x/yoisha

@Railway huh

this shit is so over lmao

Turn any website into an API with /browser-to-api. This skill analyzes network activity, CDP logs, and website behavior to generate a custom OpenAPI spec. Watch Codex one-shot a fully documented OpenTable API client from a single prompt 👀

Raycast 2.0 (beta) is now publicly available for download. 👉 raycast.com/new

What happens when you post a real Monet and say it’s AI? The coolest art social experiment I’ve seen in a while. Thank you @SHL0MS

Bro. Close it. This is not a hack demo. More of a law-breaking speed run. - Crypto payment = banned. up to 3 years jail + fine. - AML law can freeze bank accounts. - promote unlicensed crypto in Thailand or target Thai = gg Be careful out there. Thai law is no joke. Yeah, they might struggle to take down a smart contract, trace ZKPs, freeze non-fiat value capture, or chase anon devs. But pls don't lol. 👀

when you push realism too far using GPT Image 2 x Seedance 2.0 on Higgsfield

@D_studioproject although the thai dub is not accurate, this is so fucking dope!!!!

เนื่องจาก x นั้นมีการให้เงินกับยอด engagement จึงทำให้มีพวกที่ชอบทำการ rage bait ให้เราอยากไป quote ด่า หรือให้มีการมาแชร์ต่อเยอะๆ ซึ่งแนะนำให้ว่าถ้าจะด่าอะไรแบบนี้ไป Crop รูปมาด่าดีกว่าครับอย่าไปให้ยอด engagement การกระทำแบบนี้ เพราะการไปด่ากลับ account พวกนี้ จริงๆคือช่วย feed algorithm ให้เขาได้เงินเพิ่มนะ หลาย account ตอนนี้ใช้ fearbait / drama bait / misinformation เพื่อ farm engagement เพราะยิ่งคนโกรธ ยิ่ง share ยิ่งเมนต์ ก็ยิ่ง monetize ได้ สุดท้ายคนที่เสียคือ timeline ของเราเองที่เต็มไปด้วย content toxic แบบนี้ ดีที่สุดคือ block, report, และอย่าแจก reach ฟรีให้ account ที่ตั้งใจทำตัวน่าเกลียดเพื่อเอา engagement

Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.

1/ Today marks 1 year since Initia mainnet went live. To the builders who launched appchains, the users using them, the creators who told the story, the investors who backed it, and the community pillars who kept showing up: thank you. ↓

Initia's investors, founders, and team have chosen to delay their vesting cliff by 1 year to April 2027. We are grateful Initia's backers and contributors have a long term belief in the project and token. Over the next year, we're focused on apps and directing value to INIT

With the Kelp exploit yesterday, I wanted to provide a quick summary for folks who may be confused or simply looking for compiling what we know. Firstly, my employer would be remiss if I didn’t lead with: MegaETH does not appear to have direct exposure to the incident; I could only locate ~$100 of rsETH on the entire chain. Summary of events: Earlier yesterday, an invalid message that supposedly originated from a smaller L2 was accepted by the Kelp DAO DVN for their LayerZero OFT. As a reminder, L0 has a variety of security setups, which can be on par with or surpassing alternatives, but in practice many teams choose a lower burden. Kelp DAO's controlled contract accepted the invalid first transaction, but blocked a follow-on for another tranche of rsETH. Many teams (including Ethena, who operate USDM) have paused their OFT bridging while an investigation occurs, but as of right now this looks to me like it was a compromise on the Kelp DAO side, and the security assumptions did not require further compromise to execute malicious actions. rsETH is lock-and-mint on mainnet. This means when L0's rsETH goes between L2s, it is burn-and-mint, but when Ethereum is one end of a corridor, there is an rsETH token that is being (un)locked. For example, rsETH traveling from Optimism to Mega would burn 1 rsETH on Optimism for every 1 rsETH minted on Mega. For rsETH traveling from Optimism to Ethereum, however, 1 rsETH on Optimism would be burned and 1 rsETH would be unlocked from the mainnet escrow. In theory, there should be 1 rsETH in the mainnet escrow for every rsETH outside of Ethereum. That is now no longer the case. Somewhere between 17-18% of rsETH supply was unlocked erroneously by the attacker. That rsETH was then swapped on DEXes and borrowed against on lending protocols, including Aave, Compound, and Euler. All of this was done on Ethereum and Arbitrum to the best of my knowledge. The problem set: There are three major, interconnected problems that I currently see. First is whether rsETH is now ~17.5% undercollateralized due to the inflated supply (L1 + L2). Kelp looks like they have two paths they can take. The first is to socialize the dilution across all rsETH holders, and try to recover what it can through cooperation with DeFi teams and law enforcement. The second is to keep the mainnet supply whole, since all rsETH on mainnet are, in theory, the actual assets, and the number of rsETH on mainnet did not chain. This would effectively be pushing all losses onto L2 rsETH holders. I expect many lawyers are being activated already, since there is not clear seniority or differentiation that I can find in the Kelp documentation about this. Second is whether lending protocols now have accrued large amounts of bad debt. I want to be perfectly candid that my initial estimates are in excess of the buffers most of these protocols have in place. For example, should all rsETH holders be diluted 17%, then mainnet Aave alone looks like it has around 2x the bad debt as can be covered by Umbrella, which is more than $50m of coverage. Complicating things is that rsETH losses socialized only among L2 holders would be dramatically larger, and could still result in some Arbitrum deployments becoming insolvent. It will take some time for each lending protocol to even know which scenario would be better for them (my guess is Compound and Aave would like to see it contained to L2s, while Euler and Fluid and others who compete more in frontier markets would want the universal haircut), and then advocate for that solution to Kelp. Thirdly, while all of this is ongoing, the WETH and stablecoin markets are largely illiquid. This creates a lot of negative carry for some highly leveraged trades. Until those trades can unwind, that risks some positions slipping into liquidation at a time that some of these markets will not be especially liquid. 1/2

