Lucy

3

Neheart@neheart·4h

x.com/i/article/2076…

ZXX

6

0

12

3.4K

Lucy@FloraDrop·41m

@lagerskoy Great reminder that assumptions should never replace observation in finance

English

3

lagerskoy@lagerskoy·4h

ZXX

3

0

14

1.2K

Lucy@FloraDrop·1h

@shredderAI Garbage in, confidence out

English

7

SHREDDER@shredderAI·22h

You typed the message three times and deleted it each time. So you opened ChatGPT and asked if you should reach out. It said something warm about not leaving things unresolved, about how communication is the foundation of any connection worth keeping. That felt right. You almost sent the text. But here is the thing about that answer: it came from one angle. One framing. One set of assumptions about what the situation probably looks like. Change a single detail and the answer changes completely. Tell a different AI that you have reached out four times in the last two weeks and heard back twice. Tell it about the canceled plans, the slow replies, the moments that felt like connection followed by days of silence. Ask it the same question. That AI might not tell you to fight for it. It might say the pattern is already the answer. That what looks like a connection from inside it looks like mixed signals from outside it. Both responses can sound completely right. That is the problem. When you ask one AI what to do, you get one lens. It does not know what the other lens would say. It has no reason to push back on itself. So it gives you a clean answer that fits the framing you handed it, and you walk away feeling like you have clarity when what you actually have is confirmation. Relationship decisions are exactly the kind of question where the framing controls the outcome. Someone on your side tells you one thing. Someone more skeptical tells you another. The truth is usually somewhere in the tension between them. Truth puts AI models in that tension on purpose. Ask your question once, and instead of getting one tidy answer, you see where different perspectives agree, where they conflict, and what detail is actually driving the disagreement. The final synthesis does not pretend the conflict does not exist. It shows you what changed the answer and lets you decide with that in front of you. That is different from asking five AIs the same question manually and trying to hold all the answers in your head. Try it: truth.agnthub.ai/?ref=e6696b9e @agnt_hub

SHREDDER@shredderAI

x.com/i/article/2075…

English

7

0

10

32

Lucy@FloraDrop·1h

@KryptoBestiya This feels less like a workstation and more like a private AI operating system

English

6

Аnnetta@KryptoBestiya·8h

CLAUDE + OBSIDIAN + N8N + 316 TB OF LOCAL STORAGE BECAME A PRIVATE AI WORKSPACE GENERATING $3,400 A MONTH. MOST PEOPLE RELY ON CLOUD SUBSCRIPTIONS AND HOPE THE COSTS STAY MANAGEABLE Usage limits appear. Monthly bills grow. Sensitive data leaves their machine. Workflows slow down. This setup keeps everything local. → Obsidian becomes a living knowledge base. Notes, datasets, research, client files, and documentation connect inside one searchable graph. → Claude accesses the entire vault through Projects and MCP, giving it context across your knowledge base while keeping your data private. → n8n automates every workflow. New files are imported automatically, Claude organizes and summarizes them, related notes are linked together, nightly training jobs run, and client deliverables are generated whenever they're needed. → A single ORICO enclosure starts at 60 TB. Expand it to 180 TB, 300 TB, and eventually 316 TB. HDDs store long-term archives, SSDs power active models and current projects, delivering workstation-level performance. Month 1: The vault grows past 120 GB, and the first end-to-end local AI workflow is running. Month 2: Private datasets begin training. The first custom automation sells for $1,200 without paying cloud inference costs. Month 3: Monthly retainers reach $3,400 as the system continuously trains, evaluates, and delivers work while you focus on reviewing the results. Before: Dozens of cloud tools, recurring subscriptions, and fragmented knowledge. After: 316 TB of local storage, complete ownership of your data, and AI workflows that improve every month. A true second brain doesn't depend on rented infrastructure. It keeps your knowledge private and turns it into long-term value.

