Recent submissions

Comment · 2 months ago
Isn't ChatGPT benchmaxxing, then ? Responding "hmm…" isn't actually responding and latency should time to first relevant phoneme.
Comment · 2 months ago
I have zero interest in world knowledge for my LLMs but this got me wondering : are there RAGs for that kind of data ? How could a LLM like DeepSeek-v4-flash-vision-exp accurately answer you question with an indexed database of labeled landmark pictures (or even 3D models ?).
Comment · 2 months ago
Indeed ! LLM are creative writers, not journalists. Relying on overfitting for factual accuracy in not tenable. I don't understand why grounded RAG with judges in not the norm.
Comment · 2 months ago
You understand that drones are used for precision strikes instead of indiscriminate bombing, right ? It's the opposite of "sponsoring a genocide" (if there actually ever was one happening…)
Comment · 2 months ago
Everything is food and air. There is no such thing a "human labor" without food and air to sustain it. I'm not sure what your point is.
Comment · 3 months ago
BTW, I don't understand why having the driver converting AC to DC inside the LED spots is the default. For a new house, it seems to make sense to me to have at least one external driver for a ceiling of spots, if not one per floor (not one for the whole house because pf the drop in voltage). But 24V led spots seem much less common and I cannot find many reviews. Any idea why it's not more common ? It seems efficiency and thermals would be much easier to optimize without the size constraint, plus economy of scale with components. Any specific 24V led spots / bulbs to recommend ?
Comment · 4 months ago
Comment · 4 months ago
Comment · 7 months ago
Nice ! My most pressing request for VSS would be efficient binary vectors : is this on the table ?
Comment · 1 year ago
Only those who don't care/know about prompt processing speed are buying Macs for LLM inference.
Comment · 1 year ago
For LLM inference, I don't think the PCIe bandwidth matters much and a GPU could improve greatly the prompt processing speed.
Comment · 1 year ago
Indeed, recent Flash Attention is a pain point for non CUDA.
Comment · 1 year ago
The idea is presumably that you would "sell" at an artificially low price.
Comment · 1 year ago
> The Soviets had the […]first woman,[…]

That is quite the claim !

Comment · 1 year ago
You forgot the "/s", or do you actually believe that it's capitalism's fault is a mother taking care of her children is "unpaid labor" ?

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton