dabockster

Because the only place where "free" AI is viable is when the model is running on your own computer. It's not sustainable at all, from a business perspective, to let millions of connections absolutely slam a huge AI model 24/7 for free - even with usage limits and throttling.

https://en.wikipedia.org/wiki/Jevons_paradox

Better to write a solid inference engine, tell people to run it themselves, and externalize that cost.

Okay, so how would you market that to someone that's non tech savvy? What's the sales pitch?

They are all impressive tech, but not actual stuff you can sell or distribute until you can answer those questions.

> Mozilla said they wont survive

Entirely their fault, tbh. Mozilla's C suite has knowingly enriched themselves off this money for over 15 years now. If they were serious about surviving, they would have found alternative funding sources a long time ago.

Firefox isn't a true project. It's Google paying off someone to make Chrome appear not to be a monopoly at first glance.

Chrome is the "best" because all of the other browsers continue to fail at real world marketing. The best ads and marketing continue to be real life stuff - billboards, bus signage, people handing out flyers, etc etc etc. You can't just hype a browser on social media or web forums, and you can't hype it solely to those who are tech savvy.

A solid example of this right now is all of the Mullvad VPN ads I've seen on the Seattle Light Rail lately. Google used to have ads everywhere for Chrome. The only time I saw Firefox stuff was the rare t-shirt at a tech conference.

Or by people like me with LM Studio, a lightweight GGUF from Hugging Face, and maybe some kind of vector database MCP tool.

> Stack Overflow isn’t dead because of AI. It’s dead because they spent years ignoring user feedback and then doubled down by going after respected, unpaid contributors like Monica.

They also devolved into a work friendly variant of 4Chan's /g/ board. "Work friendly" as in nothing obviously obscene, but the overall tone and hostility towards newcomers is still there (among other things).

The M series chips are optimized at both the assembly language and silicon (hardware) levels for mobile use. X86 is much more generalized.

No, the problem is that you’re thinking of this like Spock - purely logical. Humans aren’t logical. We absolutely trust/distrust each another based on appearances and mannerisms. This is not limited to RMS.

People are prejudiced, plain and simple.

I actually filed a complaint with the WA Attorney General over that. My older parents got hit by that exact process. So there’s at least a public record complaint that Google is now actively blocking cybersecurity technology (because that’s what adblocking software ultimately is).

You can self-sign keys. Secure Boot has had a mechanism for that already for years (mok on Linux).

All of those ran really slow compared to Android versions at the time, or their dev tooling sucked. The only one I really enjoyed using was Sailfish, and even they had to implement an APK compatibility layer. So for the average consumer, what’s the benefit to using that over straight Android?

> easier said than done

This is true for both the engineering and business sides. Cyanogen’s failure showed that it ultimately doesn’t matter how good your software product is if your business side of things is poorly run. Same with the Pebble smartwatch - amazing product, terrible back office.

One could argue that those “cheap” devices are ewaste from the beginning, and customers needing lower cost mobile devices should be buying more expensive ones used or refurbished.

You have to pair program with them. They don't work well at all when fully autonomous.

But they do work for pair programming. Which explains a lot of the tech layoffs we've been seeing.

Pair programming with an agent is so powerful and underrated.

Open vs closed source software

Open vs closed web

Both of the open options are nice on paper, until you realize that they were largely being pushed by big tech. The same big tech that is scraping them hard for AI training, and possibly have been using them silently without respecting licensing agreements for an even longer period of time.

"Open" only works when everyone plays nice and fair. This whole kerfuffle of AI scraping has shown that this isn't happening, and likely hasn't happened for a long time if at all.

That's good to know. I'll have to look up those licenses another time.

I've built stuff in Qt before and it's great, but it's super expensive if you want to use it for proprietary work.

AI agents are making it really easy to develop in C/C++ and produce rock-solid memory safe code. Really looking forward to a new renaissance for desktop programming!

I've been using a lot of the Chinese open source models like R1 and the Qwen coder series. I've also been trying some community mods of them from HuggingFace.

I was using a combo of Ollama + Roo Code for the front/back end but Ollama is kind of dumb when it comes to memory overload protections. I've since switched to LM Studio and it has a very annoying hard timeout of 2-3 minutes on its API server. Local isn't perfect right now, but when it does run you can see the potential as clear as day.

The Chinese models like Deepseek R1 and the Qwen Coder series have been way better at coding than any of the western models. And they're open source, too.

32 billion sized R1 and Qwen models blow GPT, Gemini, and Claude out of the water for coding tasks usually. And they can be made to run on a standard gaming PC off the shelf from somewhere like Best Buy or Costco. They will run really really slowly, though, but they will run.

