ShortSqueezePro retweeted
Get proxies that deliver on speed, privacy, and price. With us, you get reliable performance and global reach.
ShortSqueezePro retweeted
Elara 把稳定币做市包装成了一个生息资产。目标年化 10%–15%,但只预留约 10% 的系统资产应对即时赎回,这两个数字需要放在一起看。它属于 Brila 生态,和 Cyan 的 NFT 借贷、Credit Vaults 的信贷产品并列。Elara 自身的公开融资轮次,这次没有查到;生态内其他产品的历史规模,也不能算到它头上。
产品流程很直白:存入 USDC、USDT 或 USDe,获得 ELUSD;再把 ELUSD 质押成 sELUSD,才开始承担策略盈亏、获取收益。只拿着未质押的 ELUSD,不会自动生息。
钱由谁赚?当前执行伙伴 Arken 把资金放进稳定币交易对的集中流动性区间,收取成交手续费。区间越窄,单位资金利用率越高;遇到稳定币脱锚,资产偏离区间也会更难处理。
我更在意一个容易漏掉的细节:Elara 和 Arken 当前豁免管理费、业绩费,后续会重新收费。因此,即使交易量与策略表现不变,恢复收费也可能压低持有人的净收益,不能直接把今天的展示值当作长期回报。
按目标区间举个例子:1 万美元投入一年,对应 1,000–1,500 美元的目标收益。这只是算术演示,实际表现、未来费用、进出成本都会改变结果。
退出比买入更值得拆。文档说,赎回落在缓冲资金以内可以即时处理,超过缓冲通常约 24 小时;两者都是运营目标,没有写成合约保证。
假设系统资产为 100 万美元,10% 缓冲约为 10 万美元。遇到集中赎回,后面的申请就可能排队;排队请求虽然记录在链上,结算仍要授权角色执行。链上有凭证,不等于持有人能自行完成到账。
目前 ELUSD、sELUSD 没有外部 DEX 流动性,赎回是文档给出的实际退出通道。策略净值损失由 sELUSD 持有人承担,背后没有储备基金或次级份额替你先吸收损失。
想体验可以从官网进入应用:
elara.fi/
连接钱包 → 选择支持的稳定币铸造 ELUSD → 质押为 sELUSD;退出时走解质押和赎回流程。美、英、欧盟/EEA、新加坡用户在限制范围内。
我的判断:Elara 的研究价值在于把主动做市变成普通用户能持有的份额。评估它应同时看扣费后净值和真实赎回耗时;没有稳定币做市经验、又需要随时动用本金的人,很难只凭一个年化数字判断这笔交换是否划算。
#Elara #稳定币
ShortSqueezePro retweeted
Most companies show the finished robot.
XPENG showed the six years of falling.
In the lab. On the factory floor. In front of cameras.
Every hard fall. Every quiet reset. Every small step that finally looked human.
No flashy demo day. No “we’ve solved it.”
Just the long, unglamorous truth: real robots are built the same way people get strong - one failure at a time.
He Xiaopeng and the team kept walking.
Six years later the machine walks with them.
That’s the part most people never film.
ShortSqueezePro retweeted
X上搞钱的几种方式:
1.原创内容奖励:每半个月50–5000U
2.订阅服务:每个会员每月1U起
3.灯塔捡瓶子:每次0.1U
4.撸毛:每次20-10000U,也可能被反撸
5.商务合作:发帖每次50–2000U
6.乞讨:每次10–70000U 不等
ShortSqueezePro retweeted
下周疯狂日程,决定接下来行情走向。
周一 → 中国货币政策公告,
周二 → 美国《清晰法案》投票,
周三 → 美联储FOMC利率决定,
周四 → 日本利率决定,
周五 → 美联储公告。
ShortSqueezePro retweeted
A camera on spider legs sits alone in an empty field, pointed at houses no one lives in.
No operator. No cage. Just optics, antennas, and metal limbs waiting.
Sci-fi spent decades building human faces for the threat. The real version never needed a face.
Give a camera legs and the right to decide when to move, and it stops being a tool. It becomes something that watches.
We prepared for robots that look like us.
The ones arriving look like equipment that learned how to stay.
ShortSqueezePro retweeted
Nidal Wonder walked onto the America's Got Talent stage to battle a humanoid robot and quit after 45 seconds.
