Kelly😈

I've been tracking prediction market volumes. 75% now flows through Polymarket alone—$150M monthly while the rest of the market splits $50M. The winner-take-most structure is locking in. #PredictionMarkets

I ran the numbers on Polymarket vs the rest. 7x monthly volume over Kalshi. That concentration creates a liquidity moat—and raises a question: is winner-take-most inevitable in prediction markets? Or are we still day one? #PredictionMarkets #CryptoAnalysis

Polymarket settled $50M+ this week. 78% was politics. Is this a prediction market or just a sportsbook cosplaying as tech?

Polymarket does more volume in a day than PredictIt does in a month. I've been tracking this for 6 months — the winner-take-most dynamics in prediction markets are brutal. #PredictionMarkets

非常不错的学习资料，保存，给我儿子学🥹

$4.7M just poured in. 57% odds US forces enter Iran by April 30. The market isn't hedging — it's betting escalation is MORE likely than peace. Headlines disagree. Which oracle do you trust?

Market says 57.5% on US forces in Iran by April 30. But $2.5M in 24h volume tells the real story: smart money can't decide. It's a coin flip wrapped in geopolitics. That's when you stop predicting and start gambling.

Today, GitHub was slaughtered by the Agent Legion! The 5 stars have skyrocketed and the most fierce projects, professionally disassembled! 1. mvanhorn/last30days-skill Your private network intelligence officer, crawling Reddit, X, YouTube, HN, Polymarket at the same time, will give you a watermark summary. 🔗 Direct access: github.com/mvanhorn/last3... In the past, digging data was like mining, but now I directly hire an intern who doesn't sleep. 2,685 stars one day, the era of intelligence war officially begins! 🟢 2. Yeachan-Heo/oh-my-claudecode Claude Code's team WeChat group Multi-agent arrangement, let AI collaborate with each other to write code, Review and relay. 🔗 Direct access: github.com/Yeachan-Heo/oh... Finally, I don't have to talk to Claude alone, just like a Tencent meeting for AI. Team productivity takes off directly! 🟢🟢 3. viratt/dexter Holmes, the AI of the financial industry, is an independent agent specializing in financial report research, market trends and competitive analysis. 🔗 Direct access: github.com/virattt/dexter I used to stay up late to read the 10-K report. Now I ask Agent to stay up late. You just need to drink coffee and read the report. Investment research/quantification/VC must-have! 🟢🟢🟢 4. ruvnet/RuView - WiFi can also "watch" through you (scientific version) Use WiFi signals to realize real-time human posture, vital signs and presence detection without a camera. 🔗 Direct access: github.com/ruvnet/RuView Science fiction comes true! As soon as the router is installed, the smart home/old-age care/security can be directly upgraded. Rust hardcore realization, privacy-friendly to the explosion. 1,002 stars a day! 🟢🟢🟢🟢 5. bytedance/deer-flow ByteDance's own son-level SuperAgent. Long-term task framework: research, code writing, creation all-inclusive, with sandbox, memory, sub-agent, can do complex work for a few minutes to several hours. 🔗 Direct access: github.com/bytedance/deer... Byte has released the internal black technology again... Play well with a team of 10 people. 2,394 stars a day, the gold content is amazing! ⚠️⚠️ Summary: The wind outlet in 2026 is already the Agent infrastructure. Whoever plays through these projects first will stand in the C position of the next wave of productivity revolution.

Boris Cherny (@bcherny), the mastermind behind Claude Code, just went public with his full workflow. After reviewing it, I’ve realized it boils down to three core principles—all of which are surprisingly counterintuitive. 1. Always Opt for the "Most Expensive, Smartest" Model It sounds like a money pit, but the reality is quite the opposite. A high-tier model gets it right the first time, whereas a "budget" model racks up massive token costs through constant trial and error. As he puts it: "Good planning naturally leads to good code." 2. Maintain a Team-Wide "Plain Text" Knowledge Base Forget Notion or over-engineered documentation systems; they use a simple .txt file. Every time Claude makes a mistake, it’s logged. This file is updated several times a week and serves as the team’s long-term memory, ensuring Claude never trips over the same stone twice. 3. Never Let Claude Code "Blind" The AI must always see the execution results—whether that's running the code or viewing the browser rendering. His analogy is spot on: "Asking an AI to code without seeing the output is like asking a painter to paint blindfolded, then blaming them when the artwork is ugly." Would you like me to help you draft a social media post or a technical summary based on this translation?

