mufeed vh

1.4K posts

mufeed vh banner
mufeed vh

mufeed vh

@mufeedvh

cyber arms dealer @winfunction (YC S24)

sf 🌉 | kerala 🌴 Katılım Mart 2017
3K Takip Edilen4.6K Takipçiler
mufeed vh
mufeed vh@mufeedvh·
this was fun, thanks for doing this! we (@winfunction) don’t usually work on smart contracts, so this was one of our first runs in the space. pretty encouraging to see it land strongest on peripheral contracts - 12 novel peripheral root causes and 0 false positives. maybe we should make our debut in web3 audit contests. more interesting than i thought! 🙂‍↕️
English
1
0
2
135
mufeed vh
mufeed vh@mufeedvh·
cool stuff! we recently added experimental smart contract support for @winfunction and just ran it on majeur. i sent you the complete audit doc in your DMs. happy to give you access to our dashboard if that makes triaging/exporting easier! by our count, it ranks #1 on strict validated smart-contract novelty yield (contract findings only; frontend excluded).
mufeed vh tweet media
English
1
0
1
72
mufeed vh retweetledi
Miranda Nover
Miranda Nover@mirandanover·
Introducing Fort, a wearable that automatically tracks strength training. Strength training is one of the best things you can do for your health and longevity. It deserves better tools.
English
263
105
1.1K
412.2K
mufeed vh retweetledi
typedfemale
typedfemale@typedfemale·
why did it change the seed from 42 to 137? well it’s actually a physically principled decision…
typedfemale tweet media
Andrej Karpathy@karpathy

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autor… Part code, part sci-fi, and a pinch of psychosis :)

English
12
16
445
43.5K
mufeed vh retweetledi
Thomas H. Ptacek
Thomas H. Ptacek@tqbf·
We are all living in the last fleeting moments where there's any uncertainty that LLMs will supplant most human-driven vulnerability research. Enjoy it, if that's your thing; it's not going to last much longer.
English
23
32
336
49.7K
mufeed vh retweetledi
Makers of Kerala
Makers of Kerala@makersofkerala·
Mufeed made this pretty interesting programming language out of M&Ms!
mufeed vh@mufeedvh

I made a programming language out of M&Ms. 🔴🟠🟡🟢🔵 You arrange candy by color and position, take a photo, and it becomes code. Built the full thing end-to-end: - interpreter - compiler to candy-sheet images - reverse parser - AST + execution trace - browser demo It is completely useless and I love it. Blog link in first reply.

English
0
1
3
506
mufeed vh
mufeed vh@mufeedvh·
I made a programming language out of M&Ms. 🔴🟠🟡🟢🔵 You arrange candy by color and position, take a photo, and it becomes code. Built the full thing end-to-end: - interpreter - compiler to candy-sheet images - reverse parser - AST + execution trace - browser demo It is completely useless and I love it. Blog link in first reply.
English
9
8
48
2.5K
mufeed vh retweetledi
Doc
Doc@caballerobrah·
They don't want you to know this but you can just generate formally-verifiable, error-free binary protocol parsers and state machines arxiv.org/pdf/1910.02146
English
2
10
201
11.7K
mufeed vh retweetledi
Thomas H. Ptacek
Thomas H. Ptacek@tqbf·
Nicholas Carlini at [un]prompted. If you know Carlini, you know this is a startling claim.
Thomas H. Ptacek tweet media
English
20
143
1.3K
194.4K
mufeed vh
mufeed vh@mufeedvh·
``` - do not write overly verbose code while making no compromise on handling edge cases either (especially for security critical code). - for performance critical code, think like a demoscene competitor without the code golfing. - emit code that would pass linters and formatters while keeping it human readable. ``` this, inside global AGENTS.md rules, seems to work pretty well for me!
English
0
0
2
675
hayden
hayden@haydendevs·
my main complaint with these code models is they still tend to over engineer everything. what codex did in 300 lines of unreadable code, I trimmed down to ~100. is this just a prompting skill issue?
English
182
9
1.6K
177.1K
mufeed vh retweetledi
siunam
siunam@siunam321·
I started playing CTFs in 2022, and LLMs definitely changed the **competitive** CTF scene a lot, especially since mid-2025. I also started using LLMs in late 2025. Yes, those models did one-shot many challenges, but what's the fun of slopping them? I learned absolutely nothing 🥲
siunam tweet mediasiunam tweet media
English
21
97
622
101.5K
mufeed vh retweetledi
kqx
kqx@kqx_io·
How a single typo led to RCE in Firefox Can you spot the bug? Read now at: kqx.io/post/firefox0d…
kqx tweet media
English
6
93
637
147.1K
mufeed vh
mufeed vh@mufeedvh·
Just open-sourced: gitctx gitctx is an MCP server that gives coding agents high-signal, up-to-date GitHub context in one shot. Be it code / PR / issues / commits / releases / deps. Works with Claude Code, Codex, Cursor, Amp, etc. This is essentially an open-source alternative to @AmpCode's Librarian and was inspired by the same. ❤️ If you use AI coding workflows and want better repo intelligence, try it and send feedback.
mufeed vh tweet media
English
3
1
6
573

Keşfet

@winfunction @AckeeBlockchain @auron_xyz @webrainsec @cantinaxyz @Certora @joranhonig @elonmusk