MattRix

5 ms·

31.
by MattRix 5mo ago
The idea that users will switch because a platform has ads has never played out in the past. The average user just doesn’t care.
32.
by MattRix 5mo ago
For now I just mean something simple like playing whichever note you press on the piano/fretboard.
33.
by MattRix 5mo ago
It’s a cool idea, but is there no way to hear the actual notes?
34.
by MattRix 5mo ago
It’s not a 3B model, it has 3B active parameters. The full model is much larger.
35.
by MattRix 5mo ago
It’s not a micro model at all, it requires 40gb of VRAM. The 3B is just the active parameters.
36.
by MattRix 5mo ago
Have you considered that maybe the experts in a field are actually correct about that field?
37.
by MattRix 5mo ago
Yeah it has a confusing clickbait title that gives the wrong impression of what the article is about. His point is that the bottleneck to making a complex app like photoshop is architectural rather than just writing basic code… and he argue
38.
by MattRix 5mo ago
The inference is absolutely not sold at a loss, at least not when paying API prices (the subscriptions are less clear). The reason frontier model companies aren’t profitable is because training the models is so costly, not inference.
39.
by MattRix 5mo ago
I mean everyone knows they need more compute. That’s not a secret or up for debate at all. They are maybe the fastest growing company in history.
40.
by MattRix 5mo ago
In my experience, old normies love AI, young normies do not.
41.
by MattRix 5mo ago
surely you can see how this is a rare specific situation you encountered and not some actual systemic mcdonald’s problem (there are LOTS of reasons to complain about mcdonald’s but this seems like a strange one)
42.
by MattRix 5mo ago
It’s not that you’re seeing old prices per se, but rather that it’s easier to make two IAP items (one with regular price, one with sale price) than it is to actually change the price of a single IAP. That’s why you see multiple items.
43.
by MattRix 5mo ago
When I look at the in app purchase prices on the App Store page, it tells me what they’re for? (1 month vs 1 year). You can usually assume that if you see a different price for the same name item, that it’s a sale price.
44.
by MattRix 5mo ago
Yes, it’s a significantly less powerful model, that’s why.
45.
by MattRix 5mo ago
This stuff has been happening in Apple software since well before the AI coding stuff came along.
46.
by MattRix 5mo ago
no thanks
47.
by MattRix 5mo ago
Everything I’ve heard about the company tells me they are obsessed about exponential growth. It might seem bad to make a change that loses you 10% of your users, but if those are your least profitable users and the rest of your userbase is
48.
by MattRix 5mo ago
Not sure why that’s bleak? It’s on top of the regular vacation days you get every year. I don’t know of any other company (at least not here in Canada!) that would give you an entire paid month off.
49.
by MattRix 5mo ago
That deal is a win-win for Google. If they develop a better coding model than Anthropic and beat them at coding, then they win. If they don’t, they still win by making a ton of money from Anthropic long term.
50.
by MattRix 5mo ago
That is literally their point.
51.
by MattRix 5mo ago
Only if you didn’t read the article…
52.
by MattRix 5mo ago
How can you say “without bringing in proof” when there is literally proof in the article?
53.
by MattRix 5mo ago
The problem isn’t that the tasks are impossible to solve, it’s that they’re underspecified and/or impossible to solve consistently (ex. because a test is expecting the solution function to have a specific name that wasn’t specified in
54.
by MattRix 5mo ago
They mention this in the article. This is why private (non public) benchmark tasks that have been made from scratch are necessary.
55.
by MattRix 6mo ago
I don’t see why this would happen when the modern models already use MoE, which gives them most of the benefits of having specialized models.
56.
by MattRix 6mo ago
The open models of similar scales (ex. the new 1T deepseek model) are a fraction of the cost per token, so I don’t see how that can be the case. Inference is profitable, it’s the training that makes it unprofitable.
57.
by MattRix 6mo ago
Yeah but in a discussion about technology it’s a little silly. It’s like someone complaining about their phone and then finding out they still use a Nokia.
58.
by MattRix 6mo ago
Overwatch also has “kill cams”, which basically create an entire alternate game state to show you how the enemy killed you, and they have the “Play of the Game” system that replays the coolest moment of the game at the end. It’s impressive
59.
by MattRix 6mo ago
Log out and log in again? That usually fixes these kind of issues for me.
60.
by MattRix 6mo ago
From what I’ve seen, once people start using these, they will do the font size thing. Then all your changes go through the same interface.

More ›