Shakedown (@Russellried1) header
Shakedown (@Russellried1) profile picture
WIAMI, Wilder
Joined August 2015
Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)Photo by Shakedown (@Russellried1)
🚨 NYAJIRA UPDATE! 🐈‍⬛🔥 Instagram has officially been added to our Dexscreener profile! The viral sensation behind NYAJIRA is now directly connected to our token. 1.3 MILLION likes in just 24 hours. This is more than just a meme. This is a movement. And we’re only getting started. 🚀
7
9
35
1,346
15M views.
ゴジラ、襲来。 ニャジラ、就寝。
Readers added context they thought people might want to know
AIで生成された画像です。 OpenAIが提供するチェッカーで、ツールを使用した事を意味するシグナル(SynthID)が検出されています。 openai.com/ja-JP/research…
1
3
12
683
GM Nyajiras!!
6
9
31
854
$NYAJIRA IS FOR THE COMMUNITY. 99% to the original creator. 1% to the team. No dev profits. No hidden agenda. The creator gives us the story. The community gives us the power. No community, no NYAJIRA. This isn’t about one person. THIS IS ABOUT ALL OF US. 🔥 $NYAJIRA is community-driven.
14
14
38
1,910
THE OFFICIAL $Nyajira 🐈‍⬛ 12 MILLION views in less than 24 hours since the original post went live. This is the ONLY authentic Nyajira token followed by the original creator, @nyankorooow. Don’t fall for copycats. There is only ONE real $Nyajira. ✅ Followed by the original creator ✅ The ONLY official $Nyajira ✅ Creator fees go to @nyankorooow CA (SOL): 8cWGfiHyKqxgLrLtkPt2NnLkSUV2JbcLBf4RN6Fapump THE REAL NYAJIRA HAS ARRIVED. 🔥
ゴジラ、襲来。 ニャジラ、就寝。
Readers added context they thought people might want to know
AIで生成された画像です。 OpenAIが提供するチェッカーで、ツールを使用した事を意味するシグナル(SynthID)が検出されています。 openai.com/ja-JP/research…
73
34
4
144
19,794
solana:8cWGfiHyKqxgLrLtkPt2NnLkSUV2JbcLBf4RN6Fapump Japanese cat memed as Godzilla. I honestly think it will work the tweet has gotten alot of attraction.
3
6
210
NEW ATH 8cWGfiHyKqxgLrLtkPt2NnLkSUV2JbcLBf4RN6Fapump
🤖 Made with AI
7
13
41
1,482
We have officially sent the Creator Rewards to @nyankorooow ! 🐱❤️ We sincerely wish Nyantaro continued success and look forward to seeing what the future holds for him. Thank you for bringing Nyajira into this world, Nyajira Father! 🦖❤️ The official $NYAJIRA community stands with you. 🤝
11
17
57
5,032
Shakedown retweeted
In a normal AI app, one company holds your account, your IP and every word you type. PARANOID splits who you are from what you ask, and hands the two halves to different parties. Your browser goes first. Emails, phone numbers, wallet addresses, names and other details that point to you are swapped for placeholders before anything leaves your device, and the real values go back into the answer on your side. You pay with a zero knowledge ticket, a proof of payment that stays anonymous. The relay sees your IP and carries only sealed bytes. Our gateway runs on Phala Cloud inside an Intel TDX chip, so it reads the words while seeing only the relay, and the chip stays shut to us too. The model gets only the words. All the gateway keeps is a list of spent tickets, so a ticket cannot be used twice, and each entry is just a random number. To tie you to a question, someone would have to break open a sealed Intel chip and get the relay's traffic at the same moment. Even then, your name, email and wallet were swapped out before the question left your device. paranoid.red
7
7
1
35
1,909
Shakedown retweeted
A lot of new people found PARANOID this week. If you are one of them, this article is the best place to start. It lists everything a normal AI app keeps behind a single question, like your card, your IP address and the faces in your photos, and explains how PARANOID removes each one inside an ordinary browser tab. Give it ten minutes and you will know how all of PARANOID works.
