Mukshud

30.9K posts

Mukshud banner
Mukshud

Mukshud

@Mukshud

frmr @MCDB_yale & MB&B @Wesleyan_u🦠cell envelopes, genomics & evolution | proteins & 🪸to stories | Mtb, biotech & marine bio 🌊 | views = mine |$myco

oceanic abyss เข้าร่วม Mayıs 2020
1K กำลังติดตาม1K ผู้ติดตาม
ทวีตที่ปักหมุด
Mukshud
Mukshud@Mukshud·
First they call you crazy and delusional. Then they isolate you from your friends. And then you can’t have privacy. And then they use privileged government and court processes to abuse you in ways nobody else would be able to. Then, when they can’t find anything, they continue to monitor out of “an abundance of caution”
Mukshud@Mukshud

“The patient is having delusional thoughts and believes that he is working for AI and that his devices are hacked” Meanwhile, the conditions for my forced removal were not met. I never missed a psychiatry appointment, the document was never served to me before I was removed in handcuffs, and the nobody told me anything after. I still don’t know who Nelson Richardson is. I had to post this document on X to learn it was USPTO et al., behind this document.

English
157
0
8
8.4K
Mukshud
Mukshud@Mukshud·
Hey, is ur model taking up too much RAM? Vectorize ur tensors and keep them on ur GPU and then say fuck u to ur haters
Mukshud tweet media
English
0
0
0
4
Mukshud รีทวีตแล้ว
Josie Zayner
Josie Zayner@josiezayner·
The smartest people I know have never taken an iq test because its just a test for normies to try and prove they are smart
English
5
2
28
3.1K
Mukshud รีทวีตแล้ว
bayes
bayes@bayeslord·
Compiled reality is coming
English
3
2
15
921
Mukshud
Mukshud@Mukshud·
Me and my bestie from the Bay Area in California (y’all better be afraid cuz Mimikimbach throws downnnn a fire trance set)
Mukshud tweet media
English
0
0
0
13
Mukshud
Mukshud@Mukshud·
GitHub ftw tho because everything is time stamped and logged
Elon Musk@elonmusk

@beffjezos First they ignore you, then they ridicule you, then when you win they assign credit to someone else. And the cycle keeps repeating.

English
0
0
0
15
Mukshud
Mukshud@Mukshud·
Yes indeed
Elon Musk@elonmusk

@beffjezos First they ignore you, then they ridicule you, then when you win they assign credit to someone else. And the cycle keeps repeating.

English
0
0
0
10
Mukshud รีทวีตแล้ว
Awni Hannun
Awni Hannun@awnihannun·
One thing I miss that I don't experience working with coding agents is that feeling of flow state. Being locked in for hours on a tough problem. The latencies are too high. Lot's of context switching. In some ways it's nice. The apple watch reminder to stand up and stretch is easier to heed. But I can't wait until latencies come down and we move past the dial-up era of agents.
English
20
6
140
7.3K
Mukshud
Mukshud@Mukshud·
@ivanfioravanti @0xSero hmmm idk why people don't look at my projects? I've been developing a lot of tools for local training
English
0
0
0
6
Mukshud
Mukshud@Mukshud·
took a lot of trial and error but I got my @Snapdragon @Qualcomm GPU to sync in distributed parallel training with my mlx training runs @Apple @surface @linuxfoundation :D.. am building a comp bio/bioinformatics/computational chemistry coding LLM built on GPT architecture :) I'm using my custom drivers! and tensors ;)
Mukshud tweet media
English
0
0
2
70
Mukshud รีทวีตแล้ว
CNX Software
CNX Software@cnxsoft·
Low-power 25 TOPS M.2 AI accelerator module. cnx-software.com/2026/03/21/rad… @RadxaComputer AICore DX-M1M is built around the DeepX DX-M1M neural processing unit (NPU) and consumes around 3W of power. Its compact size and low power consumption make it ideal for industrial robot arms, autonomous mobile robots (AMR), edge servers, drones, and AIoT devices. It relies on a PCIe Gen3 x2 interface and works with both x86 and Arm systems, including the Raspberry Pi 5 and Radxa ROCK SBCs.
CNX Software tweet mediaCNX Software tweet mediaCNX Software tweet media
English
2
22
160
9.1K
Mukshud รีทวีตแล้ว
Mathieu
Mathieu@miniapeur·
Mathieu tweet media
ZXX
6
20
430
6.4K
Mukshud
Mukshud@Mukshud·
🌊🌦️🔓🔐🔑🔏
Mukshud tweet mediaMukshud tweet media
QME
0
0
0
23
Mukshud
Mukshud@Mukshud·
my thesis works extends into cereberospinal fluid and modeling that is wild. the dielectric constant of both the csf under circumstances can be similar to the dielectric constant of the periplasm of certain bacteria. there's an interesting relationship I'm finding
English
0
0
0
28
Mukshud
Mukshud@Mukshud·
and 3 dimensions
English
0
0
0
15
Mukshud
Mukshud@Mukshud·
the ideal and minimal path length for any hyperoptimized semiconductor for protein folding is 7–8 nm in thickness for reasons I won't describe until I test it :p
English
2
0
0
26
Mukshud
Mukshud@Mukshud·
the semiconductor itself would have at least 4 layers
English
0
0
0
12