muzz @muzzdotdev

zzz learning; not ready yet github.com/muzzlol Joined July 2023

Tweets

4K
Followers

140
Following

426
Likes

7K

muzz @muzzdotdev

17 hours ago

Nobody has the bread to eval opus man. How has anthropic gotten away with only increasing their pricing

Ibragim @ibragim_bad

a day ago

Nobody has the bread to eval opus man. How has anthropic gotten away with only increasing their pricing

15 26 207 29K 72

Download Image

0 0 0 65 0

## The 5 O'clock Rule The level of abstraction for solving a problem will bubble up until it allows the average dev to stop thinking about the problem. start decleratively defining your tasks for ai model with @DSPyOSS and optimize with evals that you know matter. stop…

1 1 3 1K 5

muzz @muzzdotdev

3 weeks ago

they don't know about @astral_sh uv over at /g/

0 0 2 143 0

Download Image

muzz @muzzdotdev

4 weeks ago

GLM-4.5 used SGLang. This was my sign to look into SGLang

0 0 0 99 0

muzz @muzzdotdev

4 weeks ago

im happy context free grammar is here but it shoudl've been here way sooner

0 0 1 102 0

muzz @muzzdotdev

4 weeks ago

this openai keynote couldn't have been worse..

0 0 0 99 0

muzz @muzzdotdev

a month ago

I get completely stunlocked from doing anything else productive once a training run ensues. Is this normal? does it get better with time? don't have a choice anyway, can't wait around doing nothing for so long

0 0 3 85 0

Download Image

muzz @muzzdotdev

a month ago

once you try different models with your agent, you’ll see just how much better the claude-4.(1) models are at agentic tasks than the rest.

1 0 1 84 0

muzz @muzzdotdev

a month ago

I knew @DSPyOSS was high signal, but man i was still surprised by how well thought out and executed this project is after using it.

0 2 8 684 0

Connor Shorten @CShorten30

a month ago

GEPA is a SUPER exciting advancement for @DSPyOSS and a new generation of optimization algorithms re-imagined with LLMs! 🧩🚀 Starting with the title of the paper, the authors find that Reflective Prompt Evolution can outperform Reinforcement Learning!! 🤯 Using LLMs to write…

22 50 354 62K 263

Download Image

ManusAI @ManusAI_HQ

2 months ago

After four overhauls and millions of real-world sessions, here are the lessons we learned about context engineering for AI agents: manus.im/blog/Context-E…

29 276 1K 516K 1K

Armin Ronacher ⇌ @mitsuhiko

2 months ago

Does someone have an Open Source codebase they had little success with agentic coding I could use as a test-bed to improve the agent workflow? I think I might make a video about that. x.com/mitsuhiko/stat…

Armin Ronacher ⇌ @mitsuhiko

2 months ago

52 6 201 43K 52

19 4 100 24K 35

muzz @muzzdotdev

2 months ago

I see a lot of people developing with CUDA libraries on cloud notebooks or macOS without proper IDE support. No type hints, no linter errors, no autocomplete etc. just raw dogging and hoping it works when you finally run it on the actual GPU box... just use remote ssh to a…