code.

When you think of it in a supply chain sense, Linux is one giant outsourcing operation and a whole bunch of “not my problem” project management styles.

That wouldn't work against a really sophisticated attacker. Especially for something that's clearly being maintained for free by one overworked person in their spare time (yet again).

You'd need some kind of offline verification method as well for these widely used infrastructure libraries.

For this kind of infrastructure, some kind of real world verification may be necessary as well. Like having human ran phone verification (not AI, an actual call center) using information intentionally kept offline for securing more widespread and mission critical packages.

They can't pwn what they can't find online.

I haven't published anything to npm in over a decade. But if you still have access to git, a cli, or a browser where the login is cached and you can access it, you should do so and either take the code down or intentionally sabotage/break it.

If you're running this kind of infrastructure online these days, you have every right to require payment somehow. Don't work for free.

Here's something I generated in my coding AI for Powershell:

`Get-ChildItem -Recurse | Select-String -Pattern '_0x112fa8' | ForEach-Object { $_.Line.Substring(0, [Math]::Min(80, $_.Line.Length)) }`

Breakdown of the Command:

- Get-ChildItem -Recurse: This command retrieves all files in the current directory and its subdirectories.

- Select-String -Pattern '_0x112fa8': This searches for the specified pattern in the files.

- ForEach-Object { ... }: This processes each match found.

- Substring(0, [Math]::Min(80, $_.Line.Length)): This limits the output to a maximum of 80 characters per line.

---

Hopefully this should work for Windows devs out there. If not, reply and I'll try to modify it.

> Many people also don’t vendor their own dependencies, which would slow down the spread at the price of not being instantly up to date.

npm sold it really hard that you could rely on them and not have to vendor dependencies yourself. If I suggested that a decade ago in Seattle, I would have gotten booed out of the room.

Or they worked for a small number of large firms that have reigned in referrals.

Like how a lot of people moved to Seattle for its tech scene but then realized that all we really have for large scale employement is Amazon, Microsoft, and Boeing. All the other firms are lower headcount satellite offices.

Same. Or maybe something using PHP just as a basic templating engine (like how people used to use it). Something that can be copy/pasted to a dime a dozen cPanel powered shared webhost at a moment's notice.

Coffee is all about location, location, location. Honestly, you could sell Folgers at $1 a cup at the right location and you'd still make tons of money.

1. Disable Face ID/fingerprint unlocks and go back to using a alphanumeric pass code (letters and numbers). It's fine once you're actually in the phone, but you absolutely don't want to unlock your phone from the lock screen with biometrics.

2. Watch what kind of app permissions you grant to what apps. This is way easier to do on iPhones vs Android phones.

3. Understand that Google Apps (gApps) on Android phones are essentially a rootkit anymore. Especially with that recent news that Google is going to try to end sideloading APKs. Apple isn't that much better but they're more "out of the way" when you try to work.

4. Apple's SDK for its devices is more private than Android in certain ways. The IceBlock app's explainer is really good about this: https://www.iceblock.app/android

See if your community can establish some kind of mesh network connection to someone outside of the country. You'd have to trust everyone on it with your life (quite literally in this case), but it would work.