15
4
29
3,490
Shakedown retweeted
imagine being able to short your favorite kol that's where $perps is heading per the dev's latest alpha leak: every trader, wallet or kol becomes its own perps market each one with its own pool perps for humans, traders, even bots most requested traders go live first then the community votes on who gets added next think you know who's about to fade? soon you can put a position on it testing in pre-release now, open beta next $perps everything thesis.casino
🤖 Made with AI
2
4
9
229
Shakedown retweeted
OpenAI was caught paying hundreds of contractors $50 an hour to read real ChatGPT conversations, often with a summary of the user's memories on screen. Some of those users had asked ChatGPT to "keep this between us." This is not just OpenAI. Every AI company runs your words through servers its own people can open, and the only thing standing between them and your chats is a privacy policy they wrote themselves. On PARANOID that promise is replaced by hardware. Our gateway, the server that handles your questions, runs sealed inside an Intel TDX chip, the kind of confidential computing big clouds rent out to their most sensitive customers. The key that opens your question is born inside that chip and never leaves it. We run the server and still cannot read a single question, so there is nothing we could ever hand to a contractor. You can check this yourself. Before a word leaves your device, your browser asks the chip for a report signed by Intel and confirms the chip is genuine, debug access is off, and the code inside is the exact build our multisig recorded on Solana. If any of that fails, your question stays on your device. The model on the other end gets the words of your question and nothing that ties them to you. The gateway's code is public: github.com/paranoiddotred/pa…
🦔404 Media reports that OpenAI pays hundreds of contractors over $50 an hour to read real ChatGPT conversations. The program is called Project Lily internally. Contractors read your prompt, summarize what you were trying to do, and score four AI responses on a scale of 1 to 7. OpenAI says accounts are anonymous, but the contractor dashboard often includes a "user memories summary" that can reveal your location, profession, and personal context. The privacy filter that's supposed to catch personal details misses uncommon identifiers. Some users asked ChatGPT to "keep this between us" without knowing a human would read it. The opt-out is buried in Settings under Data Controls and it only applies to future conversations, not past ones. My Take Every AI company does this. Anthropic does it and says so. Google does it and posts a disclaimer. This is how the models get better. The problem with OpenAI is that they left the setting on by default for 900 million users and couldn't tell 404 Media where they disclosed that humans read the conversations. So it's a disclosure problem, and it's a bigger one now that they're about to ask the public to invest in this company. Most people treat ChatGPT like a private conversation. They talk about personal stuff they wouldn't say to a coworker. A contractor on the program said they don't believe users would imagine someone is reading their chats. Some users asked ChatGPT to keep things private without knowing a person would see it. OpenAI could fix this tomorrow by putting a one-line notice on the chat screen and making the setting off by default. They haven't. If you use ChatGPT, go to Settings, Data Controls, and turn off "Improve the model for everyone." It takes ten seconds. Hedgie🤗 404media.co/inside-project-l…
6
5
27
2,252
Shakedown retweeted
Upside from the 'bear' market drama? We love building in these conditions. $BTC around 69k doesn't mean anything for us building. KOLPERPS.com drops soon & we're cooking. gPERPS
3
6
34
691
Shakedown retweeted
We have already launched PARANOID's MCP server successfully. Claude Code, Cursor and the Claude desktop app can pass any question to PARANOID, and it reaches the model with a zero knowledge ticket instead of your name. PARANOID already works in any browser, and we are now preparing a release that will make it even better to use every day.
Glama, the directory that reviews MCP tools, graded PARANOID's MCP server an A and lists it as official. glama.ai/mcp/servers/paranoi…
11
9
2
43
2,134
Shakedown retweeted
The Architecture Shift Previous iterations required separate domain-specific pipelines. v1.11 replaces that with one unified cross-domain architecture trained from scratch: • Evidence-aware: distinguishes physical models from market odds and human crowds • Spread-aware: ingests member confidence widths and consensus fragility • Surprise-conditioned: models delta against baseline expectations • Direction-aware syntax: understands causality ("heatwave → hospital demand" ≠ "hospital demand → heatwave")
1
4
1
7
372
🐸🐸🐸
TIBBIR’s fee vault woke up. Two treasury keys changed. The fee wallet sent 0.69 ETH to a fresh address. We follow the money and explore the liquidity, deployment and Morpho possibilities. tibbir.nl/articles/tibbir-fe… $TIBBIR
9
174
Shakedown retweeted