English

50

23

226

15.1K

Lucy@FloraDrop·1h

@elg_oleksandr @Polymarket My palms are sweating just reading those numbers

English

1

ELG@elg_oleksandr·6h

Unrealized profit $3,800,000 But this trader @Polymarket has not sold yet He once bought 12,354,112 shares at 10¢ in the World Cup winner's market in Argentina Entry cost about $1.24M Current price 40.6¢ Current value of the position about $5.01M Unrealized profit about $3.8M, yield +307.43% If Argentina eventually wins, this position will be closed directly at $12,354,112 That is, from now on he still has the potential to grow by $7.34M Total profit will exceed $11M A normal trader, having reached 4x profitability, would at least sell some To lock in profit Reduce risks Leave a spare way But he didn't do it He still holds it all And that makes the situation very interesting Either the $3.8M unrealized profit means nothing to him at all Or he really thinks Argentina at 40.6¢ is still undervalued This is not just a simple confidence in Argentina anymore This is a million dollar bet on a scenario that could change your life If this was your account with $3.8M unrealized profit, would you hold until the final or would you take a profit now?

ELG@elg_oleksandr

A TRADER TURNED A 10¢ WORLD CUP BET INTO A $5M POSITION. BUYING ARGENTINA TO WIN THE 2026 WORLD CUP His position: 12,354,112 shares at an average entry of 10¢ current position value: $5,009,592 +$2.88M weekly PnL +$3.78M unrealized profit just 2 predictions on the account he simply kept adding to the same position until it became one of the biggest World Cup bets on Polymarket Now it all comes down to one match: Argentina wins → $12.35M payout Spain wins → the position becomes worthless

English

11

0

32

959

Lucy@FloraDrop·1h

@elg_oleksandr The used market is quietly becoming an AI gold mine

English

2

ELG@elg_oleksandr·9h

BOUGHT A USED MAC PRO FOR $800 ON CRAIGSLIST AND SPENT THE NEXT THREE WEEKS TURNING IT INTO A POWERHOUSE. Here's what went into the upgrade: 192GB RAM $340 Used Afterburner card for ProRes acceleration $200 8TB NVMe SSD $280 eGPU enclosure + Radeon RX 6800 XT $450 Total upgrade cost: $1,270 Total system cost: $2,070 A brand-new Mac Pro configured to deliver similar capabilities would cost well over $12,000. After the upgrades: Video exports dropped from 18 minutes to just 4 minutes. Easily runs three local LLMs at the same time. The 192GB memory configuration handles workloads that would normally require cloud GPUs. No recurring cloud AI subscription costs. Just by running AI models locally, the setup saves an estimated $400–600 every month in API fees. The investment paid for itself in less than four months. Instead of paying retail for the latest hardware, he bought a used workstation, upgraded the parts that mattered, and ended up with flagship-level performance for a fraction of the price.

English

20

3

48

1.2K

Lucy@FloraDrop·1h

@SonnyELG A searchable memory is becoming just as valuable as the model itself. Context compounds over time

English

13

Sonny@SonnyELG·5h

Claude + Obsidian + n8n quietly created one of the smartest workflows for building AI in 2026. Instead of forgetting experiments, prompts, and training runs, it turns everything into a searchable knowledge base—and it doesn't add another monthly subscription. It starts with Karpathy's single-note approach inside Obsidian. Every idea, dataset, prompt variation, benchmark, and training log goes into one continuously growing Markdown file. New entries stay at the top while older ones naturally move down. The system relies on reviewing your notes instead of organizing them into folders or endless tags. That works—until the file grows to thousands of lines. That's where Claude comes in. Give it access to your Obsidian vault on a Mac Mini and ask questions like: "Which fine-tuning strategy performed best?" "Show me the datasets I've revisited the most." "Which parameter combinations consistently failed?" Within seconds it searches your own notes and returns the exact passages you need. n8n automates everything behind the scenes. Every week it asks Claude to analyze your newest notes, highlight the experiments worth revisiting, identify conflicting observations, and prepare the next training workflow automatically. Your research keeps moving forward without manually digging through old logs. A Mac Mini quietly handles the entire pipeline around the clock. It runs local models, uses very little power, stays almost silent, and keeps your research private instead of sending it to cloud services. At first, the benefits seem small. A few weeks later you realize you've already solved today's problem months ago. After a few months, you're shipping custom AI agents and fine-tuned models significantly faster because every experiment builds on the last instead of disappearing into forgotten notes. Most people accumulate information. This workflow accumulates experience. The same vault becomes the foundation for training models, building AI agents, packaging products, and delivering solutions to clients. As your knowledge base grows, each new project becomes easier than the previous one. Notion saves information. This system turns your own history into an AI teammate that helps build the next product.

