3abiton

9 ms·

151.
by 3abiton 1y ago
As many pointed out, Macs are decent enough to run them (with maxxed rams). You also have more alternative, like DGX Sparks (if you appreciate the ease of cuda, albeit a tad bit slower token generation performance), or the Strix Halo (good
152.
by 3abiton 1y ago
The hope is lost for Android, there is no moving forward with google antagonizing its foss roots. Libre phone it is. We have to forcibly remove the bandage.
153.
by 3abiton 1y ago
> Unlike nanochat this is purely vibe-coded, improving vibes by 110% Karpathy clearly said that it wasn't vibe coded. Apparently it was more time consuming to fix gpt bugs than to do it by himself.
154.
by 3abiton 1y ago
To be fair finding K is highly domain dependent and I would argue should not be for the analyst (solely) to decide, but with a feedback from domain experts.
155.
by 3abiton 1y ago
Wait till it gets gobbled by the next gen training data, and embedded in the weights of upcoming LLMs. Paired a clueless vibe coder, with no token limits.
156.
by 3abiton 1y ago
> Many software engineers now spend as much (or more) time reviewing the output of their own AI tools than their colleagues’ code. This hits home.
157.
by 3abiton 1y ago
> Or, to put it another way, OpenAI’s 2025 revenue is on track to only be $3.1 billion more than last year, while its annual operational costs are set to be $24.1 billion more than last year. So, for every dollar of revenue growth OpenAI
158.
by 3abiton 1y ago
> Or, to put it another way, OpenAI’s 2025 revenue is on track to only be $3.1 billion more than last year, while its annual operational costs are set to be $24.1 billion more than last year. So, for every dollar of revenue growth OpenAI
159.
by 3abiton 1y ago
How does this differ from nanochat?
160.
by 3abiton 1y ago
That was a fun read, great planning, and testing strategy!
161.
by 3abiton 1y ago
> Probably, could explain why central banks and investors have been stocking up on gold and crypto even as interest rates decline. The main purchaser of gold in recent months has been countries like Turkey. I don't think the gold ru
162.
by 3abiton 1y ago
Looks pretty neat, what's the advantage over LLM to mermaid?
163.
by 3abiton 1y ago
I know this is a big request, but lately I've been using LLMs to quickly draft and showcase architecture decision. I mainly fallback to mermaid, but I am curious if d2 would be more flexible? I usually endup visualizing anyway with dra
164.
by 3abiton 1y ago
I am curious were the differntiation comes between linux, and other unix systems (bsd, etc ..).
165.
by 3abiton 1y ago
To be fair, as the "AI" industry came late, the dibs were called on all the cool acronyms/names (LoRa for example).
166.
by 3abiton 1y ago
Essentially there is growing interest in the "formal" math community (combinatorics, mining, etc ..) to do exactly this.
167.
by 3abiton 1y ago
That's why project like nanochat are really cool, you can get around the limitations of such gigantic libraries, while at the same time understanding the underlying architecture.
168.
by 3abiton 1y ago
It seems to me the whole AGI problem is ill posed and barely well formalized. And thus you can always move the goal post.
169.
by 3abiton 1y ago
I think this is a moot argument, especially for x86 arch. I can run nearly almost any game built for windows 98 - till now, on linux.
170.
by 3abiton 1y ago
I would argue MCP is technically a "tool calling" approach, albeit more specific.
171.
by 3abiton 1y ago
I wish this was wrong, but it really isn't. To contrast though, I would argue that is part of evolution? We just want to do things faster or better? Smartphones solved no problems, but they ushered the digital millenium.
172.
by 3abiton 1y ago
So basically its legitimacy and novelty is based on a loop hole? Nonetheless a cool hobby project, and usually will rally people behind this concept and we might see some excitment there.
173.
by 3abiton 1y ago
I low key want to see official documents stating the name of some of those threads: all the "bongland" and" have you got a loicense" threads with some of their respective comments.
174.
by 3abiton 1y ago
While the title read a bit dramatic, I find it hard to disagree on the concepts
175.
by 3abiton 1y ago
I could feel the anger of the author oozing through the writing. Great work!
176.
by 3abiton 1y ago
15B lost, but it gave us good memes. Still, unbelievable amount.
177.
by 3abiton 1y ago
It's interesting 22 parts in under a year, seems like a fun up to date project. Karpathy did something very similar with nanochat (following nanogpt).
178.
by 3abiton 1y ago
> nobody knows precisely what to do to ensure an AI writes formal emails correctly or summarises text accurately. This is a bit of a hyperbole, a lot of the recent approaches rely on MoE, that are specialized. This makes it much more usu
179.
by 3abiton 1y ago
> Also, the other reviews I’ve seen point out that inference speed is slower than a 5090 (or on par with a 4090 with some tailwind), so the big difference here (other than core counts) is the large chunk of “unified” memory. It's no
180.
by 3abiton 1y ago
As someone who hot on early on the Ryzen AI 395+, are there any added value for the DGX Spark beside having cuda (compared to ROCm/vulkan)? I feel Nvidia fumbled the marketing, either making it sound like an inference miracle, or a dev

More ›