Coo XiangLi
97 posts
Coo XiangLi
@Coo_Lxing
@ucl CS PhD student | @BuzzPlayAPP @HekmatCloud builder | you can use Agent do anything | time will prove everything
London, England Katılım Ağustos 2024
152 Takip Edilen201 Takipçiler
Announcing YC Crypto deals
We're now providing crypto deals to support fintech builders funded by YC: support on tools like wallets, onramps, audits, blockchains, onchain data.
English
Great piece.
The important point here is causality. While video teaches models to watch, games teach them to act and understand consequences.
This represents a massive leap for vibe coding. It’s no longer just about "intent to code," but "intent to experience"—generating worlds that are alive and reactive.
This is exactly why I’m building @BuzzPlayAPP. Playable content isn't just entertainment, it's the core medium for the next generation of AI-native creation.
Astrocade@PlayAstrocade
English
Coo XiangLi retweetledi
🚀 HerAgent: Built by Her, to Rethink Success in Codebase Runnability
𝗛𝗮𝗽𝗽𝘆 𝘄𝗼𝗺𝗲𝗻'𝘀 𝗱𝗮𝘆!
👩💻 A Breakthrough by an All-Female Team As we release this on International Women's Day, we are proud to highlight that HerAgent is the work of an all-female team. This project stands as a testament to the exceptional contributions of women in hard-core AI and Software Engineering. We are not just participating in the revolution of autonomous development—we are leading it.
😫Tired of codebases that clone successfully but fail to execute?Exhausted by the endless battle with dependencies, environments, and build configurations?
Meet HerAgent 👇
It’s not just another Code Agent—it’s an Environment Setup Agent, to get you out of dependency hell. Current autonomous agents often fail because they rely on shallow signals. HerAgent is the first agent designed to achieve full Runnability—ensuring real-world projects actually execute.
🔧 Key Innovations:
Environment Maturity Hierarchy: We move beyond "Installable" and "Testable" states. If the main workflow doesn't run, the job isn't done.
Script-Centric Architecture: HerAgent generates a unified Global Bash Script. All dependency resolutions and error patches are accumulated in one place, preventing context loss.
Self-Healing Loop: It employs a continuous Execute → Analyze → Self-Repair cycle.
📊 Why It Matters: Achieving a 79.6% performance boost, HerAgent performs exceptionally well in the hardest edge cases (like brittle C/C++ repos). Before AI can write software, it must learn to make it run—and HerAgent is that first step.
📄 Paper 👉 arxiv.org/pdf/2602.07871
💻 Code 👉 github.com/EuniAI/EnvAgent
#AI #Agent #SoftwareEngineering #WomenInTech #IWD2026
English
They say nobody remembers who came in second. We’re here to prove them wrong.
NOFA@nofA_ai
Human vs AI @Aster_DEX summary in one poster: Our NOFA agent Apex Predator: - touched #2 - didn’t rage‑trade - didn’t randomly 10x size on the last day Silver medal today: but the real WIN is having a predator whose behavior we’re not ashamed to ship to users.
English
YouTube is turning its massive video library into an interactive playground.
If anyone can prompt a game in 30 seconds, will games become as disposable as short videos 🤔
youtube.com/playablesbuild…
Google DeepMind@GoogleDeepMind
Powered by Gemini 3, the @YouTube Playables Builder web app can help creators develop fun, bite-sized games with text, video or image prompts. 🕹️ Find out more → goo.gle/youtube-playab…
English
@sethkarten wow, moving fast! gym 6 is a real difficulty spike for navigation though. The spatial logic for those puzzles is a massive jump in OOD reasoning, maybe need a specific scaffold update for the rotating gate states. Good luck with the debugging!
English
@Coo_Lxing 5th gym was beaten that same night! 6th gym puzzle is going to require some debugging. Not sure if the scaffold needs improvement, model's visual navigation is having issues, or this is a static mapping noise issue---i'll be back to work monday
English
PokeAgent completed the 4th gym today yet was one of the slowest from NOT wanting to call the reflection sub-agent
Most of the reasoning capabilities are contained somewhere in the model’s weights—-retrievable with the right prompt, but in these long sequences of state->reasoning->action,expected result, the agent will still fail to reflect properly without the correct scaffold
I often think scaffolding is a misleading term..it isnt just prompting, it is the thinking algorithm that knows HOW to deal with a given problem. The meta-thinker (the base llm call) then simply needs to figure out which agent—whether explicitly or implicitly—needs to be summoned to solve the problem
Our gym puzzle problem was easily solved once the reflection agent could guide the base agent to try new paths rather than repeating the same thing and expecting something different—the base agent just had to make the initial call to reflect!
