gm machine smol update: >increased i2c speed from 100khz to 400khz & removed frame delay >left video (new) vs right (old). both recorded at 60fps but old version shows those scanning artifacts going down the screen >you can run smoothly even on small OLED screens

my gf gifted me this beauty, i love my gf

https://t.co/jckCvsQEos

the vods from Zigtoberfest are going out

november 3rd, it’s snowing, yes

doxxed one of me languages i speak lol

u can also gen dataset from a yt video (pic1) but u have to make smol changes in their parser in order to make it work (pic2, they're using deprecated api (pic3)) this brings nostalgic memories of "fixing" python libs in offline env

https://t.co/35gnH7ZCGQ

v cool repo from meta (huh?) that allows u to gen synthetic qa, qa + cot datasets from pdf, docx, txt etc. v useful for sft warmups

i hate how gov tech works, the bureaucracy part we have a pipeline, everything’s working fine etc. new task arrives: we need to send the data we’re receiving to another team which is part of the same gov entity, what the fuck

ever wondered what an RLVR environment is? in 27 min I’ll show you: - what they made of - how RLVR differs from RLHF - the performance gain it gives to small models - and a walkthrough of the verifiers specs to define them. by the end you will be able to make your own 👺🦋

new ios is so inconvenient from a ux perspective, where i need extra taps/clicks to use it. e.g. in photos i now need to tap 2 times to get into the collections tab, in safari to see all open tabs it's again 2 taps, whereas before it took 1! i don't like that they're sacrificing

first exams done, now i have more free time for interesting things

1st place on meta llama accelerator

so many drops, and i see little to no linkedinism, interesting

pod: https://t.co/1NsuXV4gOY blog posts:

disco pod, with the disco blogpost series zig was mentioned btw

i enrolled in uni, they making me learn deutsch und c++

CORE is launching a weekly challenge with prize $ Week 1 has been launched If I can trust you to not use LLMs for this challenge, link is in first :) (btw, I do actually read the entry questions)

vouch!!!!

https://t.co/bQ1VWsZnnL