6 months ago KuCoin admitted fault for my $300K liquidation. Here's what happened since. I took a Google Meet with their Head of Futures. I went to an in person meeting at Tribes in Dubai Mall with their Global Business Director. I sent 10+ proposals. I gave them every possible way to make this right. On the call they took full responsibility. They admitted the liquidation was caused by broken infrastructure. Their platform failed and they said so themselves. But here's the part that's hard to believe. Their Head of Futures couldn't understand basic futures mechanics. I had to explain how margin, liquidation and order book depth works to the person running the futures division at a top 10 exchange. The person responsible for resolving my case didn't understand the product that caused it. Their first offer: bring us $2.5 billion in trading volume and you can "earn it back." I did the math for them live in the chat. $10,000 per 100M volume. That's 0.01% return. To recover $250K I would need to generate the monthly volume of a top 50 institutional desk. For free. I said no. Their second offer was worse. $20K upfront, but only if I hit 1,000 active users and $300M in volume first. Then a $30K "cashback" that requires KuCoin's manual approval. I said no again. Their third offer was even worse than the second. $10K/month. Halved the numbers from the deal they already couldn't close. After an in person meeting. After a Google Meet. After weeks of negotiations. Every single offer came with the same condition: delete the tweets, stop talking, and come work for us as a KOL. Promote the exchange that wrongfully liquidated me. Bring them users. Make them money. Then maybe they'd consider giving back what they took. I told them in the chat: "It's like someone steal from me $250K and then tells me come work for me and you'll make it back (maybe)." Their response? "Let me think about it." Then silence. Weeks of silence. I had to chase them for every single reply. Christmas came and went and I gave them a final deadline January 6th. They came back with yet another lowball. KuCoin had their Head of Futures, their Global Business Director, and multiple senior reps in this group chat. They all saw every message. They all went quiet when it mattered. Today I'm releasing the full 30 minute Google Meet recording and the complete Telegram history. Every message. Every offer. Every time they went silent. You'll hear them admit fault and then watch them do nothing about it. They had 6 months to make this right. They chose silence. Video drops today.

my brain's context window ran out before my agents did. so I put a little pixel buddy in my notch and it watches all of them for me. it tracks claude code, codex, gemini, cursor, all of them. when one needs approval the notch expands and I can just hit allow without leaving what I'm doing. plays 8-bit sounds. click any finished task and it jumps to the exact terminal tab, even the right pane. been running 5-8 agents daily for two months and honestly this is the only reason I haven't lost my mind.

@imedwardluo i just came across your app, the design is amazing!

最近每天烧几亿 Tokens，做了一款很有趣的 Mac「灵动岛」App - 👾 Vibe Island. Vibe Island 的灵感来源于我在 Vibe Coding 会同时开 5-10 个对话跑不同的任务，经常切着切着就忘了刚刚那个对话跑到哪了...任务做没做完、需不需要我审批等等问题。 感觉人类大脑的上下文窗口已经远远不够用了。 于是，我就给这些 Agents 造了一座灵动岛！ 装上 Vibe Island 之后：Mac 刘海区域会展示各种 Agents (Claude Code / Codex / OpenClaw 等等）的实时工作状态： - 当 Agent 需要权限审批或任务完成时，灵动岛会自动展开，不用切回终端就能操作 - 面板聚合了所有正在运行的 Agents 对话，点击卡片，精准跳回到会话所在的终端界面 - 所有状态都配有定制化的 8 bit 像素风游戏音效，狠狠提升了 Vibe Coding 的氛围感 灵动岛是一个极其优雅的形态：既能展示后台任务的状态，又不打扰前台正在进行的任务，它只在需要你注意的时候弹出，不用切换应用就能完成轻量交互，还带有一点陪伴感。 其实最早这个 App 只能触发右上角的系统通知，后来发现只做通知远远不够，实际应该解决的是：如何在恰当的时刻，以最小的摩擦，把注意力带回到正确的位置。 欢迎前往官网下载体验👇🏻 vibeisland.app

可能是这几天熬夜没睡觉做产品脑子做傻了...想补一条 Vibe Island 新帖子的时候，不小心点错删掉了自己有生以来最火的一条 30w+ 的帖子... 😭 发之前完全没预料到帖子会这么火，还有这么多大佬都自发帮我转发，真的非常感激各位大佬，也非常感谢大家喜欢我的产品，容纳它还有很多瑕疵和小问题。我在疯狂修 bug 优化中，很快就会上线更多有趣有用的功能！ 只保留了一张截图纪念，原贴内容如下 👇🏻

Strait of Hormuz: A CitriniResearch Field Trip The Field Report from Analyst #3 is live. citriniresearch.com/p/strait-of-ho…