科技推特精选 - 2026-01-28

今日科技动态：随着月之暗面（Moonshot AI）发布 Kimi k2.5，开源 AI 格局发生重大变革。作为一款强大的多模态智能体模型，Kimi k2.5 正面挑战顶级商业闭源系统。与此同时，OpenAI 推出用于科学研究的 Prism，谷歌在全球范围内扩大了 Gemini 3 Pro 的访问权限，市场焦点依然集中在开源模型性能的飞速增长。在资本市场，随着微软和 Meta 财报临近，两家巨头正面临证明巨额 AI 投资合理性的压力；与此同时，新型编程智能体正引领软件工程领域经历一场根本性转型。

1. lmsysorg (Group Score: 90.3 | Individual: 28.9)

Cluster: 4 tweets | Engagement: 45 (Avg: 36) | Type: Tech

🎉 Congrats to @Kimi_Moonshot on the release of Kimi K2.5! We’re excited to announce day-0 support for Kimi-K2.5 in SGLang ✨

Kimi K2.5 is a powerful open-source, native multimodal agentic model, trained on ~15T mixed vision–text tokens. It’s a unified architecture that brings together 👀 Native multimodality (vision + language) 💻 Vision-to-code workflows and tool chaining 🤖 Agent swarm execution for parallel, coordinated task solving

Related PR: https://t.co/mZwAjDV1hl You can now run Kimi K2.5 with SGLang using the following command 👇

See 3 related tweets

@mervenoyann: RT @Kimi_Moonshot: Kimi K2.5 has arrived! 🥝 Here are 2 things to know: Aesthetic Coding x Agent Swar...
@ModelScope2022: 🚀 Meet Kimi K2.5! 🌙 This is Kimi’s most intelligent and versatile model to date, achieving SOTA perf...
@AdinaYakup: Kimi K2.5 from @Kimi_Moonshot is more than just another large model🤯

https://t.co/rlPwVO2EHN

✨ Na...

2. GenAI_is_real (Group Score: 80.4 | Individual: 30.6)

Cluster: 3 tweets | Engagement: 33 (Avg: 32) | Type: Tech

kimi k2.5 just dropped and the agent swarm beta is actually insane. 1,500 tool calls at 4.5x speed? openai is still trying to get o1 to follow simple instructions while moonshot is literally building an autonomous workforce. hle 50.2% is the new ceiling. if you aren't thinking about multi-agent scaling yet, you're building for 2024. @kimi_moonshot @sama

See 2 related tweets

@aakashgupta: Moonshot just told you the agentic era requires swarms, not single-agent chains, and nobody’s repric...
@FireworksAI_HQ: Kimi K2.5 is now live on Fireworks with full parameter RL tuning support!

This is @Kimi_Moonshot Mo...

3. kevinweil (Group Score: 66.4 | Individual: 32.0)

Cluster: 3 tweets | Engagement: 723 (Avg: 338) | Type: Tech

💥 Today we’re introducing Prism—a free, AI-native workspace for scientists to write and collaborate on research, powered by GPT-5.2.

Accelerating science requires progress on two fronts:

Frontier AI models that use scientific tools and can tackle the hardest problems
Integrating that AI into the products scientists use every day

Prism is free to anyone with a ChatGPT account, with unlimited projects and collaborators.

Try it today at https://t.co/cmEsJYRtW9—would love to hear your feedback.

See 2 related tweets

@tszzl: RT @OpenAI: Introducing Prism, a free workspace for scientists to write and collaborate on research,...
@TechCrunch: OpenAI launches Prism, a new AI workspace for scientists https://t.co/y8hBOmJHnR...

4. eliebakouch (Group Score: 65.5 | Individual: 30.1)

Cluster: 3 tweets | Engagement: 578 (Avg: 280) | Type: Tech

very nice release by the kimi team, benchmarks are on par with opus 4.5, gpt 5.2 xhigh, gemini 3.0 pro

there is also some nice details on the parallel RL part in the tech blog explaining how they build K2.5 agent swarm https://t.co/yAFLdjPR49

See 2 related tweets

@mark_k: Kimi K2.5 benchmarks are here, and the data suggests a new leader in the open-source space. 📉

@Kimi...

@victormustar: RT @eliebakouch: very nice release by the kimi team, benchmarks are on par with opus 4.5, gpt 5.2 xh...

5. Reuters (Group Score: 55.6 | Individual: 20.3)

Cluster: 3 tweets | Engagement: 25 (Avg: 66) | Type: Tech

Microsoft and Meta will kick off Big Tech earnings this week under pressure to prove that their costly bets on AI can power another year of strong growth as a resurgent Alphabet leads the high-stakes race https://t.co/wXrJRN3YbH https://t.co/7SsZQUyTVv

See 2 related tweets

@ReutersBiz: Microsoft and Meta will kick off Big Tech earnings this week under pressure to prove that their cost...
@Reuters: Big Tech earnings to test AI rally as resurgent Alphabet takes lead https://t.co/90Ah9JvEcd https://...