Paramount+

* Star Trek everything (except Section 31 - worse than Star Trek V)

* South Park specials

* Frasier sequel

* Beavis and Butthead revival

Before someone comes in and says something moronic (because physical media isn't a VC funded B2B SaaS business), I'd like to remind everyone that Redbox was going strong until Chicken Soup for the Soul bought them. Even in the late 2010s when streaming was really taking off, you could often go down to your local corner store and rent a DVD or BluRay out of a Redbox machine.

The only reason people went to streaming in the first place was that Netflix was cheap and, arguably, a heavily subsidized monopoly in the streaming space.

So yeah, some kind of DVD/BluRay business could very well work in the US and Canada.

Yeah I know about Model Context Protocol. But it's still only a small part of the AI puzzle. I'm saying that we're at a point now where a whole AI stack can run, in some form, 100% on-device with okayish accuracy. When you think about that, and where we're headed, it makes the whole idea of cloud AI look like a dinosaur.

LLMs in general are going to make C/C++ development viable again for the masses. A big reason everything went to web frameworks was memory safety - even big frameworks like Electron and React Native are memory safe overall. Writing low level code by hand was a migraine experience, even for experienced devs. AI changes this relationship entirely, even if you just use it as a pair programmer. I've had even small 8b models correctly call out memory unsafe code and suggest fixes. Larger models are even better.

Local LLMs? A new renaissance. All that power without having to pinky swear with a cloud provider that they won't just take your generated code and use it for themselves.

Expect to see some awesome Windows and Mac apps being developed in the coming months and years. 100% on device, memory safe, and with a thin resource footprint. The 1990s/2000s are coming back.

> diminishing returns

I think this is a hard kick below the belt for anyone trying to develop AGI using current computer science.

Current AIs only really generate - no, regenerate text based on their training data. They are only as smart as other data available. Even when an AI "thinks", it's only really still processing existing data rather than making a genuinely new conclusion. It's the best text processor ever created - but it's still just a text processor at its core. And that won't change without more hard computer science being performed by humans.

So yeah, I think we're starting to hit the upper limits of what we can do with Transformers technology. I'd be very surprised if someone achieved "AGI" with current tech. And, if it did get achieved, I wouldn't consider it "production ready" until it didn't need a nuclear reactor to power it.

> We believe that all agents will long more like this in the future - long running, asynchronous, more autonomous. Specifically, we think that they will:

> Run asynchronously in the cloud

> cloud

Reality check:

https://huggingface.co/Menlo/Jan-nano-128k-gguf

That model will run, with decent conversation quality, at roughly the same memory footprint as a few Chrome tabs. It's only a matter of time until we get coding models that can do that, and then only a further matter of time until we see agentic capabilities at that memory footprint. I mean, I can already get agentic coding with one of the new Qwen3 models - super slowly, but it works in the first place. And the quality matches or even beats some of the cloud models and vibe coding apps.

And that model is just one example. Researchers all over the world are making new models almost daily that can run on an off-the-shelf gaming computer. If you have a modern Nvidia graphics card, you can run AI on your own computer totally offline. That's the reality.

I can replicate almost all of cloud based vibe coding with the Roo Code extension for VS Code and a Qwen Coder model file from HuggingFace, all running on my gaming computer. And I usually get better code quality than almost every cloud based coding agent I've tried.

That right there says a lot about where we are with AI now. We have AI models that can run on off-the-shelf name brand computers (think a Dell or HP tower at a Best Buy) with about the same memory footprint as a few Chrome tabs. Yeah, it'll be slow and stutter like crazy at only a few tokens per second of output. But it still works in the first place.

And that's the important part here. The fact that local on-device AI is here, it can produce output superior to cloud platforms in general, and it's only going to get better over time (especially zero shot training techniques - look it up). These firms' products are expensive developer front ends to some of the most expensive and proprietary models on the market - that they have to rent themselves. Why use them as a middle man when I can either go to the model provider directly or bring my own model?

Also, this trend of fully forking VS Code needs to stop. These coding apps could have been VS Code extensions, which would let devs have access to all of Microsoft's proprietary tooling. And the apps feel vibe coded themselves overall.

Especially if you only deal in government issued currency and refuse to use crypto. The US dollar is the world's reserve currency for a reason - we have tons of opaque and invisible systems in place to make sure it's rock solid. Systems that other countries either don't have entirely or rely on a series of privately owned systems that feel like chewing gum and string.

> Whats wrong with verifying the ID?

- IDs usually contain enough information to totally steal a person's identity (beyond just their name and mailing address).

- IDs usually have secret security mechanisms beyond what's publicly known for a government to verify if an ID is real or not.

- 3rd party business systems can only verify the publicly known security stuff. And because that security stuff is public, it can get faked easily.

- IDs are super easy to fake.

- The only entity that can totally verify an ID is a government, because you'd have to verify the secret security mechanisms as well.

And there's a million reasons why you wouldn't necessarily want a government to verify an ID for a private business transaction.