AMAZINGINDEX.COM 每天向上一点
59.8
VOL. 2026.11
2026.11.13
← 返回 2026.11.13 日报
日报快照 · Daily Snapshot
NO. 027

Google 连发 Gemini 4 与 AlphaGenome 2

#TWEET_DIGEST X (Twitter) 2026.11.13
推荐指数 76.0 NO. 027 · 2026.11.13
推文26点赞32,197转发2,286Tweets26

Google 密集发布 Gemini 4 Carbon 决策模型和 AlphaGenome 2 DNA 序列模型,前者以极低 token 消耗登顶 CursorBench,后者可预测细胞类型特异的 DNA 变异影响。对 AI 工程师意味着推理成本可能再降一个数量级,生物计算赛道出现可落地的预训练模型。

Hassabis 把两条线放在同一天发不是巧合:Gemini 4 Carbon 的 " barely uses any tokens " 暗示 Google 在推理效率上找到了新压缩路径,这可能是对 DeepSeek R1 成本攻势的直接回应。

AlphaGenome 2 更值得长期跟踪——之前 AlphaFold 解决了结构问题,但变异功能预测一直是临床应用的瓶颈,这次明确开放学术 API,说明 Google 想把生物计算从「发论文」推进到「被药厂调用」的阶段。

做生物信息学的团队建议优先申请 API 测试其在罕见病变异解读上的表现;做模型推理优化的工程师应该拆解 Gemini 4 Carbon 的 token 消耗数据,看 Google 是否采用了新的注意力机制或投机解码策略。

@demishassabis Likes 10.3k Reposts 609

RT @OfficialLoganK: Introducing Gemini Decision 1, our new frontier decision model in the Gemini API. It is $0.10 per 1M input tokens wit…

查看原文 →
@demishassabis Likes 5.1k Reposts 292

RT @cursor_ai: Gemini 4 Carbon is now available in Cursor! At high effort, it's the top model on CursorBench at 60.4% https://t.co/4I32noz…

查看原文 →
@demishassabis Likes 3.5k Reposts 437

Super proud to announce AlphaGenome 2 - our new DNA sequence model! 🧬 It predicts the impact of DNA variants in different specific cell types - unlocking crucial information for understanding complex diseases. Available for academic research via our API: https://t.co/dEyrvpTxIl

查看原文 →
@demishassabis Likes 2.9k Reposts 95

RT @LuminaBench: Gemini 4 Carbon is ridiculously efficient It barely uses any tokens, it's super cheap and incredibly fast for a model not…

查看原文 →
@demishassabis Likes 2.5k Reposts 153

Very excited to release Gemini 4 Hydrogen! Incredible speed and intelligence for everyday coding, agentic work and all-round use. I think you'll love this model. More info in the blog - enjoy! https://t.co/952K09CUnd

查看原文 →
@demishassabis Likes 1.6k Reposts 43

RT @willcb: i think i like carbon more than fable

查看原文 →
@OpenAI Likes 1.3k Reposts 177

RT @jeffreygwang: (1/10) Today, we’re sharing new research from @OpenAI: How does CoT monitorability (no-CoT capability and CoT controllabi…

查看原文 →
@demishassabis Likes 1.1k Reposts 49

RT @martin_casado: Ok. Carbon is *really* good at code.

查看原文 →
@DrJimFan Likes 718 Reposts 87

Autonomous, uncut, 1x speed. This is GR00T 2 controlling real dexterous hands to perform a chemistry lab experiment (and clean up after itself) end to end. Episode 1 of a new series! I’ll keep dropping real-time robot videos, with none of the “4x” or “8x” labels you usually see in the corner. No one can speed up reality, except Dr. Strange ;)

查看原文 →
@demishassabis Likes 480 Reposts 31

RT @levie: Ok this is a big deal. We’ve been testing Gemini 4 Hydrogen in early access with the Box AI Complex Work Eval, and it’s a seriou…

查看原文 →
@demishassabis Likes 415 Reposts 11

RT @SahilPanhotra: Gemini 4 carbon is fucking awesome results soon : )

查看原文 →
@demishassabis Likes 359 Reposts 26

Glasses are one of my favourite Gemini integrations - congrats to @izadi_shahram and the team & our fantastic partners!

