Philip Nuzhnyi
LightNews LightNews
banner
callmephilip.com
Philip Nuzhnyi
@callmephilip.com
210 followers 290 following 1.1K posts
🧜‍♂️ Surftware engineer. Stochastic parrot.

CODE: https://github.com/callmephilip
MD: https://callmephilip.com/
Posts Replies Media Videos
callmephilip.com
This one goes pretty hard
November 8, 2025 at 12:02 PM
callmephilip.com
Apparently abiogenesis is also sometimes referred to as "biopoesis" - which is a. poetic b. poesque
November 7, 2025 at 11:10 AM
callmephilip.com
What a great episode! Chris Lattner is a beast.
November 7, 2025 at 10:20 AM
callmephilip.com
I appreciate this sentiment and attitude.
tedunderwood.com Ted Underwood @tedunderwood.com · 9d
I would find far more wearisome than letting people do it for themselves.

Are people overusing and abusing block lists? Probably.

Do I subscribe to any? No.

But a little link rot is a fair price to pay for giving my fellow users self-determination. And that means a right to make mistakes.
November 4, 2025 at 11:09 AM
callmephilip.com
"Each program inside tmux gets its own terminal managed by tmux, which can be accessed from the single terminal where tmux is running - this called multiplexing and tmux is a terminal multiplexer." - from github.com/tmux/tmux/wi...
November 3, 2025 at 11:28 AM
callmephilip.com
🤖 "then, man made the machine" - open.spotify.com/track/7H2CS2...
Ex Machina
open.spotify.com
October 31, 2025 at 10:42 AM
callmephilip.com
Was "attention is all you Heed" taken?
October 31, 2025 at 9:24 AM
callmephilip.com
Great news! 🎈
brianmc.bsky.social Brian McCullough @brianmc.bsky.social · 16d
🧵Some personal news: After a 6 year hiatus, the Internet History Podcast is back! Yes, the original tech history podcast is back with the first three episodes below, followed by a new one every week for the foreseeable future. Search your podcast app now youtu.be/sjkkdcMp2Pk
Internet History Podcast | Jimmy Wales, Founder Of Wikipedia
YouTube video by Internet History Podcast
youtu.be
October 28, 2025 at 1:56 PM
callmephilip.com
No moar splits, y'all. h/t @wesbos.com
October 24, 2025 at 10:59 AM
callmephilip.com
Quanto vale a tua alma?

Quanto vale a alma de um bairro?