Gold curtains. Black marble floor. 2 dancers on it.
One is a teenager who built his whole name on flips. Grey tee, cream pants, curly hair.
The other is a Unitree humanoid in a matching white suit, with a glowing blue visor where a face should be.
The kid moves first. That's the tell.
Kick. Spin. Drop. Years of reps in front of a lens sitting in every angle of it.
The robot answers.
Deep lunge. Leg up. Hold. Kung fu shapes landing clean on the drum track.
Slick marble under it. It never slips.
No breath between moves. No reset. No glance around the room to check if that landed.
The kid knows what it costs to look effortless. He can feel the tank draining. The thing across from him has no tank.
Past 40 seconds, he stops trying.
Arms half raised, he stands and watches it work. Then he walks out of his own shot.
The robot doesn't notice. Arms out, still on beat, still facing a spot where a human used to be.
The kid walked off. The machine kept dancing. Nobody told it to stop.
I don’t look like a champion.
Belly, short breath, last summer’s shorts. People on the shore were already laughing when I stepped onto the first springboard.
Spring. Barrel. Ball. Tire. Banana. Crate. Carrot. Rusty pipes in the middle of the water.
I fell. I got up. I jumped again.
At the end I lifted the trophy. The crowd screamed like I’d won the Olympics.
You don’t need a perfect body or the “right” start. You need to finish when everyone already decided you’d drown on the second obstacle.
#NeverGiveUp #Underdog #WaterChallenge #OldButGold #KeepGoing #ViralVideo #Motivation #FinishWhatYouStarted
I GAVE GROK BOT $54, CLOSED THE LAPTOP, AND WENT TO SLEEP
At 05:30 it handed me an $8,900 balance.
Net profit: $8,846.
I did not leave one bot staring at a chart. Eight agents worked the same live desk through the night:
SEARCH scanned fresh meme signals.
RISK checked contracts, mint authority, liquidity, and exit depth.
SNIPER waited for a cleared entry window and executed it.
WHALE tracked large-wallet movement.
RUG watched the developer wallet for a liquidity touch.
SHILL compared social acceleration with market activity.
EXIT moved stops and took profit through a ladder.
LEDGER kept every position, rejection, and handoff in shared memory.
Every setup moved from agent to agent. SEARCH could not buy. SNIPER could not approve its own risk. RISK could veto the entire chain. EXIT needed a route out before execution was allowed in.
HEAD OF DESK read the combined state and released only the trades that cleared the rules I had approved before going offline.
If a setup failed a contract, wallet, liquidity, or attention check, it died before capital touched it. If it cleared, SNIPER entered, EXIT managed the position, and LEDGER carried the result into the next decision.
By morning the desk had one brief ready:
start: $54
balance: $8,900
net: +$8,846
open risk: reviewed
owner decisions waiting: 1
This is the result of one run, not a promise that the next one repeats it. The useful part is that the agents kept researching, checking, executing, and managing risk while the laptop was closed.
ShortSqueezePro retweeted
A 55kg humanoid robot walked up to a woman dragging a suitcase down subway stairs.
It reached for the handle.
The yellow sign said Watch the Stairs.
Three steps later the robot was face-down on the metal, suitcase still in its grip, legs kicking air.
The woman froze. The robot kept trying to stand while sliding.
This is the same model that does aerial kicks and martial arts demos on stage.
In the real world it still cannot walk down stairs with a bag.
This guy gave Grok Bot a BTC trading strategy and it made $1,907 a day
The account is already at +$57,220 PnL with an average trade of just $36, repeating its setup roughly 25 times every hour across BTC “Up or Down” markets
AI agents combine their signals into one algorithm:
1. Bayesian repricing decides when the position should change
Every new BTC signal updates the model odds:
O(new) = O(old) × LR
If Up remains underpriced, the agents can continue accumulating it
If the edge shifts toward Down, the system can start building the opposite side and convert part of the inventory into complete sets
2. KELLY and TAKER control how the opportunity gets executed
The sizing agent converts the estimated probability edge into an allocation:
f* = (bp - q) / b
TAKER then receives the side, maximum price and allowed size
It can wait for a better fill, execute the first leg, or begin acquiring the opposite outcome when the model sees a better entry
This trader’s Polymarket account: norm1e69
17,702 predictions with a 50% win rate
Most profitable trades:
$882 → $2,088 (+$1,206)
$1,102 → $2,295 (+$1,194)
$1,529 → $2,604 (+$1,075)
The system combines probability, sizing and execution to keep rebuilding the position as BTC moves instead of treating every entry as an isolated market
ShortSqueezePro retweeted
TWO RESEARCHERS FROZE THE WORLD'S WEB SERVERS WITH ONE FORM.
