rends_east

Nikita Kinelovsky

Security & smart contracts. Getting into LLMs, posting what I figure out along the way. Occasionally breaking things.

[rends_east @rends_east]

👋 Hi, my name is Nikita. I'm from Tbilisi, Georgia.

I'm a security and blockchain specialist, and right now I'm actively getting into AI.

I'm also the ex-founder of DEX @BidaskProtocol, where I raised $1M+ in funding.

I take on development and audit projects, and help with writing secure code. Always happy to connect :)

[rends_east @rends_east]

It was a nice week

[rends_east @rends_east]

@thecomfeed Probably a good way to launder money
He might have made the mistake on purpose to profit from it himself
Needs to be checked

[rends_east @rends_east]

Funny thought: because Codex refuses to discuss security topics, its code is less secure by default than code from other models
It reminds me of how security researchers used to identify an app’s stack just to know where its weak points were

[rends_east @rends_east]

@shivangmauryaa Do not use codex for bb
It will ban you for security questions

[rends_east @rends_east]

damage control?

[FFmpeg @FFmpeg]

Several FFmpeg developers have received six free months of the Claude Max 20x plan.

Thank you @ClaudeDevs and @AnthropicAI for supporting FFmpeg through the Claude for Open Source program!

So far, Claude has helped find missing backports for the upcoming 9.0 release.

[rends_east @rends_east]

@haydenzadams Honestly it is hard to see what there is not to understand about v4.
Concentrated liquidity is complex by itself, while hook logic often feels quite natural

[rends_east @rends_east]

@__phantom1 About 4 days, i guess

[rends_east @rends_east]

When I joined the waitlist for a Kimi K3 subscription, I stated that I’m a software engineer.
After I submitted it, I thought that was probably a stupid move. Surely software engineers would be let in last since we’re the ones who burn all tokens, all limits on all subscriptions.
But no, I got access pretty quickly. Now I’m happily sitting here with my subscription!

[rends_east @rends_east]

@hetmehtaa what are the business risks of your vulnerability?

[rends_east @rends_east]

@socialforging Interested. I previously built my own CLMM DEX as CTO and also worked on security audits. I’d be happy to chat!

[rends_east @rends_east]

@jessepollak Literally me.

[rends_east @rends_east]

@issanode If you were running 10 agents at once, that means you got 10× the experience!

[rends_east @rends_east]

Yesterday I had a thought: life would be much easier if .env files were committed encrypted
I googled and, turns out, I wasn’t the first person to think about it. That already exists
:(

[rends_east @rends_east]

X is talking about Anthropic’s fight against open source right now
But everyone seems to forget that Anthropic open-sourced сlaude сode less than a year ago!
People have short memories...

[Ryan Brewer @ryanbrewer]

Bad take, being against banning open source models is a fine position to hold while not open sourcing your own products.

No one is forcing OpenAI or Anthropic to open source their models, they are simply against lobbying the government to ban open source. It’s very easy to understand.

[rends_east @rends_east]

@QuixiAI @AnthropicAI I hope you cancelled your OpenAI subscription first

[rends_east @rends_east]

@agupta For some reason, people keep denying there’s a difference between “allowing something” and “banning everything else”
Funny crossover seeing the same debate on AI twitter...

[rends_east @rends_east]

@shevagro Честно говоря, никогда особо не помогало.
Во первых, галлюцинации всё еще есть, а для такой базы знаний это бывает критично
Во вторых, поиск для агента по такой базе данных сложен
Короче почти как проблема P=N(P).

[rends_east @rends_east]

@Dinosn guess what percentage of repo is AI slop