查看原文 →
@demishassabis Likes 300 Reposts 22

RT @haider1: well damn google is having one of its craziest release weeks ever: nov 9 — tpu 8tx and 8i max nov 10 — gemini robotics 2.5 no…

查看原文 →
@demishassabis Likes 289 Reposts 28

RT @danielmckinn0n: Big G remains 👑 of rare-disease evals. @GoogleDeepMind Gemini 4 Carbon and @AnthropicAI Fable 5.2 tie for first on @Ga…

查看原文 →
@ylecun Likes 211 Reposts 11

RT @jbhuang0604: @ylecun @TimDarcet @gabriberton in a nutshell https://t.co/hP2RxZKInv

查看原文 →
@ylecun Likes 184 Reposts 88

RT @ReutersWorld: EXCLUSIVE: Covert Russian network exploited French riots on social media, EU report says https://t.co/lZxd71ugKX

查看原文 →
@ylecun Likes 181 Reposts 38

RT @NYU_Courant: We're hiring! New York University expects to have several tenure-track faculty positions in Computer Science and invites q…

查看原文 →
@svpino Likes 134 Reposts 3

n8n just released a personal agent. Open source! Seems pretty awesome: 98% task completion using GPT-6.1 Sol in October. Separate from this launch but I love it: apparently, @n8n_io has a small division (Labs) where they are interacting quickly on different ideas. Love that!

查看原文 →
@svpino Likes 104 Reposts 3

Here is a big problem with agentic coding: The more code you write, the more complex it is to keep the application from blowing up in users' faces. For every new feature you implement: • You need to determine how to measure it • You need to plan who gets it first • You need to decide when to let more people use it • You need a rollback strategy for when things go wrong This is a huge problem, but we can fix this very easily! ↓

查看原文 →
@ylecun Likes 93 Reposts 15

RT @WIRED: Moving AI from text to 3D space, world models simulate real-world physics to teach robots and autonomous cars how to navigate re…

查看原文 →
@svpino Likes 92 Reposts 33

This gives you a Linux machine with Claude Code, Codex, and OpenCode already installed. The beauty of this is how much of the agent workflow comes already assembled and ready to use: • Preinstalled coding agents and development tools • Commands to start and monitor background tasks • Credential handling without exposing them • Automatic pauses when agents are idle

查看原文 →
@ylecun Likes 87 Reposts 18

RT @nickclegg: Getting too close to Trump was a mistake – how could Silicon Valley have gotten (US) / got (UK) it so wrong? Read the full…

查看原文 →
@svpino Likes 86 Reposts 10

I built an agent and tested it without writing a single line of code. Here is how to create a digital twin of your agent, mock every tool it touches, and simulate every function it runs without affecting production. I'm using Claude Code here. Probably the closest thing to magic you'll see this month. This will test and discover problems with your agent that you didn't even know were possible. It will then tell you how to fix everything it finds.

查看原文 →
@svpino Likes 71 Reposts 5

This new model is a big deal: It was trained to return the exact passages from the source document that answer a question. Think about applications that work with contracts, invoices, emails, or other documents. You always want to answer questions with direct traceability back to the source. $0.01 per 1M input tokens with free output. Awesome price for processing documents at scale!

查看原文 →
@svpino Likes 55 Reposts 2

A new open-weight 100B-parameter model that runs on your laptop. It's fast and good enough for vibe coding. They shared a few metrics that look pretty impressive: • Only needs 29GB of RAM, which is good for a 100B model • Beats Claude Opus 4.6 on Terminal-Bench 2.1

查看原文 →
@svpino Likes 53

Wow, Jev is giving Google a run for its money. $0.20 cheaper per 1M tokens. It's kind of surprising that Google can't compete on price here.

查看原文 →
查看原文 →