与「Refi」相关的搜索结果

posting the same announcement ten times can make someone look active. but does that mean they've contributed anything useful? that's the problem @rarelistio is trying to address. on Rarelist, creators earn points called RARS for contributing to project campaigns. the team is refining how it awards those points, giving more weight to original posts that show proper research. they're also adjusting filters to catch spam and low-effort copies, which can lose points or get filtered out before rewards are paid. these changes don't mean every well-researched post will automatically rank higher. but i find the direction interesting. i wonder if it'll make creators spend more time finding something useful to share instead of repeating the same information.

🚨重磅！GLM 5.5最快将于本周发布！将掀起开源模型全新高度！ 智谱Zai创始人唐杰清华早已暗示两大飞轮RSI、后训练将铸就 GLM-5.5 成为无敌的存在！ 🔹飞轮已跑通：500M代码仓库→千万级Issue-PR合成→GLM-5 Agentic Engineering；合成沙盒→GLM-5.1自主执行；Agent自己造动态环境→GLM-5.2长程任务；网络安全环境，漏洞+PoC+Exploit→GLM-5.3。每一步都在用上一代模型生产下一代的训练数据。 🔹数据墙被击穿：Chinchilla定律明确，固定数据无法消除数据依赖损失。精选Web数据（RefinedWeb）远超原始语料，为飞轮提供高质量燃料。 🔹GLM 5.5是闭环产物：GLM-5.3已驱动Infra Agent在10万国产卡上自己优化推理系统，两周吞吐提升3倍+。飞轮继续转，5.5的后训练+自我进化环境只会更硬核，量化金融、代码沙盒、长程Agent全线拉满。 从Vibe Coding到Agentic Engineering，再到数据飞轮自我强化，5.5不是简单迭代，是这套系统第一次真正闭环后的产物。 #GLM55# #智谱# #唐杰# #数据飞轮# #RSI# #AgenticAI# #开源大模型# #清华# #AI# #Zai#

❄️ What makes NICE1® different from traditional ice? Dr. Eric McCarty demonstrates how NICE1® combines cold + compression in an iceless system. No bags of ice. No constant refilling. Less hassle. Recovery technology designed to fit the patient’s routine. 💪 #SportsMedicine#

Elon Musk on whether we're alone in the universe. "We have no evidence of alien life whatsoever. Nothing, nothing, nothing. Flat damn nothing." Elon sticks to the scientific method: no evidence, no claim. And his bar for proof is simple. "All it would have taken would be for an alien civilization to put a one-inch cube of pure titanium." Titanium never shows up pure in nature. Refining it takes real industry. One small cube would prove someone built it. No signals. No artifacts. Not even a cube. He told us this in our 2022 interview. The 𝕏 Takeover at Giga Texas is this Saturday, 10/10. @elonmusk @theXtakeover Do you think we're alone?

🚁 Tell the #PhaserGameAgent# you want a synthwave rescue-chopper game. Keep refining. You land on something like "Choppy". By Jeff Marsh: clear five bases, carry every soldier home over the wall 🪖 Play now 👉 #gamedev# #indiegames# #madewithphaser# #ai#

📣 Diesel or net zero, how long can we have both for? 📣 📰 Donald Trump has told Europe to release emergency diesel or risk being cut off from American supplies.  Britain joined crisis talks after he threatened a ban on diesel exports to France and Germany.  The White House aides say American farmers, truckers and businesses should not carry a global diesel shortage.  Britain imports 55 per cent of its diesel, about a third of that from the United States.  There are more than 15 million diesel vehicles on the road, and the price has broken £2 a litre.  Net zero and green policies led us to shut refineries, Britain closed two last year, and stocks have fallen.  Trump’s stance is that a shortage you built is a shortage you answer for.  Are we going to reopen the refineries, now that fuel is a national security problem, not an export CO2 emissions to hit a climate target problem?

近期手臂的任務對我來說困難度大大提升阿 各位高級鉗工有沒有秘訣傳授一下 我最近看 @axisrobotics 的技術架構，有一個地方其實很值得注意 它沒有把整套 Robotics 工作流綁死在單一 Simulator 或單一電腦環境裡。 這件事乍看很工程師，但對 Physical AI 來說影響很大。 Axis 的做法是把不同環境負責的工作拆開。 前端可以直接在 Browser 裡用 MuJoCo WebAssembly 做 Teleoperation，Web、Mac、Windows 都能參與，不需要 GPU 或特殊軟體。 完成後，Trajectory 會用統一格式傳到後端，再交給 Linux GPU Server 執行更高負載的 Simulation。 這裡就可以接上 IsaacSim，進一步做更真實的場景、物理條件與 Domain Randomization，最後再進入 Data Refinement、Model Training 和 Sim-to-Real Evaluation。 簡單畫成一條線 Browser → MuJoCo → Unified Trajectory → GPU Server → IsaacSim → Training → Real Robot 我覺得這個架構真正有意思的地方，是把「資料怎麼收」和「資料怎麼變得更真實」分開處理。 收集資料需要的是低門檻與高擴展性。 Simulation Augmentation 則需要大量 GPU 運算與更複雜的環境。 兩者如果全部塞在同一套環境裡，Contributor、Simulator、硬體與運算資源很容易互相限制。 Axis 把這些環節拆開，再透過統一的 Trajectory 格式串起來，等於讓不同機器、不同 Simulator、不同運算環境之間可以開始協作。 這也是我覺得 Robotics Infrastructure 很重要的一點。 真正能長期擴張的 Data Engine，最後拼的可能不只是誰收集了多少資料，而是這些資料能不能跨環境流動，並且持續進入下一個階段。 Axis 目前的 Cross-Simulator / Cross-Machine 架構，就是在解這個問題。

City at night by Opus 5.5 Made from scratch with blender, image-gen, and Three.js. Still amazing that it runs in the browser. Also refined some combat, animations, and music. Game dev with AI is a reality now.

Up, up, and away. 🚀 Today, in partnership with @planet, we launched a prototype satellite carrying four TPUs into orbit on @SpaceX's Transporter-18 rideshare mission. This launch is the first step of Project Suncatcher, our long-term research moonshot to see whether we can one day host scalable machine learning infrastructure in space. Over the coming weeks, we'll gather in-orbit data on how our TPUs handle the physical stress, radiation, and thermal extremes of space. Whatever we learn, we'll use it to refine our future designs.