HNHacker News
TopNewBestAskShowJobs

0-_-0

4,368 karma · joined October 10, 2018

submissionscomments
0-_-0··on DeepSeek v4
Weights are the source, training data is the compiler
0-_-0··on DeepSeek v4
Weights are the source, training data is the compiler.
0-_-0··on Mythos is shaping up to be a nothingburger
"We also haven't seen any holes that couldn't have been dug by an elite construction worker." Therefore excavators are nothingburger
0-_-0··on XOR'ing a register with itself is the idiom for zeroing it out. Why not sub?
So then the question is: which pipeline is used less? Bit or add?
0-_-0··on Kimi K2.6: Advancing open-source coding
Pun intended?
0-_-0··on All phones sold in the EU to have replaceable batteries from 2027
1 mm thickness is a fine trade-off
0-_-0··on 4-bit floating point FP4
A 256 element lookup table is much bigger than a simple multiplier
0-_-0··on The RAM shortage could last years
Exactly, that's why they are not building more capacity and that's why RAM prices will stay up for years.
0-_-0··on 4-bit floating point FP4
You want to make multiplication cheap, it's not just about compression
0-_-0··on 4-bit floating point FP4

  00 -> 0.0
  01 ->-0.0
  10 -> Inf
  11 -> -Inf
0-_-0··on EU age verification app hacked, 2 minute How to posted
My understanding is that this is much more privacy friendly than showing your id
0-_-0··on AMD's Ryzen 9 9950X3D2 Dual Edition crams 208MB of cache into a single chip
Instead of a cache you could put down an SRAM buffer, it would be more efficient than a cache and just as fast. And addressable. Interesting idea.
0-_-0··on Kotlin creator's new language: talk to LLMs in specs, not English
Besides, you can deterministically generate bad code, and not deterministically generate good code.
0-_-0··on Inside the M4 Apple Neural Engine, Part 1: Reverse Engineering
Why would Google need graphics functionality to train neural networks?
0-_-0··on Inside the M4 Apple Neural Engine, Part 1: Reverse Engineering
When you already have a GPU in a system, adding tensor cores to it is much more efficient than adding a separate NPU which needs to replicate all the data transfer pipelines and storage buffers that the GPU already has. Besides, Nvidia's tensor cores are systolic.
0-_-0··on /e/OS is a complete, fully “deGoogled” mobile ecosystem
You have no idea how BAT ads work in brave, do you?
0-_-0··on Thanks a lot, AI: Hard drives are sold out for the year, says WD
So it's like humans then
0-_-0··on Expensively Quadratic: The LLM Agent Cost Curve
What's in the prompt cache?
0-_-0··on Expensively Quadratic: The LLM Agent Cost Curve
The cache gets read at every token generated, not at every turn on the conversation.
0-_-0··on Americans want heat pumps – but high electricity prices may get in the way
"The standards will require minor updates to gas-fired storage (gas tank) water heaters."
0-_-0··on The Perry Bible Fellowship
My favourite PBF comic: https://pbfcomics.com/comics/no-survivors/
0-_-0··on Richard Feynman Side Hustles
Had to read it 3 times but it makes sense
0-_-0··on Ask HN: Notification Overload
Your main problem is you have to much FOMO. Just keep notifications off for things you don't need.
0-_-0··on Ask HN: Notification Overload
Switch browser to desktop mode, works for me
0-_-0··on Booting from a vinyl record (2020)
Very unlikely, you need the browser for that
0-_-0··on TinyTinyTPU: 2×2 systolic-array TPU-style matrix-multiply unit deployed on FPGA
CPU branch predictors use neural networks
0-_-0··on CUDA Tile Open Sourced
Or even just pointers...
0-_-0··on Stranger Things creator says turn off “garbage” settings
Hear hear. 24 FPS is an abomination for fast movement.
0-_-0··on Stranger Things creator says turn off “garbage” settings
But the high FPS version is only in cinemas
0-_-0··on CUDA Tile Open Sourced
This is basically the nvidia equivalent of cooperative_matrix_2 in Vulkan which is vendor agnostic and should get much more hype that it's getting.
← PreviousPage 3 of 34Next →