Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
rao-v
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
10 ms
·
211.
▲
by
rao-v
11mo ago
I’ve been struggling to find an AM5 board that can run three MI50s at 4x. This is perfect thank you. Him are you sure about some of the PCI slots? I think some marked as 4x get downgraded to 1x on these boards… Further edit - this maybe acc
212.
▲
by
rao-v
11mo ago
I was genuinely impressed by the Antigravity browser plugin for "agentic" work. I ran into a neat website and asked it to generate a similiar UX with Astro and it did a decent-ish job of seeing how the site handled scrolling visua
213.
▲
by
rao-v
11mo ago
I mean it's even simpler. Almonds are entirely non essential (many other more water efficient nuts) to the food supply and in California consume more water than the entire industrial sector, and a bit more than all residential usage (~
214.
▲
by
rao-v
1y ago
I wonder if in the long term, with compute being cheap and parameter volumes being the constraint, if it will make sense to train models to be robust to different activation functions that look like ReLU (I.e. swish, gelu etc.) You might ev
215.
▲
by
rao-v
1y ago
I have to admit this really does seem like beautiful syntactic sugar despite me not being a fan of accumulating keywords in languages. Your example of writing a function instead of a neat little lambda is clunkier to quickly scan for correc
216.
▲
by
rao-v
1y ago
Exactly this! I’d love a modern C++ like syntax with the expressiveness of python and a mostly functional approach. C# is not that far I suppose from what I want
217.
▲
by
rao-v
1y ago
All I want is terminals to share config across MacOS, Linux and ideally Windows and have great support for SSHing into low end boxes with some minimal persistence solution like tmux on them. In an ideal world I would not have to think about
218.
▲
by
rao-v
1y ago
The reality is that poison is dose dependent, and we’ve mostly identified all the dangerous things and their doses. That lead in the bottle and the formaldehyde in your cheap clothes is bad … but not in a way that will detectably cause you
219.
▲
by
rao-v
1y ago
You could probably speed this up a lot by getting the token log probs and finding your category based on the highest log prob token that is in the category (you may need multiple steps if your categories share token prefixes)
220.
▲
by
rao-v
1y ago
It does sometimes feel that all of these systems are slowly rediscovering that the OG, Aider ( https://github.com/Aider-AI/aider ), had a near perfect architecture for pair programming with LLMs from the start.
221.
▲
by
rao-v
1y ago
I’d love to see an analysis of how much typing hurts LLMs that need to read / edit your code (due to increased context) vs helps (due to more clear type context). I want to believe that corrected typed python code is easier for smaller
222.
▲
by
rao-v
1y ago
I’m amused to see a for loop in a function that is purportedly branchless (not a critique)
223.
▲
by
rao-v
1y ago
I'm kinda confused as to the licensing on this. Can I use this as a sandboxed python for random (non-web) projects like I would pydiode?
224.
▲
by
rao-v
1y ago
It does suggest that there are many important problems out there that are amenable to relatively cheap brute force search at this point.
225.
▲
by
rao-v
1y ago
It is possibly (superb subreddit) but painful to convert a modern model and takes ages for them to be supported. The NPU is energy efficient but no faster than CPU for generation (and has lousy software support). I’m mostly interested in th
226.
▲
by
rao-v
1y ago
Yup! Build and ignore KleinAI and Vulkan etc. I’ve found that a clean CPU only build is optimal
227.
▲
by
rao-v
1y ago
Nice! Cheap RK3588 boards come with 15GB of LPDDR5 RAM these days and have significantly better performance than the Pi 5 (and often are cheaper). I get 8.2 tokens per second on a random orange pi board with Qwen3-Coder-30B-A3B at Q3_K_XL (
228.
▲
by
rao-v
1y ago
Why not just buy a put on TSLA of the right size? It's probably cheaper to Vanguard + a put than to buy a more expensive specialized ETF
229.
▲
by
rao-v
1y ago
This generally is a version of what economics and game theory knows as countersignalling. A classic paper is “Too Cool for School” https://host.kelley.iu.edu/riharbau/cs-randfinal.pdf Always worth pondering when it wor
230.
▲
by
rao-v
1y ago
How good is your esp32 support? This could make a lot of home automation hobbyists very happy!
231.
▲
by
rao-v
1y ago
Not at Google (anymore), but Gemma3n is a radically different (and very cool) architecture. The MatFormer approach essentially lets you efficiently change how many parameters of the model you use while inferencing. The 2B model they release
232.
▲
by
rao-v
1y ago
Fabulous stuff! Oh my request … the vision head on the Gemma models is super slow on CPU inferencing (and via Vulcan), even via llama.cpp. Any chance your team can figure out a solve? Other ViTs don’t have the same problem.
233.
▲
by
rao-v
1y ago
Lovely and simple … you’d think it would have become the best practice in most libraries by now
234.
▲
by
rao-v
1y ago
The directly into working memory bit is nonsense of course, but it does point to a problem that is probably worth solving. What would it take to make the KV cache more portable and cut/paste vs. highly specific to the query? In theory
235.
▲
by
rao-v
1y ago
Don’t give them ideas
236.
▲
by
rao-v
1y ago
I kind of want Python to have a desktop / web neutral library that does this and always am annoyed to learn it does not exist. Do folks understand why?
237.
▲
by
rao-v
1y ago
I’d have assumed that somebody had a format for just the basic operations … but it looks like even that is just being considered
238.
▲
by
rao-v
1y ago
I think this maybe why! Of course it’s great for vendor lock-in. But given how many companies need to work with diverse suppliers, there must be a whole bunch of re-creating models happening. There is no chance that everybody is using the s
239.
▲
by
rao-v
1y ago
This is helpful but surely we can save a consistent description of the process step and the input parameters alongside the final geometry in the same file format?
240.
▲
by
rao-v
1y ago
The challenge is that she’s right on all three of these major points … to some extent. Unfortunately at times she seems eager to fling babies without even encountering bathwater.
More ›