6. chddaniel (Group Score: 54.6 | Individual: 31.2)

Cluster: 2 tweets | Engagement: 40 (Avg: 20) | Type: Tech

I built a better vibe coding tool

Bolt .new rasied: $100M I raised:$ 0

Bolt's ARR: $40 MILLION My ARR:$ 84k

Bolt's investors: Google Ventures, etc My investor: my mom's basement

Their Figma import: left @Shipper_now Figma import: right

As we speak my dev is making this even better..

Drop your Figma links below, I'll put it through Shipper and show you the result

See 1 related tweets

@chddaniel: Bolt .new's ARR: $40 MILLION My ARR:$ 84k

Bolt rasied: $100M I raised:$ 0

Bolt: 5 million Me: 190 ...

7. KirkDBorne (Group Score: 51.7 | Individual: 13.5)

Cluster: 4 tweets | Engagement: 12 (Avg: 26) | Type: Tech

Learn Python for #MachineLearning comprehensively in this massive 772-page book: => #Python Machine Learning (3rd Ed.) => https://t.co/5CSkCaSTJM by @rasbt ———— #DataScience #ML #AI #NeuralNetworks #DeepLearning #DataScientist #ReinforcementLearning https://t.co/sFIpgQqlh3

See 3 related tweets

@KirkDBorne: Get one of these #ML with #Python books, depending on your level:

🌟1000 5-star ratings🌟>> “Ma...

@KirkDBorne: #DeepLearning with #TensorFlow and #Keras — Build and Deploy Supervised, Unsupervised, Deep, and Rei...
@KirkDBorne: 🌟Massive 774-page book by @rasbt 🌟 "#MachineLearning with #PyTorch and Scikit-Learn: Develop #ML and...

8. GeminiApp (Group Score: 51.1 | Individual: 22.1)

Cluster: 3 tweets | Engagement: 1376 (Avg: 850) | Type: Tech

Today we’re launching Google AI Plus in 35 new countries and territories, making it available everywhere Google AI plans are. 🌎

Google AI Plus includes: • More access to Gemini 3 Pro and Nano Banana Pro in the Gemini app • More access to Veo 3.1 in @FlowbyGoogle and the Gemini app • Expanded access to @NotebookLM • 200 GB storage that you can share with up to 5 family members

Existing Google One Premium 2 TB subscribers in these countries will also automatically get access to all the benefits of Google AI Plus in the next few days.

Learn more about Google AI Plus: https://t.co/4eUUK2ZKr8

See 2 related tweets

@TechCrunch: Google’s more affordable AI Plus plan rolls out to all markets, including the U.S. https://t.co/K1Fn...
@Google: RT @GeminiApp: Today we’re launching Google AI Plus in 35 new countries and territories, making it a...

9. AdinaYakup (Group Score: 50.3 | Individual: 20.2)

Cluster: 3 tweets | Engagement: 184 (Avg: 208) | Type: Tech

Big day in open source AI!!

✨ Kimi K2.5 just landed 🔥 @Kimi_Moonshot https://t.co/smd9IQQlIr

✨ DeepSeek released OCR2 💥 @deepseek_ai https://t.co/HWD8etwvBA

With the Chinese Spring Festival 3 weeks away,

what’s coming next?👀

See 2 related tweets

@HuggingPapers: RT @AdinaYakup: Big day in open source AI!!

✨ Kimi K2.5 just landed 🔥 @Kimi_Moonshot https://t.co...

@TechCrunch: China’s Moonshot releases a new open-source model Kimi K2.5 and a coding agent https://t.co/8e3OG8bu...

10. yoavgo (Group Score: 49.2 | Individual: 28.8)

Cluster: 2 tweets | Engagement: 111 (Avg: 45) | Type: Tech

OpenAI prism is a disastrous tool for science. Really the worst I could imagine. And it's going to be a huge success.

See 1 related tweets

@kimmonismus: The shift is clear: OpenAI wants to position ChatGPT as the most important product for research and ...

11. Reuters (Group Score: 47.3 | Individual: 23.7)

Cluster: 2 tweets | Engagement: 31 (Avg: 66) | Type: Tech

From TikTok sealing a new deal to avoid a ban in the US to a multifaceted robot who knows how to speak and sing, get all the latest tech updates in Tech Weekly https://t.co/emag5C05uS https://t.co/b1teMJxObf

See 1 related tweets

@ReutersBiz: WATCH: From TikTok sealing a new deal to avoid a ban in the US to a multifaceted robot who knows how...

