←
bunny54,602 posts
bunny avatar✓

bunny ✓

@ConejoCapital

Conejo 🐇 Bueno | in it for the culture & tech | founder @clawpumptech @LyberAI @TektonicCompany @AIPSummit | love @solana | fun @pleasrDAO

◷ Joined September 2017▧ 6,303 Photos and videos
bunny avatar
bunny@ConejoCapital·Nov 11

With the birth of abliteration, we are at a pivotal point for crypto security. I used @LyberAI's uncensored models to find exploits on @toly's Percolator and it just got merged. If you don't bring your A-game, attackers will. Secure your contracts at Lyber.ai

Post attachment
ludo avatar
⟳ bunny retweeted
ludo@ludo_txtx·4h

🧵 Formal verification on @solana Absolutely thrilled by this 1st release. @HarmonicMath, founded by @vladtenev and @tachim, just added support for Rust to their product Aristotle, with Solana programs being a FIRST-CLASS citizen. They worked closely with 15 teams from ecosystem, calibrating Aristotle on the typical properties you'd expect from an on-chain program on Solana

0:20
bunny avatar
bunny@ConejoCapital·3h

Absolutely sick to my stomach. AI hackers got into my iMessage and texted my ex “I Love You, I’m Sorry” 100 times at 3 am before dozing off and sending her a tiktok edit of “We are never getting back together” by Taylor Swift. AI is taking us into a dystopia.

bunny avatar
bunny@ConejoCapital·14h

We finally have full ZDR for the most powerful uncensored zero refusal Kimi K3.1 Pro on @LyberAI. We also brought back @dealignai's famous Deepseek V4.1 Pro and we'll bring full ZDR to it also incredibly soon! Keep your research, defense workflows, private from us & everyone!

Post attachment
Lyber.ai avatar
⟳ bunny retweeted
Lyber.ai@LyberAI·14h

Kimi K3.1 Pro Abliterated Zero Refusal is live with Full ZDR. $4 in/$0.40 cache/$15 out/1M tokens. Uncensored DeepSeek V4.1 Flash is back. $0.29 in/$0.029 cache/$0.99 out/1M tokens. Full ZDR soon. Get up to $1,000 credits with a subscription: lyber.ai

Post attachment
bunny avatar
bunny@ConejoCapital·18h

THE MOST POWER AND ONLY TRUE ZERO REFUSAL @Kimi_Moonshot K3.1 PRO ABLITERATED IS ONCE AGAIN FULLY AVAILABLE AND WITH EXTRA GPUs ATTACHED JUST TO FULLFILL THE DEMAND FOR IT. GO USE THE MOST INSANELY POWERFUL ABLITERATD MODEL ON EARTH AT @LyberAI!!!

Post attachment
Lyber.ai avatar
⟳ bunny retweeted
Lyber.ai@LyberAI·18h

Kimi K3.1 Pro Capacity Increased. Abliterated. Uncensored. No KYC. Per 1M tokens: $4 input | $15 output Cached input $4 Existing users: $100+ inference gifted. Founders Scale: $399/mo → $1,200 credit × 3 paid cycles. Ends Sun Nov 22, 11:59pm GMT. lyber.ai/pricing/

Post attachment
bunny avatar
bunny@ConejoCapital·Nov 17

whats something good to do in London for someone just getting into London

MTS avatar
⟳ bunny retweeted
MTS@MTSlive·Nov 17

.@dealignai argues Chinese AI labs are releasing open-weight models to get as much of the world's hardware as possible running their software, ultimately locking users into their AI ecosystem: "I don't think that they're releasing models with any purposeful intent of getting us to abliterate models. Brutally honest, if they just wanted to make a completely cyber attack focused model and release it, they would." "I think they truly are playing the dominance game of getting as much hardware in the world running their models. You can have all the hardware, but if you don’t have the model to run on it, it doesn’t matter." "They really are trying to push specifically to get people locked in to using their software, their models." @LyberAI

1:05
0:06
Wesam avatar
⟳ bunny retweeted
Wesam@wesamo__·Nov 16

make a moss garden

0:50
MTS avatar
⟳ bunny retweeted
MTS@MTSlive·Nov 17

.@dealignai says he'd stop removing AI safety restrictions if models became dangerous enough for anyone to cause serious harm, but believes we're nowhere near that point: "I come from a heavy security background. At the current moment, it doesn't feel like models are close to the point where it is that easy or that dangerous." "If I ever deem the open-weight models that do come out that dangerous, that if we abliterate it, truly someone not even that motivated or capable can do really big harm with it, I would for sure stop, to be honest." "I don't think we're anywhere close to that level yet." "It's single streams of words, and through those words they think. We branch out with ideas and we merge multiple streams of thought together. "Until there is a drastic change in the architecture of how these models work, I don’t even think personally that they will be that dangerous." @LyberAI

1:46
0:06