QwenGlazer9000

What Quants do you run?

Earlier in march/aprile, there was a regression in Claude code.

Unintentional tbf.

Last time they were called out, it was a regression in Claude code itself.

At least that's their explanation. Either way, it wasn't a good look for "vibecoding" but it got brushed over.

You can find everything on X, from dumpster fires to peak intellectual, everything in between.

If you look at some of the other comments replying to my original comment, you'll get the context that the project started out as a hobby project, purely for fun.

Call me chronically online or whatever, but I've looked at his twitter, and I really don't like his attitude. What specifically grinds my gears is one of tweets where he looks down on programmers who rewrite and program just for the fun of programming.

It's so hypocritical of him, when his entire project started off as just for the fun of programming.

The cherry on top is this was all a part of the whole "agentic shift". I'm fine with someone using AI, and he seemed to use it semi responsibly (at least last I checked), but his hubris and hypocrisy make me not care about the project anymore.

I'm glad we have an alternative to ladybird, since I haven't been a fan of the direction of the project for a bit.

My org also adopted openspec, massive downgrade in workflow. Every change is so much more tedious now.

Two actually*

Mistakes happen but this same buffoon got elected twice. That's no fluke, deep down the american population likes him enough to give him two chances.

Yeah no it does not pay for itself just comparing to cloud. Not at these prices at least, people far richer than you or I buy these things wholesale, no scalper, bought a significant amount at cheaper prices, and are wired up the ass with VC money.

The premium is not having your million dollar prize and career stolen by billionaires.

> But you see, now you'll have more time for the actual important things!

> Like what?

> Cleaning shit out of clogged toilets!

It's because anthropic vibemathed it. I forgot the name but some other guy is working on a handwritten version of it and I bet it'll be more than just 1 magnitude faster.

After looking at some of his recent interviews, he's acknowledged his previous predictions that 1. He was naive enough to think the investment won't keep coming despite lack of profitability and 2. He acknowledges that AI in it's current form is a billion dollar industry, not a trillion dollar one, but also not zero.

Definitely less extreme than he used to be.

Was it?

Given the timing, I think they A. shat their pants since Deepseek flash just came out with insane pricing before the price hikes, and B. Anthropic is really struggling in model tiers below opus.

It was smart for them to cut prices regardless of whether they had 80% efficiency gains or not

For most of the use cases I would need an app for, koreader's plugin system works just fine for it. You'd be surprised the gems you can find for koreader.

It's that good because the language model they use is expensive I'm pretty sure. I haven't seen many details on what they use, but it's apparently a loss leader for them and already can no longer be used for free.

Meanwhile google translate is probably served on the cheapest model they could muster.

Funny how that works out, because AI is the solution to a problem significantly worsened by AI. The web didn't used to be this terrible before chatGPT, now it's just so much slop. At this point AI is required to sift through all the bullshit caused at least in part by AI.

You know now that I think of it me too. I absolutely never used the free google one because I never trusted it (you know what they say about free things).

If I'm paying for it through Kagi, I could at least trust their incentives are aligned with mine instead of whoever is paying them the most.

I just wish it were easier to keep up with which models are good for what. Especially with the introduction of the research assistant, I have no idea whether I should be using that or one of the standalone models. And don't even get me started on trying to pick the best open model, it's always fluctuating.

I wouldn't say it's better, it's a great alternative though. I find that for some tasks I much prefer doing a search myself as I may not trust the sources it's pulling from, but for others it's not only adequate but much faster.

I think it's a great complement to human lead search.

Everyone else mentioned the app support which is true, but for graphene specifically, they do not like desktop Linux at all because they don't like its security. They would much rather build on AOSP than desktop linux.

That shit used to be frowned upon, now it's celebrated!

From my experience they have been getting less repairable. Look at recent thinkpads.

My headcannon is that the only reason the T14G7 is as repairable as it is, is due to the whole right to repair movement, including framework.

> the neat part is that you don't have to, millions will do it for you!

You're not wrong. We're already seeing something similar with all the outages and buggy software updates these days.

We little choice in the matter, you will have ze (software) bugs!

Because they are still dealing with memory safety issues to this day. They ported it to rust, completely introduced a bunch of bugs, went through all that effort, just for the codebase to be not much safer.

I'm actually basing my opinion off the blog post, specifically the code. While they do get a few changes for free with the port, most of the code looks identical. And the number of unsafe statements certainly agrees.

And now they're going little by little, playing whack a mole with seg faults... My point is that the step they're at right here is the important part, and that mechanically converting the entire codebase to rust was an optional step when incrementally rewriting would've worked just as well if not better.

Being "only" 165k USD doesn't mean anything, because you didn't do a complete port, you did a port in a trench coat. If they went for the more targeted module by module approach, and did a clean and proper rewrite, LLM assisted or not, they would reach their end goal much less turbulently and without the effort of that initial frankenport.

I didn't say don't use LLMs. In fact, I'm quite certain they could help quite a bit.

But I don't think they did it in a way that actually provided a meaningful gain. My point is that the state of the codebase right after the code is just a worse version of the original zig code with few of the rust advantages. Going from that Frankenstein rust code to actual memory safe rust is a similar leap to going from zig.

It does matter, because that's the entire point of the language. You know, memory safety?

The vibe coded rust port did not get them any closer to full compiler verified memory safety. They still need to go file by file, bit by bit, and make it memory safe, at which point, why not just do that from zig.

That's good to hear. Having a proper protocol is the hardest and most important part. If everyone agrees on the protocol, different implementations can exist, forks can compete, and we get all the benefits of such without network fragmentation.

> He does make a good point that most of the progress so far have not been material in the sense of providing real positive outcomes for ordinary people.

Even worse, it has made a significant number of people's lives worse. Some major, like layoffs, some less major, like PC prices, some are just annoyed by AI slop.

And then Dario himself wouldn't stop yapping about white collar work being gone in X years...

Problem with that is that it's not proper rust code, it's some kind of franken zig style rust port.

The actual hard part is getting it into idiomatic, safe rust, and I don't believe the LLM port makes the full transition any easier than doing things the old fashioned way: a dual lang code base like Linux.

I had a lot of hope for reticulum, but it's a one man project with a near burned out maintainer, taking on an absolutely monstrous task.

I'm now cautiously optimistic. This or something similar is likely the long term future of mesh networks, but in the short to medium term, meshcore is the way to go.