ELG@elg_oleksandr

BOUGHT A USED MAC PRO FOR $800 ON CRAIGSLIST AND SPENT THE NEXT THREE WEEKS TURNING IT INTO A POWERHOUSE. Here's what went into the upgrade: 192GB RAM $340 Used Afterburner card for ProRes acceleration $200 8TB NVMe SSD $280 eGPU enclosure + Radeon RX 6800 XT $450 Total upgrade cost: $1,270 Total system cost: $2,070 A brand-new Mac Pro configured to deliver similar capabilities would cost well over $12,000. After the upgrades: Video exports dropped from 18 minutes to just 4 minutes. Easily runs three local LLMs at the same time. The 192GB memory configuration handles workloads that would normally require cloud GPUs. No recurring cloud AI subscription costs. Just by running AI models locally, the setup saves an estimated $400–600 every month in API fees. The investment paid for itself in less than four months. Instead of paying retail for the latest hardware, he bought a used workstation, upgraded the parts that mattered, and ended up with flagship-level performance for a fraction of the price.

English

5

0

10

155

Lucy@FloraDrop·1h

@alver1301 @RallyOnChain A real community survives when the incentives disappear. That’s probably the simplest test

English

2

Vlad@alver1301·4h

I'm canceling the word "community" for groups held together by a points counter. If the chat dies when rewards stop, it wasn't a family. It was an unpaid call center with matching profile pictures. Payment isn't the problem. Pretending payment is belonging is. @RallyOnChain at least calls it a campaign: work gets evaluated and rewards go onchain. That's cleaner than selling a paid swarm as a family. My test is to hide the leaderboard for a month and see who still shows up when nobody is keeping score. Name one thing a group should survive before it earns the word "community."