Every wallet address you paste into ChatGPT is saved under your name, right next to your email and your card. On PARANOID the AI never even sees it. Before your question leaves your device, your browser swaps each wallet address for a placeholder like [WALLET_1], and when the answer comes back it puts your address back in, right there on your screen. Names, emails, phone and card numbers get the same treatment. The AI answers a question about [WALLET_1], from nobody in particular, and nothing about you is saved anywhere. paranoid.red
2
4
16
1,271
Shakedown retweeted
What if we could tokenize a forecaster? Back a person or system’s ability to understand what comes next. Follow their calls. Inspect their evidence. Share in a clearly defined season’s performance. A proposal for an economy built around the future. 🧵
4
12
1
17
669
Shakedown retweeted
A month ago Vitalik listed what paranoid mode takes: private payments, a hidden address, cover traffic and prompts cleaned up on your own machine. PARANOID has all four in a browser tab. A zero-knowledge proof pays for every question. Three switches do the rest: a relay hides your address in place of Tor, decoy questions go out between your real ones, and names and numbers turn into placeholders before anything leaves your device. This week we tightened the part he was least happy with, the traffic itself. Answers now go out on a fixed clock in frames of one size, with filler while the model thinks, so the wire looks the same whether it is writing or waiting. paranoid.red
Wrapped zkAPI + Tor + extra cover traffic + local LLM prompt rewriting in an open-AI compatible daemon, and made a "paranoid mode" wrapper for my agent harness (pi): Each conversation gets assigned a separate Tor circuit, your requests are masked with random cover traffic, zkAPI ensures you don't leak your identity through payments, and all prompts are locally rewritten (bring-your-own-local-model; I'm using qwen3.8-flash-next) github.com/vbuterin/zkapi/tr… My agent harness is wrapped with a bwrap sandbox (also experimenting with MicroVM) and wrapped again with torsocks, so everything it does goes through Tor. Here's what me asking it to do a simple research task looks like. It's getting quite usable! Though you definitely still pay quite a bit in increased latency, especially when starting a new conversation. Also, Tor is far from optimal here for privacy; the cover traffic mechanism is okay against passive adversaries but much weaker if the LLM provider is colluding with the guard and actively manipulating timing. This kind of task really requires a more specialized mixnet, probably take the tradeoff of a higher bandwidth multiplier and built-in constant cover traffic. Also also, much cheaper zkAPI deposit and withdraw gas costs are coming to mainnet soon.
2
22
1,094
Shakedown retweeted
For everyone new here, here is what PARANOID does and why we built it. Every time you ask an AI something, the company on the other side knows it was you. Your account, your card and your IP address ride along with every question, from health worries to money problems to the business plan you have told nobody about. All of it sits in a log on someone's server, where it can be read, leaked or handed over. Vitalik Buterin and Davide Crapis designed a way out called zkAPI, and the Ethereum Foundation and Open Anonymity built it. You deposit once. After that, every request you send carries a zero knowledge proof that says "this is paid for" while keeping secret which deposit paid. The AI receives a paid question from an anonymous member. If anyone tries to spend the same ticket twice, the math reveals their key and their deposit can be taken, so the service always gets paid and every honest user stays a stranger to it. This is a big deal because private AI has always meant trusting a company's promise. zkAPI turns that promise into math a computer can check: the server confirms every request is paid, and every user looks exactly the same to it. We built the Solana version and open sourced all of it, from the vault program to the circuits to the app. You pay in USDC, PARANOID or ZEC and ask the AI models you already use from a browser tab, with your name left out of the conversation. github.com/paranoiddotred/pa… paranoid.red
Wrapped zkAPI + Tor + extra cover traffic + local LLM prompt rewriting in an open-AI compatible daemon, and made a "paranoid mode" wrapper for my agent harness (pi): Each conversation gets assigned a separate Tor circuit, your requests are masked with random cover traffic, zkAPI ensures you don't leak your identity through payments, and all prompts are locally rewritten (bring-your-own-local-model; I'm using qwen3.8-flash-next) github.com/vbuterin/zkapi/tr… My agent harness is wrapped with a bwrap sandbox (also experimenting with MicroVM) and wrapped again with torsocks, so everything it does goes through Tor. Here's what me asking it to do a simple research task looks like. It's getting quite usable! Though you definitely still pay quite a bit in increased latency, especially when starting a new conversation. Also, Tor is far from optimal here for privacy; the cover traffic mechanism is okay against passive adversaries but much weaker if the LLM provider is colluding with the guard and actively manipulating timing. This kind of task really requires a more specialized mixnet, probably take the tradeoff of a higher bandwidth multiplier and built-in constant cover traffic. Also also, much cheaper zkAPI deposit and withdraw gas costs are coming to mainnet soon.
4
7
23
1,899