aabhay

8 ms·

121.
by aabhay 11mo ago
Great user name for sure!
122.
by aabhay 11mo ago
I think the phrase is apt, actually, but it's not perfect: 1. We have an embedded agent that can read, edit, organize, and take actions on your files. This means it can read almost any media type, which is why it's "cursor fo
123.
Launch HN: Poly (YC S22) – Cursor for Files
66 points by aabhay 11mo ago|68 comments
124.
by aabhay 11mo ago
Then switch devices. Oh but you might need to find an alternate solution for your iMessage groups. And your apple watch. And your family icloud plan. And the apple ebooks you bought. And your airpod maxes. And your protection plan. And the
125.
by aabhay 11mo ago
Except that this same team built a similarly named software package for Nvidia GPUs as well. It’s bright researchers doing what they do best if you ask me.
126.
by aabhay 11mo ago
They are, but the problem is that shifting an organization whose lifeblood is yearly hardware refreshes and chip innovation towards a ship-daily software culture is challenging. And software doesn’t “make money” the way hardware does so it
127.
by aabhay 11mo ago
There’s also the fact that both Wikipedia and LLMs are non-stationary. The quality of wikipedia has grown immensely since its inception and LLMs will get more accurate (if not explicitly “smarter”)
128.
by aabhay 11mo ago
You don’t need to cargo install anything. You just need cargo itself, which is linked on the main page. Once you have that, here’s an example google search that gives you all the info you need to run the project (hint: `cargo run`) https:&
129.
by aabhay 11mo ago
For anyone that’s curious, the underlying Torch library is also a joy to work with, as are the many other torch bindings. For example, Rust has tch and Burn which both work with libtorch. PyTorch of course has the benefit of being dynamical
130.
by aabhay 11mo ago
The latency of 5s for the basic tier search request is very confusing to me. Is that 5s per request or 5s per 1k requests? If it is indeed 5s per request that seems like a deal breaker
131.
by aabhay 1y ago
Nuxt is amazing, but it was recently bought by vercel, so we’re all anticipating slow enshittification of it to match next.js
132.
by aabhay 1y ago
That was against the boy scouts but the reverse lawsuit hasn't been filed against the girl scouts to my knowledge. That said, law in the US and one’s opinions on what constitutes discrimination are different things.
133.
by aabhay 1y ago
So what about the girl scouts, is that also discriminatory?
134.
by aabhay 1y ago
OpenAI has always had the stance of “commercialize narrow AI that is research aligned with AGI development”. In fact they used to ask this as an interview question — “should we commercialize narrow AI or aim to put all resources into AGI”.
135.
by aabhay 1y ago
The fact that you’re cognitively aware is evidence that this is nowhere near diffusion. More like rumination or thinking tokens, if we absolutely had to find a present day LLM metaphor
136.
by aabhay 1y ago
History is full of technology doing things that go beyond human possibility as well. Think of microscopes, guns, space shuttles. There has been technology that explicitly replaces human labor but that is not at all the whole story.
137.
by aabhay 1y ago
One thing about MCP that some people forget is that the models are post trained on MCP-based rollouts. I think people imagine that MCP was something people discovered about models but it’s deeper than that — models are explicitly trained to
138.
by aabhay 1y ago
I mean how well do you reckon you can denoise a jpg by hand until its a piece of art? That way of thinking isn’t helpful to understanding AI IMO
139.
by aabhay 1y ago
Amazing example. AI turns the bedgrudging third rate API UX into a must-win agent UX
140.
by aabhay 1y ago
I mean ultimately you’re in OpenAI’s world, they have even more innate control of language, meaning, and truth
141.
by aabhay 1y ago
Swift, while being slightly more friendly for extreme beginners than rust, is just so kneecapped and bloated that it has the opposite problem as rust. At some moderate level of competence it actively hinders learning. All of the weird closu
142.
by aabhay 1y ago
What’s so much more compelling about a graph db than a “with recursive” sql call? And even such recursive sql calls should be fairly rare since you would rather cache frequently accessed foreign key relationships anyway.
143.
by aabhay 1y ago
It is? Which ones give you distributed post training of MoE LLMs?
144.
by aabhay 1y ago
You think too highly of us (humans)
145.
by aabhay 1y ago
People can laugh at Gary Marcus all they want, but there’s one aspect that people don’t understand. If you have a high conviction belief that is countervailing to the mainstream, you suffer a great deal. Even the most average conversation w
146.
by aabhay 1y ago
The repo mentions approximate NN search but the article implies this is mainly brute force. Is there any indexing at all then? If not, is the approximate part an app-space thing e.g. storing binary vectors alongside the real ones? In additi
147.
by aabhay 1y ago
If you fire someone, you get to save on even more software licenses!
148.
by aabhay 1y ago
Not only that but they’re lossy. A hex representation is strictly more information as long as comments are included or generated.
149.
by aabhay 1y ago
You’re right. I gave the wrong number. My model implies 2^12000 concepts, because you choose whether or not to include each concept to form your dimensional subspace. I’m not even referring to the values within that subspace yet, and so onc
150.
by aabhay 1y ago
My intuition of this problem is much simpler — assuming there’s some rough hierarchy of concepts, you can guesstimate how many concepts can exist in a 12,000-d space by taking the combinatorial of the number of dimensions. In that world, ea

More ›