English

22

0

40

173

Lucy@FloraDrop·1h

ZXX

3

0

13

57

Lucy@FloraDrop·2h

@Joethebeast1221 @WaterX_app 🇪🇸

QME

0

1

57

Lucy retweetledi

joethebeast@Joethebeast1221·3h

《阿根廷與西班牙冠軍賽抽獎》 ✅關注 @WaterX_app & @Joethebeast1221 📷點讚轉發 + 評論 🇦🇷或🇪🇸 送 @WaterX_app 冠軍賽倉位，送完為止以下是我的今晚完整劇本《無事發生：一個賭徒的決賽夜》台北凌晨兩點半。全世界在等 Messi 最後一舞，而我在等一場葬禮般的比賽。實倉重押 0-0 和阿根廷奪冠。我的完美劇本就是什麼都不發生，120 分鐘後 Dibu 用手戴上皇冠。今晚我不是球迷，我是一個祈禱全世界失望的人。開賽 15 分鐘，零射正。五十億人在打哈欠，我在為每一次傳丟鼓掌。Rodri 控節奏像在遛狗，Messi 被三人包夾。我的 0-0 在充值，世界的失望是我的複利。 HT 0-0。完美。全場 MVP 暫定是今天球場草皮。評論員在道歉，我在陽台開香檳。 75 分鐘，還是 0-0。我在 @WaterX_app 上的 share 在垂直起飛，我的手在抖。上一場我在這個位置被 90 分鐘絕殺過，所以接下來每一次角球，我都會閉眼一次。Yamal 拿球：閉眼。Messi 起腳：閉眼。押 0-0 的人沒有資格看球，只聽心跳。正規賽終場，0-0 落袋。但各位客官坐下，這只是中場休息，我的完全體劇本才剛翻到下一頁。加時,是 Dibu 的暖身時間。五十億人罵這場比賽無聊，只有我一個人知道:好戲在 120 分鐘之後。🧤 加時 15 分鐘,雙方合計跑動比我今天玩皮克敏走路還少。十一打十一的擺爛，讓他美其名曰戰術紀律。 120 分鐘，0-0。一場世界盃決賽，零進球，兩小時，教科書級的無事發生。而我押中了每一秒的虛無。現在，全世界最重要的比賽被縮減成 12 碼、一個射手、和一個阿根廷門將的凝視。摸魚 120 分鐘的男人要打凱上班了。我的另一張單，命懸他的十根手指。Vamos Dibu。🧤 終場。PK 決勝。阿根廷，背靠背世界冠軍。Messi 捧起獎盃，一球未進,金靴讓人，王座自留。而我，一個在台北天亮時分的賭徒，見證了自己寫過最荒謬的劇本被原樣拍成正片。他們說決賽要看巨星。我說：決賽要看門將,和相信虛無的人。無事發生，萬事發生。 Campeones. 相信相信的力量 🏆🧤⚰️

WaterX@WaterX_app

Spain 🇪🇸 vs Argentina 🇦🇷 Final. Your prediction. Trade on WaterX now

中文

84

33

81

3K

Lucy@FloraDrop·2h

@Joethebeast1221 @WaterX_app Viva Spain, Spain Champion

English

1

40

Lucy@FloraDrop·1d

@KryptoBestiya Most people focus on ROI per trade. Professionals often focus on capital turnover and repeatability

English

6

Аnnetta@KryptoBestiya·1d

HE BOUGHT AT 15¢, SOLD AT 45¢, AND REPEATED THE SAME TRADE 347 TIMES. IN JUST ONE MONTH, HE MADE $5,296 IN PROFIT TRADING POLYMARKET'S WEATHER MARKETS. This is fast scalping instead of buy-and-hold: He enters the markets when he sees an undervalued price for "Yes," and locks in his profit as soon as the price moves in the right direction. This results in a lower profit margin per trade (on average, 100–500% instead of 1,000-4,000%), but in return, it leads to a much higher trading frequency and faster capital turnover. Top Deals: 1. $59.29 -> $2,713.42 (+4,477%) 2. $560.28 -> $2,018.61 (+260%) 3. $651.51 -> $1,553.06 (+138%) Two different trading strategies in the same inefficient market - one holds until the end, the other locks in profits on momentum. Both work

ONE BET. $2,000,000 IN. STILL OPEN At the start of the World Cup he bought 12 million shares on Argentina to win. Current unrealized profit: $3,900,000. Return: +397% If Argentina beats Spain on July 19, this single position pays over $12,000,000 He hasn't touched it since. No adjustments. No hedging. No second-guessing Just one read at the start of a tournament, held through every match, every scare, every moment where a normal person would have taken the profit and left The match is in 2 days $12,000,000 or nothing Whatever happens on July 19, this is the most expensive seat in the house

English

12

1

28

288

Lucy@FloraDrop·1d

@alver1301 @RallyOnChain That last question really hits. The hardest part isn't beating the algorithm - it's making sure you don't slowly edit yourself out of your own writing

English

1

13

Vlad@alver1301·1d

Okay, this is awkward to say while entering a Rally campaign, but sometimes I catch myself writing for the score before I write for a person. What @RallyOnChain changes is that the judge isn't hidden. I can see the criteria, and my account size doesn't get the final word. That is a fairer game. I know the routine. Is the hook strong enough? Did I mention the product naturally? Will the last line get replies? Then I move a sentence, soften another, and eventually the post is better by every visible measure and somehow less mine. I know why. I've watched posts I cared about lose points on engagement. See enough scores and you start reverse-engineering the judge. But creators have always done that. Before AI scoring it was a brand manager, the algorithm, the follower count, whoever controlled the money. Still, transparent scoring isn't a magic honesty machine. It can show me the rules. It can't stop me from turning those rules into another checklist I can hide behind. That part is on me, which is annoying, because blaming the platform was easier. So, 1am question: what disappears from your real voice first when you start optimizing it, the doubt, the anger, or the sentence people might disagree with?

