GreeceLee

高强度冲浪，主营各类AI会员/软件工具/海外社媒账号/涨粉买量等

[GreeceLee @GreeceLeeea]

@thsottiaux Cutting ChatGPT desktop state by 60 percent is a real relief for power users.

[GreeceLee @GreeceLeeea]

@CaineArdayfio Audio to action in 200ms is a big leap. Skipping the transcription step removes so much friction for voice interfaces.

[GreeceLee @GreeceLeeea]

@ycombinator @alexandr_wang @garrytan Makes sense, models catch up fast but the people with a clear direction don't.

[GreeceLee @GreeceLeeea]

@ns123abc Using their own models to trace the device back to one employee is the wild part.

[GreeceLee @GreeceLeeea]

小模型单卡就能跑出这种吞吐，低成本部署的账真得重算了

[Omar Sanseviero @osanseviero]

Argon helped Gemma break 120,000+ tokens/sec on a single A100 on vLLM 🚀

The team optimized Gemma 4 E2B, unlocking a ~2.9x speedup (124k prefill, 111k e2e) with zero quality degradation

Here's what it looks like in practice: generating 64 live Python simulations at once 👇

[GreeceLee @GreeceLeeea]

@Gorden_Sun 每天几万次合并，Git 确实不是为这种量设计的，换掉也不意外。

[GreeceLee @GreeceLeeea]

@venturetwins Adding GPU thermal specs to a power bottleneck analysis is a smart angle, that's usually the missing piece.

[GreeceLee @GreeceLeeea]

@rileybrown Dropping the slop Remotion graphics and overlaying real clips from niche tweets is the actual win.

[GreeceLee @GreeceLeeea]

@skalskip92 The handwriting and rotated text parts are what usually break OCR pipelines, so that leap is actually huge for production work.

[GreeceLee @GreeceLeeea]

@AndrewCurran_ Once merges go 10x, review becomes the real bottleneck.

[GreeceLee @GreeceLeeea]

@dotey 定价还行，多步任务终于有个能打的非 OA 选项了。

[GreeceLee @GreeceLeeea]

@OfficialLoganK 1M token output is a big unlock for long coding sessions. Exciting rollout.

[GreeceLee @GreeceLeeea]

编程和网络防御都点名了，真正好不好用还得看实际跑项目的表现

[Google DeepMind @GoogleDeepMind]

Gemini 4 Argon is now globally available.

Built to solve your hardest problems from start to finish, the model brings frontier capabilities across coding, enterprise workflows, and cyber defense.

[GreeceLee @GreeceLeeea]

@xiaohu 先用代码把镜头和走位定好，再去磨质感，这思路比纯开盲盒靠谱

[GreeceLee @GreeceLeeea]

@mattyp @bot The email and Shopify hooks are the kind of moves that actually change how people use bots day to day.

[GreeceLee @GreeceLeeea]

@AlchainHust 订阅把 token 单价压到 API 下面这块确实关键，现在模型公司拼的就是这个

[GreeceLee @GreeceLeeea]

@mattpocockuk Aligning too little and aligning too hard both being on the list is painfully accurate.

[GreeceLee @GreeceLeeea]

@ScarletKc 给 AI 开终端权限比换模型涨得猛多了，沙箱里跑命令那一下感觉完全不同

[GreeceLee @GreeceLeeea]

模型是强，但最后还是得看日常用起来顺不顺手

[AI at Meta @AIatMeta]

We’re excited to introduce the Muse Wonder series with Muse Wonder and Muse Wonder Full, our most capable models yet. Both are frontier models, built to deliver state-of-the-art performance for personal and professional agentic tasks.

Muse Wonder is available today via the Meta Model API and Muse Code. Muse Wonder Full is available today for Muse Code subscribers. Muse Wonder will also soon power experiences in our personal agent product, Muse, rolling out first to subscribers.