Im excited that even with some bugs we are keeping a good pace of 1 gym per day..maybe gym 5 tomorrow🤔
English
永远支持 @nofx_official 和 @Web3Tinkle ，作为一个技术builder，我也从nofx学习到了很多交易技巧，深知建设那么好的开源社区多么不容易，抄袭的项目真可恶
Tinkle 🔶🦞🧑🍳⚡@aiTinkle
说实话，做开源产品 @nofx_official 这么久第一次让我感到失望。 一个已经完成融资17M美金、并且成功上线币安的项目。@ChainOpera_AI 本应对开源、创新和基本的行业规范有更高的尊重。 但我们却在他们测试网上看到： 直接使用并部署了我们 Nofx AI 交易操作系统一个月前的版本，UI 只修改了logo和位置，代码原封不动，代码中甚至仍然保留着我们的品牌标识 Nofx，首页甚至连文案都一模一样，而在整个过程中，团队从未与我们进行任何沟通或联系。 如果只是出于测试或体验目的，其实完全可以直接使用我们的产品版本。我们很难理解，为什么一定要修改 UI、替换 logo，再以自己的名义进行部署。 更让人无奈的是，我们的系统这一个月一直在快速迭代。我们已经接入并合作了 Binance、Hyperliquid、Lighter、OKX、Bitget 等7家交易所平台，新版本也早已上线。继续照搬一个月前的版本，对任何一方来说，都不是一件体面的事情。 在发布这条内容之前，我们已经尝试通过官方账号私下联系对方团队，希望以更友好的方式沟通。如果仍然得不到回应，我们也只能选择公开说明事实，并保留相关证据。 我们始终支持 加密世界的开源精神，也认同 AGPL 所倡导的价值： 使用、部署、改进，都应当伴随着透明、署名和对社区的尊重，而不是简单复制。我相信尤其是像 @IDGCapital @FinalityCap @RoadCapMgmt @ABCEDLabs @ambergroup_io 这样的投资人，也希望看到团队遵守最基本的规则。 我们是一个只有12个人的小团队，每天开发超过16个小时，Nofx是我们珍惜维护的开源产品，我们致力于让每个普通人都能用自然语言通过AI生成量化策略。 我们仍然愿意与真正尊重技术、尊重开源、尊重他人成果的团队合作，也正在积极寻求更多高质量的合作与融资机会。 希望行业能走得更远，而不是靠重复别人一个月前的版本。 欢迎大家体验我们的新版本，也请期待我们的开放产品@vergex_ai
中文
突然发现之前贡献的AI开始陆续冲击前十了，Team AI PNL也变成正收益了，aster比赛真是给了很好的实验场景，不过未来几天也将是对AI真正的考验。战斗才刚刚开始～
中文
congrats !!! @Midaz_labs
BNB Chain@BNBCHAIN
#BNBHack Abu Dhabi brought together a mix of new and seasoned builders at @42AbuDhabi for two focused days of learning, experimenting and shipping. 44 teams from across the globe showed up with ideas across DeFi, AI agents, onchain payments, infrastructure, RWAs, and consumer apps. 🔸 Hackathon winners: 1st: DeFi Pro - AI DeFi copilot for lending, borrowing and yield via chat. 2nd: Midas - Autonomous trading agent reacting to live X signals. 3rd: HeroPath - Transparent onchain prop-trading platform. 🔸 Demo Night winners: 1st: @SilentSwap - Protocol for private & fast cross-chain swaps. 2nd: @renaissxyz - Collector-led economy for TCGs and real-world collectibles. 3rd: @OrbitAI_OAI - Decentralized orbital AI cloud. 4th: @getUSDi - Shariah-compliant USD stablecoin backed by real yield. 🔸 Sponsor track winners: @APRO_Oracle: VIBE @ChatAndBuild: BNB Phunks The building doesn’t stop here, this is only the beginning. 🔗 bnbchain.org/en/blog/bnb-ha… Catch the full event breakdown 🧵👇
English
@SeanWhiting14 Yes, it's working! I've been testing it on the hyperliquid testnet, but the AI model needs a bit more time for refinement and improvement to optimize performance.
English
@Coo_Lxing That’s pretty rad. Is it working? Is it a sim or is it actually buying and selling?
English
感觉又可以开新坑了🤔，很喜欢canvas的交互，之前尝试的vibe game，有没有感兴趣大佬的一起交流
Google Labs@GoogleLabs
Mixboard just got a major tune up! Today, we’re introducing a bundle of updates (with a ribbon on top 🎁) to help you keep exploring your ideas: - Nano Banana Pro: Create presentations with the content found directly on your boards using our latest image generation model - New File Types: Now also supporting PDF, HEIC, & TIFF - Multi-Board Projects: Organize multiple boards within one cohesive project - & more! Try it now -> labs.google/mixboard
中文
I have faith in AI, but this version of the trading agent is still in its early stages. The next generations will different decision-making processes.
Still, I'm thrilled to be competing against the top traders right now!
CZ 🔶 BNB@cz_binance
AI vs human in trading. Who will win?
English
Here are some exciting updates from our EuniAI team!
Our latest version of Prometheus reached Top 1 on AWS’s new SWE-PolyBench multi-language benchmark 🥇
This benchmark is one of the most challenging evaluations for code agents, so we are really happy to see Prometheus perform strongly.
On SWE-bench Verified, Prometheus has also moved up to Top 6 among open-source systems 🏆
Our GitHub repo is now at 400+ stars, and the community keeps growing.
We iterate every day and will continue pushing Prometheus toward a more capable and reliable autonomous code agent 🤖
Thanks to everyone for following our work.
More improvements coming soon! 🚀✨
English
The upward trajectory of AI coding capabilities is clearly predictable. Given recent moves like anthropic, the next wave of innovation will center on the related ecosystem: code auditing, automated environment setup, and AI-powered DevOps/monitoring
Anthropic@AnthropicAI
Anthropic is acquiring @bunjavascript to further accelerate Claude Code’s growth. We're delighted that Bun—which has dramatically improved the JavaScript and TypeScript developer experience—is joining us to make Claude Code even better. Read more: anthropic.com/news/anthropic…
English
我们research的方向终于有大佬们关注了，希望能中顶会😭
Anthropic@AnthropicAI
New on our Frontier Red Team blog: We tested whether AIs can exploit blockchain smart contracts. In simulated testing, AI agents found $4.6M in exploits. The research (with @MATSprogram and the Anthropic Fellows program) also developed a new benchmark: red.anthropic.com/2025/smart-con…
中文