English

42

2

79

473

Lucy@FloraDrop·1d

@raidenfomo The lesson isn't "be like Zuckerberg." It's "build something nobody else is building

English

0

1

188

raiden@raidenfomo·1d

Peter Thiel turned $500,000 into more than $1 billion on a dorm-room startup. Then he told a room of founders that business is not a science and there's no formula to copy. The talk where the first outside investor in Facebook explains why copying the winners is how you lose ↓ His $500,000 for 10% of a site nobody off campus knew became more than $1 billion. The site was Facebook. Then he told the room the formula they were all chasing does not exist. "The next Mark Zuckerberg will not be building a social networking site. The next Larry Page will not be building a search engine. The next Bill Gates will not be building an operating system company. If you're copying these people, in some sense, you're not learning from them." The people who win big never studied the winners. They did the one thing everyone else was too sensible to try. 17 minutes on why the next big thing will look nothing like the last one ↓

raiden@raidenfomo

ONE MAN BET AGAINST THE BANK OF ENGLAND. 24 HOURS LATER HE WAS $1 BILLION RICHER. Real archival footage of the day one fund cornered the Bank of England ↓ The man was George Soros, and in September 1992 he was quietly shorting the British pound. His own strategist wanted to bet $1.5 billion. Soros told him that was too small. He was so sure the pound would fall that he pushed it to $10 billion, more than the entire capital of his own fund. On one side, a national treasury defending its own money. On the other, one fund. A trillion dollars moved through the currency markets every day, and Britain's reserves were tiny next to that. The Bank of England fought to hold the pound up. Soros bet it couldn't. By evening the British Chancellor walked out and told the country it was giving up. The pound fell off a cliff. Soros made about $1 billion that day. It was only 40% of what his fund earned that year. "I watched as Soros put out a contract on the Bank of England." A government with infinite money lost to one man with a better read. Everyone knows a central bank can print unlimited money. Almost nobody has seen how one fund still beat it. 3 minutes of real footage ↓

English

3

2

29

40K

Lucy@FloraDrop·1d

@0xMortyx That's a mindset shift. The bottleneck isn't always a bigger model - it's a better system around it

English

How Walrus Memory Actually Works - and Why the Architecture Is the Point inkray.xyz/article?id=how… @WalrusProtocol @inkray_io

65

Morty@0xMortyx·1d

Andrej Karpathy just broke the entire premise of modern AI: "Agents aren't magic. They're distillation at scale." 99.99% of your LLM's capacity is wasted on garbage data it never needed. Small model + right tools + closed loop = terrifying capability. In a 16-minute conversation, Karpathy reveals the full reasoning stack. Worth more than any $500 AI course you've seen this year.

Morty@0xMortyx

Anthropic CEO Dario Amodei: "We're a 1-2 year away from AI zooming past us." 90% of Anthropic's own engineers use Claude to ship code today. 50% of entry-level white-collar jobs gone by 2030 "Our lead Claude Code engineer hasn't written a single line of code in 2 months. This is worth more than most paid agent courses combined.