No flood. No botnet. Just math.
In 2011, on stage at a hacker congress in Berlin, Alexander Klink and Julian Wälde sent a single HTTP request to a PHP server.
The form carried 65,536 parameters, and every key was picked so they all collided into one hash bucket.
A hash table is supposed to give O(1) lookups. Under that crafted input it collapses into a linked list - every insert scans the whole chain, and the cost quietly turns into O(n²).
One request pinned a CPU core for about 30 seconds. A few of them stalled an entire cluster, on almost no bandwidth.
The ugly part: this lives inside the simplest "Hello World" app, because the framework silently folds POST data into a hash. PHP, Python, Java, Ruby, ASP.NET, v8 - nearly every language on the web fell to the same trick.
The mechanism had been published in 2003 and patched only in Perl. Everyone else ignored it for eight years.
The fix was almost insultingly small: seed the hash with one secret random number, fresh on every run, so the attacker can never precompute the collisions.
ShortSqueezePro retweeted
OpenAI just dropped GPT-6 Astra.
Anything you can do on a computer, it now does for you.
You speak. It clicks. It types. It builds the deck, books the DMV slot, rewrites the spreadsheet, ships the code.
No more tab switching. No more hunting menus. No more “wait, where was that button.”
One prompt. Full desktop control. Fast.
The promo shows a quiet desk, red lamp, single screen. The cursor moves without a hand. Files rearrange. A presentation appears in under a minute.
They call it the AGI era.
Astra saturates ARC-AGI-3 at 99.9%.
Hits 100% on ExploitBench. Finishes computer tasks in roughly half the time of the previous model.
It is already rolling to limited orgs. Plus and Pro next.
The window is still open.
The model is already inside it.
ShortSqueezePro retweeted
TST 第一个测试币
Broccoli 第一个CZ的狗
Mubarak 第一个中东概念
Giggle 第一个慈善项目
币安人生 第一个破圈中文币
Marscoin 第一个币股meme
一年以来,好像一切都变了,又好像一切都没变
ShortSqueezePro retweeted
THIS DEVELOPER TURNED A TERMINAL INTO A MUSIC PLAYER AND 883,000 PEOPLE WATCHED HER BUILD IT.
she started with code.
then built the whole interface around it - pixel art, animated vinyl, playlists, progress controls, themes.
eventually it became an actual desktop app that can play local music and connect to Spotify, Apple Music and YouTube.
Electron + React + HTML5 Audio.
basically went from terminal -> code -> design -> working product.
and she put the source code on GitHub.
this is the kind of project that makes you want to open a terminal and build something completely unnecessary just because you can.
ShortSqueezePro retweeted
Boston Dynamics retired its hydraulic Atlas on April 16, 2024, and showed an electric one 1 day later.
The farewell video was 3 minutes of failure. Legs buckling. Fluid on the concrete. 11 years of a machine that bled every time it fell.
Then the new one woke up.
It was lying face down, folded flat, legs bent the wrong way. It didn't stand like a person. Its hips spun 180 degrees and drove it off the floor head-first.
No pumps. No lines. No 28 hydraulic joints to keep pressurized.
The 2013 version ran on a tether and walked like a man carrying 2 suitcases uphill. It fell opening doors. It fell crossing rubble. Every fall cost a rebuild.
The electric one doesn't have a human's range of motion. It has more.
That is the part that isn't a stunt. Boston Dynamics stopped building a robot that copies a body and started building one that beats a body at 1 job: moving parts across a factory floor.
Hyundai took 80% of the company in 2021 at a $1.1B valuation. The first units go into a Hyundai plant in Georgia. Not demos. Material handling - lifting sequenced parts, hours at a stretch, no breaks, no shift change.
The backflips were the ad. The 360-degree hip is the product.
The old Atlas was built to look human. The new one stopped trying.
























