@bytes032.xyz
6.2K posts

@bytes032.xyz
@bytes032
chief smart contract security cook @zenith256
Need a security audit? → Inscrit le Nisan 2020
1.5K Abonnements20K Abonnés

Cursor raised $2.3B to resell a open source model
Fynn@fynnso
was messing with the OpenAI base URL in Cursor and caught this accounts/anysphere/models/kimi-k2p5-rl-0317-s515-fast so composer 2 is just Kimi K2.5 with RL at least rename the model ID
English

I stay getting 0 likes, no one getting in my booty hole
shat@shatmypants10x
if u think 3 likes arent alot think of 3 people trying to rape u
English

added @cantinaxyz scan + @certora formal verification + @joranhonig grimoire skill run
making moloch/majeur the most AI audited code - ever.
ross.wei@z0r0zzz
English

My daughter came home with a black eye yesterday
"Dad, Ahmed hit me" she was crying
I sat her down immediately
"Sweetheart, you need to understand something"
"Ahmed comes from a different cultural background"
"He's still adjusting to European norms"
She started sobbing
"But papa, he punched me and called me a slut because I wanted to wear leggings"
My wife pulled me aside and told me to call the police
"And ruin a young man's integration process?" I asked
We are a tolerant and inclusive European family
I won't let one incident change that
English
@bytes032.xyz retweeté

i expanded this eval to 25 languages and evaluated gpt-5.4 medium /fast in codex
you can see the full results below
github.com/banteg/minigit…

Yusuke Endoh@mametter
I benchmarked which language works best with Claude Code. Ruby, Python, and JavaScript came out cheapest, fastest, and most stable! See my article in detail: dev.to/mame/which-pro…
English

I met an Iranian man in Luton Airport. He was in tears. Asked him what was wrong.
He said: "My family is trapped in a literal battlefield. There's violence in the streets, killings in the name of Islam, collapsed public infrastructure, the government are all criminals..."
I told him to keep strong, we're all praying for the citizens of Iran.
He looked confused.
"Iran? Bro, I live in Birmingham."
English
@bytes032.xyz retweeté

Latest V12 results on EVMBench (detect): 70.6% recall
GPT and Opus achieve ~40%, 45%.
EVMBench has a lot of problems. The grader is extremely flaky. Doesn't really measure FPR. No holdout set, invalid bugs. But anyways we're approaching saturation 😈🚀
More gains yet to come


OpenAI@OpenAI
Introducing EVMbench—a new benchmark that measures how well AI agents can detect, exploit, and patch high-severity smart contract vulnerabilities. openai.com/index/introduc…
English

@thsottiaux It would be great when the TUI gives a file path, when I cmd + click it would open in my default editor. I think this is big QOL improvement.
English

@gakonst hatchet is pretty decent lightweight temporal alternative
English

github.com/earendil-works…
github.com/earendil-works…
extreme alpha in these two repos imo
English
@bytes032.xyz retweeté