12. AISafetyMemes (Group Score: 46.4 | Individual: 19.4)

Cluster: 3 tweets | Engagement: 316 (Avg: 295) | Type: Tech

Andrej Karpathy: "This is easily the biggest change in ~2 decades of programming and it happened over the course of a few weeks."

"I rapidly went from about 80% manual+autocomplete coding and 20% agents to 80% agent coding and 20% edits+touchups."

"I am bracing for 2026 as the year of the slopacolypse."

"LLM agent capabilities (Claude & Codex especially) have crossed some kind of threshold of coherence around December 2025 and caused a phase shift in software engineering."

"I am slowly starting to atrophy my ability to write code manually."

"It's so interesting to watch an agent relentlessly work at something. They never get tired, they never get demoralized, they just keep going and trying things where a person would have given up long ago to fight another day. It's a "feel the AGI" moment to watch it struggle with something for a long time just to come out victorious 30 minutes later."

See 2 related tweets

@svpino: “LLM agent capabilities (Claude & Codex especially) have crossed some kind of threshold of coher...
@jiayuan_jy: 2026 would be the start of AGI.

From Karpathy:

I rapidly went from about 80% manual+autocomplete...

13. bindureddy (Group Score: 45.8 | Individual: 26.4)

Cluster: 2 tweets | Engagement: 109 (Avg: 300) | Type: Tech

Open Source Tsuname Is Real - Kimi K2.5 Is The Best OSS Model In The World

I know DeepSeek is preparing to launch their model, but Kimi K2.5 just beat them to it! They are by far the best open-source model in THE WORLD.

Still, there is a considerable gap between them and the Anthropic models, but it's definitely closer than ever before

It is available RIGHT NOW on ChatLLM

See 1 related tweets

@abacusai: RT @bindureddy: Open Source Tsuname Is Real - Kimi K2.5 Is The Best OSS Model In The World

I know ...

14. huggingface (Group Score: 45.6 | Individual: 22.9)

Cluster: 3 tweets | Engagement: 114 (Avg: 60) | Type: Tech

RT @allen_ai: Introducing Ai2 Open Coding Agents—starting with SERA, our first-ever coding models. Fast, accessible agents (8B–32B) that ad…

See 2 related tweets

@allen_ai: RT @Tim_Dettmers: We release SERA, the first model part of Ai2’s Open Coding Agent series. SERA is a...
@allen_ai: RT @Kyle_L_Wiggers: 🚀 Big day at Ai2! We’ve launched Open Coding Agents, starting with SERA—our firs...

15. LiorOnAI (Group Score: 42.9 | Individual: 42.9)

Cluster: 1 tweets | Engagement: 1438 (Avg: 232) | Type: Tech

Claude can make blue1brown animations in minutes.

Education is about to explode. https://t.co/3a1dfqAlfO

16. Reuters (Group Score: 42.2 | Individual: 25.1)

Cluster: 2 tweets | Engagement: 20 (Avg: 66) | Type: Tech

Exclusive: A plan to build a synthetic diamond plant in the United States is a prime prospect in Japan's $550-billion investment package, as the allies push to expand production of a material vital to chip and high-precision manufacturing https://t.co/xR48EUE0U7

See 1 related tweets

@Reuters: Exclusive: Japan, US eye synthetic diamond production under $550 bln investment plan https://t.co/no...

17. matteocollina (Group Score: 41.8 | Individual: 41.8)

Cluster: 1 tweets | Engagement: 6770 (Avg: 391) | Type: Tech

RT @karpathy: A few random notes from claude coding quite a bit last few weeks.

Coding workflow. Given the latest lift in LLM coding capab…

18. brunoborges (Group Score: 41.4 | Individual: 41.4)

Cluster: 1 tweets | Engagement: 670 (Avg: 41) | Type: Tech

RT @aakashgupta: Anthropic employs world-class engineers who could build an HR system in weeks. They use Workday anyway. The reason tells y…

19. TechCrunch (Group Score: 39.6 | Individual: 22.3)

Cluster: 2 tweets | Engagement: 26 (Avg: 47) | Type: Tech

AI chip startup Ricursive hits $4B valuation two months after launch https://t.co/jzUO7V07PW

See 1 related tweets

@jm_alexia: RT @annadgoldie: Thrilled to announce our $300M Series A at a$ 4B valuation!

Chips are the fuel for...

20. haoningtimothy (Group Score: 39.5 | Individual: 39.5)

Cluster: 1 tweets | Engagement: 451 (Avg: 97) | Type: Tech

We are really taking a long time to prove this: everyone is building big macs but we bring you a kiwi🥝 instead. You have multimodal with K2.5 everywhere: chat with visual tools, code with vision, generate aesthetic frontend with visual refs...and most basically, it is a SUPER POWERFUL VLM