Peter Sobot @psobot

I teach computers to listen to music. 🇨🇦🎶👨🏼‍🔬🥁🎹🎸 petersobot.com Brooklyn, NY Joined January 2009

Tweets

13K
Followers

4K
Following

3K
Likes

39K

Peter Sobot @psobot

6 days ago

love to do machine learning work and see my PyTorch file identified as a SNES emulator page table by mime type detectors iana.org/assignments/me…

0 0 2 232 0

Dillon Mulroy λ @dillon_mulroy

a week ago

as an experienced engineer, the worst part of AI right now is the illusion of productivity and constantly battling the sunk-cost fallacy

102 182 3K 120K 258

Paul Butler @paulgb

a week ago

I haven’t written much Swift code but I hear a lot of good things. Congrats to the happy couple.

2 1 28 2K 0

The insiders I talk to at the big AI labs all seem to be on the same page about this: the only thing that matters is the pace of progress. Each model is just a checkpoint to show people how that progress is going. If progress slows significantly, all of the big labs are cooked.

François Chollet @fchollet

2 weeks ago

99 359 2K 227K 313

1 0 4 617 0

Peter Sobot @psobot

2 weeks ago

joshuavaldez.com/the-unbearable…

1 0 3 224 1

Peter Sobot @psobot

2 weeks ago

I love it when a plan comes together. Huge congratulations to all of my colleagues around the world who came together to make this product happen! newsroom.spotify.com/2025-08-19/mix…

1 0 8 519 0

Tim Hopper @tdhopper

3 weeks ago

Imagine explaining to my younger self that I’m struggling at work because my code editor, forked from an open source Microsoft editor written in typed JavaScript that can write code for me from text based prompts, is eating up the 64 GB of RAM on my Mac laptop.

2 2 8 711 0

vicki @vboykis

a month ago

i feel like, even though there is a LOT of good content out there, there still is not a good holistic overview of concurrency in python

11 2 95 8K 11

Peter Sobot @psobot

a month ago

friends have started turning down $1-2M/year offers to join big AI companies as engineers. three so far this summer all early-30s with 10+ years of big tech experience. all saying they don’t want to be stuck working 80-hour weeks at this stage of their lives. logical, but wild

0 0 11 609 0

Peter Sobot @psobot

a month ago

lack of design signals "I'm too focused on the mission to focus on design, so I'm just going to write pure HTML" but always peek under the hood to find out if that's true or not

0 0 4 374 0

Download Image

Peter Sobot @psobot

a month ago

wedding planning as a software engineer in 2025 means reverse-engineering the links that hotels give you so you save your guests from the dark patterns that hotels use on them when attempting to book in your room block

0 0 14 450 0

Download Image

Peter Sobot @psobot

a month ago

You know you're in too deep when the solution to your problem appears to lie in a GitHub repo with one commit, one file, and zero comments by someone who hasn't touched GitHub in 8 months.

0 0 10 380 0

Charles 🎉 Frye @charles_irl

2 months ago

beep boop, let's compute

0 1 7 734 0

Peter Sobot @psobot

2 months ago

macOS sometimes likes telling me how old my laptop is for some reason?

0 0 3 332 0

Download Image

Peter Sobot @psobot

2 months ago

Anecdote time: I spent 5 minutes using OpenAI Codex while waiting for a bagel this morning. It "fixed" three bugs in my codebases. As soon as I got home, I realized they each introduced new bugs. The real fixes took 10 seconds each when I took a moment to read the code.

METR @METR_Evals

2 months ago

236 1K 7K 3.6M 3K

Download Image

1 0 13 666 1

roon @tszzl

2 months ago

IME really good software ppl who deeply care find the least use from LLM coding and are often ideologically opposed to it because they like to exert editorial control over every line. slop research coders such as myself don’t care as much and have much larger gains

METR @METR_Evals

2 months ago

236 1K 7K 3.6M 3K

Download Image

78 21 1K 80K 160

Erik Bernhardsson @bernhardsson

2 months ago

A lot of the engineers with the deepest experience that I work with also have an absurdly wide span: everything from memory allocators to CSS selectors. They just like to build cool stuff, and that’s why they picked up so many things.

Gergely Orosz @GergelyOrosz

2 months ago

100 84 2K 328K 205

20 28 509 41K 75

Peter Sobot @psobot

3 months ago

Working in ML gives one the knowledge to know what’s possible, but not the wisdom to avoid wasting a full afternoon trying to get some obscure, experimental, low-quality research repo running.