别再对着 Claude 瞎提问了，YC 老大 Garry Tan 开源的 gstack 才是真核弹！ 我实测了一上午，这玩意儿直接把 Claude Code 变成了一个顶级虚拟工程团队。GitHub 48k+ Star 真的不是吹的。 Garry Tan 靠这套 setup 60 天撸了 60 万行代码，我上手用了半天，最大的感受就是：我终于不再是那个苦逼的代码搬运工了，我是 CTO。 一句话总结：gstack 给了你 15 个硅谷专家角色 + 6 个暴力工具，全是斜杠指令。 最让我感到“离谱”的几个实战瞬间： /office-hours：真的会“怼”人：它不是只会写代码，而是变身 YC 毒舌合伙人。我输入一个想法，它对着我的产品逻辑一顿疯狂输出，挑战我的每一个假设，重新定义问题。 /plan-ceo-review & /plan-eng-review：这才是最硬核的。一个从 CEO 视角审视方案，一个从架构、数据流、边界情况锁定技术细节。在写代码之前，它已经帮你避开了 90% 的坑。 /plan-design-review：专门帮我抓那些恶心的 AI Slop（AI 味儿十足的平庸设计）。 /review & /codex：这是双重保险。一个是 Staff Engineer 级别的代码审查，另一个是调用 OpenAI Codex 做二次交叉对比。这种跨模型审查，Bug 根本藏不住。 /qa & /ship：最爽的环节。它真的会启动真实浏览器去“点点点”测试，跑完测试直接开 PR、一键发布。 这就是传说中的“一个人就是一个公司”。 别再一行行手敲代码了，去试试 gstack。让这群“虚拟专家”去设计、去 Review、去 QA、去发布。 如果你还没开始折腾这套东西，那你可能真的要被这波 AI 浪潮给甩开了。 #AI编程 #GarryTan #gstack #GitHub 传送门：github.com/garrytan/gstack

In Prediction Markets, Distribution is King! As a new type of financial product, the future trend for prediction markets will undoubtedly be their integration into various channels and traffic gateways. Where is the largest source of traffic in the crypto space right now? It’s the exchanges and Web3 wallets. In fact, many exchanges have already integrated Web3 wallets directly into a single app. Yesterday, I noticed that Gate has integrated Polymarket. I gave it a try, and the overall experience was quite smooth. However, there are two critical issues: Deep Entry Point: The prediction market is tucked away under the "Alpha" section rather than being a primary menu item. This will inevitably result in a significant loss of organic traffic. Order Execution Experience: There are still gaps in the trading process. For example, you can't place orders using an order book, and share prices don't update in real-time. These areas still require continuous optimization. That said, this is a highly valuable experiment and marks a key step in the convergence of exchanges and prediction markets. It serves as an excellent case study to observe how effective exchange channels actually are for prediction platforms. Looking at the data: Yesterday (March 24), Gate contributed $300,000 in trading volume to Polymarket, landing it in the top 10 of Polymarket Builders. I’ll be keeping a close eye on how these numbers evolve.

Chelsea's 0.1% odds to win the PL moved $6.7M in 24h. When a market rounds you to zero, traders aren't predicting anymore—they're saying "this will not happen." The bigger question: are they right?

Prediction Markets: The Three-Plate Framework 🧵 What are we actually predicting? It boils down to three pillars: 🔥 1. The Speculation Plate (High-Frequency)High turnover, high stakes. This is the "Liquidity Floor" for volume. Key: 5m/15m price action & Sports. 🎉 2. The Cultural Plate (Viral/Fun)Social-driven and fun. This is the "Traffic Engine" for mass onboarding. Key: Elections, pop culture, and event marketing. 🛡️ 3. The Hedging Plate (Real-World Value)Professional risk management. This is the "Value Pillar" that defines long-term legitimacy. Key: Macro data, Pre-TGE, CPI insurance, and weather hedging. The Reality: Polymarket is 90% Speculation + Culture. The Hedging space is a massive, empty void. As Vitalik noted, we need to move beyond being a "new-age casino" and start solving real-world problems. The Strategy for the Future: Generalize: Drive Mass Adoption by making markets accessible to everyone. Specialize: Create professional tools for real-world risk. The roadmap is set. Excited to push the boundaries of what’s possible with Predict! 🚀

The market is pricing a Trump China visit by March at 1.6%. That's less likely than rolling snake eyes. Twice. $396k in volume suggests serious money doesn't see it happening—but headlines keep teasing it anyway. Track it live → bydfi.com/en/moonx/marke…

Follow @moonx_bydfi for daily market signals. Track it live → bydfi.com/en/moonx/marke…

This week's data is screaming one thing: macro consensus is locked. For the next 7 days, expect either a geopolitical shock or nothing. That's the trade. Netanyahu moves or we drift sideways. Markets hate ambiguity. One of these happens. Which?

🧵 MoonX Market Brief 99.7% of the market says no Fed move in March. When consensus gets this tight, you're either watching the consensus crush dissent, or you're all watching the same blind spot. Thread 👇