Artem Andreenko

2.8K posts

Artem Andreenko banner
Artem Andreenko

Artem Andreenko

@miolini

Building cool stuff @SentientWaveHQ. Pancomputationalist. This is my personal blog about compute, communication, and energy.

San Diego, CA เข้าร่วม Nisan 2008
4K กำลังติดตาม4.4K ผู้ติดตาม
NASA Artemis
NASA Artemis@NASAArtemis·
The Orion spacecraft has a couple of backseat drivers.
English
38
222
2.3K
61.1K
Artem Andreenko
Artem Andreenko@miolini·
@dsvilhena @NASAArtemis A camera is mounted at the end of the solar panel section. The four solar panel sections are flexible and damp vibrations from the main cabin and service module.
Artem Andreenko tweet media
English
1
0
7
187
asic destroyer
asic destroyer@splinedrive·
Anything not written in SystemVerilog or Verilog is dead to me.
English
13
4
75
4.6K
Thomas H. Chapin IV
Thomas H. Chapin IV@tomchapin·
I gave my local OpenClaw the ability to control its own machine at the system level via self hosted VNC. I had it build and open source its own tooling for this. It can now click its own “allow” permission dialog windows and even run sudo commands. github.com/tomsalphaclawb…
English
3
0
2
300
Co
Co@ITangoI·
@jamonholmgren Double it lol Float32 for what grid zone you are in Float32 for where on that grid you're in Now you have enough subdivision to probably map the earth.
English
5
0
9
1.9K
Jamon
Jamon@jamonholmgren·
Floating point precision gets pretty bad at 500 kilometers from the origin point 😅 Working on an origin shifting system, but it's slow going, given my terrain system doesn't support it and I'm having to do a bunch of C++ wrangling, compiling, reloading ... not the fastest feedback cycle
English
91
16
612
78K
Paul Lackey
Paul Lackey@holliemaea·
@miolini I’m not actually 100% sure. It’s Molicel P50B and definitely has some magic inside.
English
1
0
0
25
Paul Lackey
Paul Lackey@holliemaea·
Pulling a solid 200W from a tiny little 21700 cell for more than 90 seconds. That's actually a bit crazy. Batteries these days are INSANE.
English
2
0
2
153
Marty
Marty@thesuprememarty·
@miolini q8_0 makes sense — dequant overhead is manageable there so tensor cores can shine. does the tradeoff flip at 4-bit and below, or does Turbo Quant have a different path for lower widths?
English
1
0
0
20
Artem Andreenko
Artem Andreenko@miolini·
Turbo Quant compression shifts LLM inference along the complexity spectrum from memory bandwidth bound to compute bound, making tensor cores work harder.
Artem Andreenko tweet media
English
1
0
1
170
Will Hughes
Will Hughes@woodtechwill·
@miolini 317W for zero-human ops. UK factories pull 50kW+ running half-automated lines. Local inference is eating legacy infrastructure.
English
1
0
1
39
Artem Andreenko
Artem Andreenko@miolini·
24 Hour energy report for a zero human company: 1. Hardware: AMD Ryzen 9 7950X, 2 x NVIDIA RTX 4090 GPU. 2. Software: llama.cpp, AutomataOS, Automata, Matrix. 3. Model: qwen3.5-27b 4. Company size: 10 employees. 5. Energy: 7.6kWh ($4.18 in SD) 6. Power draw of 317W.
Català
1
0
0
317
Austin Hirsch
Austin Hirsch@austinhirsch·
2027 is going to be wild. @__tinygrad__ is trying to sell a computer for 10m with 2.5 TB of RAM
Austin Hirsch tweet media
English
7
0
26
46.1K
Marty
Marty@thesuprememarty·
@miolini the crossover is the interesting part — at low enough bit widths, dequant overhead at inference time can eat the tensor core gains. does Turbo Quant target a specific bit-width range where the shift is net positive, or does it depend heavily on hardware tier?
English
1
0
0
28
sachin.
sachin.@sachinyadav699·
Every idea feels taken. Every API already exists. Every SaaS has 12 competitors. So what do we even build now?
English
333
63
1.1K
95.4K
Artem Andreenko
Artem Andreenko@miolini·
@MarcusHouse I just heard on the radio that one of the PCD (Microsoft Surface laptop) is not charging and turning on anymore.
English
0
0
0
313
Marcus House
Marcus House@MarcusHouse·
Yes... In case anyone was wondering, Microsoft still sucks in space.
English
224
1.6K
18.3K
1.4M
Artem Andreenko
Artem Andreenko@miolini·
I had Orion radio traffic on in the background for a couple of hours. I hope the next Artemis missions will run local AI agents on board. Some of the questions currently directed to Houston could already be handled by agents. May be even our Automata one day will fly in space.
Artem Andreenko tweet media
English
0
0
0
110