Published on

科技推特精选 - 2026年2月14日

Authors

今日科技动态：随着 MiniMax-M2.5 正式开源并为智能体及编程提供顶尖性能，AI 产业格局正经历快速演变。OpenAI 报告指出，AI 现已几乎包揽其所有内部代码编写，显著提升了生产效率；与此同时，该公司正向美国立法者就中国 DeepSeek 带来的竞争威胁发出警示。另一方面，Anthropic 因涉嫌阻碍模型互操作性的策略而面临审查。在工程标准领域，Cloudflare 推出了旨在实现服务无缝重启的全新 Rust 库，而 AI 引发的市场波动则持续影响着全球 IT 行业的估值。

1. robbinfan (Group Score: 75.6 | Individual: 44.5)

Cluster: 4 tweets | Engagement: 842 (Avg: 91) | Type: Tech

我早就说过，ClaudeCode 就不要想着用别家的模型。

现在他们恶心的做法终于出现了：在里面随机增加一些空格，从而阻止其他模型使用 ClaudeCode，导致缓存命中率大幅度降低。这种恶心的小动作反映出 Anthropic 这家公司的价值观真的有问题。

所以我认为，与其用 ClaudeCode 还不如去用 OpenCode https://t.co/zBpeWSXa33

See 3 related tweets

  • @bearliu: 我靠，效率太高了！用 Claude Code 将我拖了两个月的项目网站，今天一下午就做完并上线。中间还去海里游了一个泳，做了个饭，带了组娃。

唯一阻挡进度的就是 Token需要重置，每次用完需要两三...

  • @lifesinger: 不少产品经理看了这条不舒服 何必呢

人就是人 有 AI 加持后 懂 Claude Code 的产品经理 就不是产品经理了

身份认识转变后 前途无量

如果依旧对产品经理念念不忘 那确实 基本完了...

  • @PandaTalk8: 我好喜欢claude code 的写的文章 。 相比于Chat GPT 的文章， 我觉得Claude Code 就像一个老实巴交的码农， 实实在在。 而ChatGPT 就像一个油腔滑调的老工贼。...

2. MiniMax_AI (Group Score: 71.9 | Individual: 56.9)

Cluster: 2 tweets | Engagement: 1728 (Avg: 192) | Type: Tech

MiniMax-M2.5 is now open source.

Trained with reinforcement learning across hundreds of thousands of complex real-world environments, it delivers SOTA performance in coding, agentic tool use, search, and office workflows.

Hugging Face: https://t.co/zfu7Am7yOg GitHub: https://t.co/uF3FNnb5AX Coding Plan: https://t.co/FDhZBBjQrX

Intelligence with Everyone

See 1 related tweets

  • @victormustar: RT @MiniMax_AI: MiniMax-M2.5 is now open source.

Trained with reinforcement learning across hundre...

3. lennysan (Group Score: 64.5 | Individual: 64.5)

Cluster: 1 tweets | Engagement: 2000 (Avg: 209) | Type: Tech

My biggest takeaways from @sherwinwu:

  1. AI is writing virtually all code at OpenAI. 95% of the engineers use Codex, and engineers who embrace these tools open 70% more pull requests than their peers, and that gap is widening over time.

  2. The role of a software engineer is shifting from writing code to managing fleets of AI agents. Many engineers now run 10 to 20 parallel Codex threads, steering and reviewing rather than writing code themselves.

  3. The average PR code review time has dropped from 10-15 minutes per PR to 2-3 minutes. Every pull request at OpenAI is now reviewed by Codex before human eyes see it, and Codex surfaces suggestions and catches issues up front. This allows engineers to focus on more creative and strategic work while dramatically increasing productivity.

  4. The models will eat your scaffolding for breakfast. When building AI products, don’t optimize for today’s model capabilities. The field is evolving so rapidly that the scaffolding (vector stores, agent frameworks, etc.) that seems essential today may be obsolete tomorrow as models improve.

  5. Build for where the models are going, not where they are today. The most successful AI startups build products that work at 80% capability now, knowing the next model release will push them over the line.

  6. Top performers become disproportionately more productive with AI tools. AI tools amplify the productivity of high-agency individuals, so the gap between top performers and everyone else is widening. The ROI on unblocking and empowering your best people compounds faster than ever in an AI-augmented environment.

  7. Most enterprise AI deployments have negative ROI because they’re top-down mandates without bottom-up adoption. Success requires both executive buy-in and grassroots enthusiasm. Sherwin recommends creating a “tiger team” of technically-minded enthusiasts (often not engineers) who can explore capabilities, apply AI to specific workflows, and create excitement throughout the organization.

  8. The one-person billion-dollar startup is coming, but with unexpected second-order effects. As AI makes individuals more productive, we’ll see not just billion-dollar solo founders but an explosion of small businesses: hundreds of 100Mstartupsandtensofthousandsof100M startups and tens of thousands of 10M startups. This will transform the startup ecosystem and venture capital landscape.

  9. Business process automation is an underrated AI opportunity. While Silicon Valley focuses on knowledge work, most of the economy runs on repeatable business processes with standard operating procedures. There’s massive potential to apply AI to these workflows, which are often overlooked by the tech community.

  10. The next two to three years will be the most exciting in tech history. After a relatively quiet period from 2015 to 2020, we’re now in an unprecedented era of innovation. Sherwin encourages everyone to engage with AI tools and not take this moment for granted, as the pace of change will eventually slow.

  11. AI models will soon handle multi-hour tasks coherently. Today’s models are optimized for tasks that take minutes, but within 12 to 18 months we’ll see models that can work on complex tasks for upward of six hours. This will enable entirely new categories of products and workflows.