Quanto vale o SC⚡️DC?

~~~

How deep is your love?
johnfisherman.bsky.social Fred Rocha @johnfisherman.bsky.social · 23d
Já ninguém tem hobbies
Todos temos side-hustles
Comecei um curso de baking
na Domestika
Agora abro um perfil no Insta
Vi uns tutoriais no YouTube
Agora abro uma loja
De knitting
No Etsy.

fredrocha.net/2025/10/21/l...
Lisboa sem Graça - fredrocha.net
Entre perfumes de musk / O Capital do chacal / Arreganha os dentes / Perde a graça.
fredrocha.net
October 21, 2025 at 11:35 AM
callmephilip.com
"According to the DeepSeek paper, you can pull out 10 text tokens from a single image token with near-100% accuracy." - via www.seangoedecke.com/text-tokens-...
October 21, 2025 at 10:17 AM
callmephilip.com
"Want to make OpenAI transcriptions faster and cheaper? Just speed up your audio. [...] Run your audio through ffmpeg at 2x or 3x before transcribing it. You’ll spend fewer tokens and less time waiting with almost no drop in transcription quality." -
george.mand.is/2025/06/open...
October 21, 2025 at 10:06 AM
callmephilip.com
This ⤵️ should be very useful for Claude skills
callmephilip.com Philip Nuzhnyi @callmephilip.com · Jul 1
What would you call this? "Just in time distribution"?
koaning.bsky.social Vincent D. Warmerdam @koaning.bsky.social · Jul 1
Wrote a more detailed blog about how/why here:

koaning.io/posts/uvx-p...
October 18, 2025 at 10:44 AM
callmephilip.com
"There is nothing wrong with sobriety in
moderation." — John Ciardi
October 14, 2025 at 8:46 AM
Reposted by Philip Nuzhnyi
timkellogg.me
Karpathy: nanochat

A small training+inference pipeline for creating your own LLM from scratch

$100 will get you a somewhat functional model

$1000 is more coherent & solves math

detailed walkthrough: github.com/karpathy/nan...

repo: github.com/karpathy/nan...
Andrej Karpathy & @karpathy X.com Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single, dependency-minimal codebase. You boot up a cloud GPU box, run a single script and in as little as 4 hours later you can talk to your own LLM in a ChatGPT-like web Ul. It weighs ~8,000 lines of imo quite clean code to: - Train the tokenizer using a new Rust implementation - Pretrain a Transformer LLM on FineWeb, evaluate CORE score across a number of metrics - Midtrain on user-assistant conversations from SmolTalk, multiple choice questions, tool use. - SFT, evaluate the chat model on world knowledge multiple choice (ARC-E/C, MMLU), math (GSM8K), code (HumanEval) - RL the model optionally on GSM8K with IPDDOI - RL the model optionally on GSM8K with "GRPO" - Efficient inference the model in an Engine with KV cache, simple prefill/ decode, tool use (Python interpreter in a lightweight sandbox), talk to it over CLI or ChatGPT-like WebUl. - Write a single markdown report card, summarizing and gamifying the whole thing. Even for as low as ~$100 in cost (~4 hours on an 8XH100 node), you can train a little ChatGPT clone that you can kind of talk to, and which can write stories/poems, answer simple questions. About ~12 hours surpasses GPT-2 CORE metric. As you further scale up towards ~$1000 (~41.6 hours of training), it quickly becomes a lot more coherent and can solve simple math/code problems and take multiple choice tests. E.g. a depth 30 model trained for 24 hours (this is about equal to FLOPs of GPT-3 Small 125M and 1/1000th of GPT-3) gets into 40s on MMLU and 70s on ARC-Easy, 20s on GSM8K, etc. My goal is to get the full "strong baseline" stack into one cohesive, minimal, readable, hackable, maximally forkable repo. nanochat will be the capstone project of LLM101n (which is still being developed). I think it also has potential to grow developed). I think it also has potential to grow into a research harness, or a benchmark, similar to nanoGPT before it. It is by no means finished, tuned or optimized (actually I think there's likely quite a bit of low-hanging fruit), but I think it's at a place where the overall skeleton is ok enough that it can go up on GitHub where all the parts of it can be improved. Link to repo and a detailed walkthrough of the nanochat speedrun is in the reply. nanochat
October 13, 2025 at 6:06 PM
callmephilip.com
The subtle art of letting go
simonwillison.net Simon Willison @simonwillison.net · Oct 11
I'm beginning to suspect that a key skill in working effectively with coding agents is developing an intuition for when you don't need to closely review every line of code they produce. This feels deeply uncomfortable!
October 11, 2025 at 12:45 PM
callmephilip.com
My buddy Dostoevsky is cycling from Lisbon to Malaga and I am rawdogging it! whereisdostoevsky.com source: github.com/callmephilip...
October 11, 2025 at 10:29 AM
callmephilip.com
Good conversation with good takes on LLM assisted programming, language choices, and common sense. No hype/fluff. newsletter.pragmaticengineer.com/p/python-go-... via @gergely.pragmaticengineer.com
October 9, 2025 at 9:34 AM
callmephilip.com
It’s like if Rosetta Stone were a rap song - open.spotify.com/track/5gpyyq...
September 25, 2025 at 7:50 PM
callmephilip.com
From "not everything is a chat" category for LLM apps. A couple of interesting experiments from Sean Goedecke:

www.endlesswiki.com/wiki/main_page (an infinite, AI-generated wiki)
www.autodeck.pro (AI driven spaced repetition app)
September 25, 2025 at 9:21 AM
callmephilip.com
Frenchies took over MS's docs department - www.typescriptlang.org/docs/handboo...
September 22, 2025 at 6:17 PM
callmephilip.com
A bitter lesson about the bitter lesson - www.dbreunig.com/2025/08/01/d... via @dbreunig.bsky.social. Excellent as always
September 22, 2025 at 12:13 PM
callmephilip.com
1k
September 11, 2025 at 2:15 PM
Reposted by Philip Nuzhnyi
sirhamy.bsky.social
How to Setup an Ergonomic Workstation - More productivity, less back / neck / wrist pain

hamy.xyz/blog/2025-09...
September 10, 2025 at 10:15 PM
callmephilip.com
Claude code agentic loop breakdown blog.promptlayer.com/claude-code-... h/t @timkellogg.me
September 9, 2025 at 11:25 AM