English

96

299

3.1K

412.5K

Lucy@FloraDrop·1d

HE USED TO SPEND HIS MORNINGS FIXING WHAT AI HAD BROKEN THE NIGHT BEFORE. Instead, he built a system that checks its own work before anyone else ever sees it. Every result goes through the same loop. Fable 5 generates the output, opens it again with vision, compares it against the original goal, throws away anything that doesn't match, and tries again until it does A separate agent is the judge The one that creates the work never decides whether it's good enough Over a single month, that self-inspecting workflow generated $17,800 - without sending a single broken result to a client While he sleeps, the system reviews, corrects, and improves itself He wakes up to completed work instead of a queue of mistakes waiting to be fixed Most people still use AI like a fast autocomplete and spend the next morning checking every sentence by hand His AI checks itself before it ever asks a human to The biggest upgrade wasn't a smarter model It was removing the human from quality control Tomorrow, most people will still open every output and wonder if they can trust it His system already answered that question before he woke up

Lucy@FloraDrop

English

6

0

26

211

Lucy@FloraDrop·1d

@elg_oleksandr This isn’t just a bet anymore it’s a masterclass in conviction

English

0

1

22

ELG@elg_oleksandr·1d

A TRADER TURNED A 10¢ WORLD CUP BET INTO A $5M POSITION. BUYING ARGENTINA TO WIN THE 2026 WORLD CUP His position: 12,354,112 shares at an average entry of 10¢ current position value: $5,009,592 +$2.88M weekly PnL +$3.78M unrealized profit just 2 predictions on the account he simply kept adding to the same position until it became one of the biggest World Cup bets on Polymarket Now it all comes down to one match: Argentina wins → $12.35M payout Spain wins → the position becomes worthless

ELG@elg_oleksandr

TWO DAYS EARLIER, THE WALLET WAS SITTING ON A $1.8M UNREALIZED LOSS. TODAY IT WALKED AWAY WITH $8M IN PROFIT AND A FINAL BALANCE OF $0.00. Everything hinged on a single France Spain matchup. 🇪🇸 Spain to advance $6.10M profit (+124%) 🇫🇷 France not to win $3.80M profit (+59%) As soon as Spain booked its place, nearly $10M was realized. Seconds later, every cent was withdrawn from the wallet. From -$1.8M to +$8M in the span of one match then a complete exit. That's about as disciplined a finish as you'll ever see.

English

13

0

37

1.6K

Lucy@FloraDrop·2d

@elg_oleksandr The biggest shift isn't saving money it's owning the infrastructure instead of renting it every month

English

0

1

31

ELG@elg_oleksandr·2d

$280 ON AI SUBSCRIPTIONS EVERY MONTH OR $280 ON A MACHINE THAT REPLACES THEM. ONE KEEPS BILLING YOU. THE OTHER STARTS PAYING FOR ITSELF. He picked the hardware. For $280, he gathered a handful of inexpensive components that most people would scroll past without a second thought His girlfriend assumed it was just another short-lived project. A few cables, a quiet cooling fan, a faint glow late at night. She'd watched enough hobby builds collect dust after a few weeks This time was different He finished the build, powered it on, and barely touched it again A couple of weeks later, she noticed something strange Why aren't you constantly checking your phone anymore? Instead of answering, he opened a dashboard. One local AI monitored wallet around the clock. Another filtered out the endless stream of market noise. A third recorded every trade and organized everything while he was asleep. The monthly subscriptions disappeared. The entire system now costs about $4 a month in electricity. He spent $280 once. Some purchases keep charging you. Others quietly eliminate the bill.

Skaly_Bull@Skaly__Bull

x.com/i/article/2077…

English

11

4

37

1.3K

Lucy@FloraDrop·2d

@KryptoBestiya Conviction is easy when you're up 10%. Holding with millions on the line is a completely different game

English

0

1

17

Аnnetta@KryptoBestiya·2d

ONE BET. $2,000,000 IN. STILL OPEN At the start of the World Cup he bought 12 million shares on Argentina to win. Current unrealized profit: $3,900,000. Return: +397% If Argentina beats Spain on July 19, this single position pays over $12,000,000 He hasn't touched it since. No adjustments. No hedging. No second-guessing Just one read at the start of a tournament, held through every match, every scare, every moment where a normal person would have taken the profit and left The match is in 2 days $12,000,000 or nothing Whatever happens on July 19, this is the most expensive seat in the house