  12. Audio is the next frontier for multimodal AI. While coding and text get most of the attention, audio is hugely underrated in business settings. Improvements in speech-to-speech models over the next 6 to 12 months will unlock significant new capabilities for business communication and operations.

4. Reuters (Group Score: 59.6 | Individual: 24.4)

Cluster: 3 tweets | Engagement: 228 (Avg: 122) | Type: Tech

OpenAI has warned US lawmakers that Chinese artificial intelligence startup DeepSeek is targeting the ChatGPT maker and the nation's leading AI companies to replicate models and use them for its own training, a memo seen by Reuters showed https://t.co/xxSfndc03p

See 2 related tweets

  • @Reuters: OpenAI has warned US lawmakers that China's DeepSeek is targeting the ChatGPT maker — and the nation...
  • @Reuters: OpenAI says China's DeepSeek trained its AI by distilling US models, memo shows https://t.co/IBWLaA9...

5. annadgoldie (Group Score: 56.9 | Individual: 56.9)

Cluster: 1 tweets | Engagement: 2381 (Avg: 17) | Type: Tech

Analogue 3D. A reimagining of the Nintendo 64. In 4k.

Limited Editions shipping now

Extreme Green. Ghost. Ocean. Glacier.

6. rohanpaul_ai (Group Score: 55.2 | Individual: 26.4)

Cluster: 3 tweets | Engagement: 12 (Avg: 114) | Type: Tech

The AI dev ecosystem has been moving fast as hell recently.

Another beautiful launch from @cline, its Cline CLI 2.0

From prompt to production. All in your terminal.

Run AI coding agents in parallel across your project, on real terminal, and inside CI/CD pipelines with ACP support for any editor.

You can also use local models via Ollama or any OpenAI-compatible endpoint. Point the CLI to your local server and it works offline.

Will be so useful, because it matches how devs actually work, where 1 thread is refactoring, another is updating docs, another is doing a quick investigation, and CLI 2.0 is explicitly built to run multiple isolated sessions without mixing context.

See 2 related tweets

  • @dr_cintas: Oh wow…Cline CLI 2.0 just dropped: open-source AI coding from your terminal.

It lets you run parall...

  • @cline: RT @dr_cintas: You can now run open-source AI coding agents without paying for API keys 🤯

Cline CLI...

7. business (Group Score: 55.0 | Individual: 28.2)

Cluster: 2 tweets | Engagement: 308 (Avg: 77) | Type: Tech

The rise of AI billionaires in China has ushered in a new kind of tech elite

China's new AI tycoons have amassed a collective $100 billion, rivaling Bill Gates's net worth. Diana Li explains https://t.co/xZxYRGR9Xq https://t.co/oDxTPf5dzO

See 1 related tweets

  • @business: China’s rising AI billionaires are challenging US dominance — and building huge fortunes on the back...

8. FireworksAI_HQ (Group Score: 54.7 | Individual: 29.7)

Cluster: 2 tweets | Engagement: 64 (Avg: 77) | Type: Tech

MiniMax M2.5 just dropped. It's already on Fireworks.

Excited to announce Fireworks AI as the day-0 launch partner for the brand new @MiniMax M2.5!

M2.5 is built for production agents at frontier performance at one-tenth the cost:

🟣Production-ready coding across the full dev lifecycle, not just bug fixes 🟣Autonomous agentic workflows with advanced search & tool use 🟣Deliverable office work - formatted docs, presentations, financial models 🟣Blazing speed: 100 TPS, matching Claude Opus 4.6 on complex tasks 🟣Run agents 24/7 without cost anxiety. $1/hour continuous operation.

We’re thrilled with MiniMax's velocity and excited to deepen our partnership, keeping the latest innovations flowing on Fireworks.

We're thrilled to be MiniMax's day-0 partners. Another day-0 launch on Fireworks! Fastest way to try what's new. -> https://t.co/s4ZPBKDcpg

See 1 related tweets

  • @MiniMax_AI: Proud to have @FireworksAI_HQ as our official day-0 launch partner for MiniMax M2.5!

If you're bui...

9. TheEconomist (Group Score: 51.3 | Individual: 17.5)

Cluster: 3 tweets | Engagement: 14 (Avg: 92) | Type: Tech

How can you incorporate AI into your role as a boss? How can your employees use it? Ask our management columnist. He’ll answer questions about the tech in our “Boss Class” podcast. Email your queries to podcasts[at]economist[dot]com https://t.co/FNMQgzfMh4

See 2 related tweets

  • @TheEconomist: Have questions about AI in the workplace? Our management columnist will answer them in a bonus episo...
  • @TheEconomist: Do you have questions about how AI could work at your organisation? Or in your job? Our management c...

10. bcherny (Group Score: 48.5 | Individual: 39.1)

Cluster: 2 tweets | Engagement: 3049 (Avg: 745) | Type: Tech

Love seeing how Spotify is shipping with Claude Code.

Their best developers haven't written a single line of code since December, they fix bugs from their phones, and they shipped 50+ features from Slack during morning commutes

https://t.co/rYTVJBHE0s

See 1 related tweets

  • @bcherny: RT @TechCrunch: Spotify says its best developers haven’t written a line of code since December, than...

11. googledevs (Group Score: 48.1 | Individual: 27.4)

Cluster: 2 tweets | Engagement: 3016 (Avg: 1145) | Type: Tech

RT @GeminiApp: Today, we’re releasing a significant upgrade to our specialized reasoning mode, Gemini 3 Deep Think.

Deep Think is built to…

See 1 related tweets

  • @GeminiApp: Ready to try the updated Gemini 3 Deep Think?

Google AI Ultra users can go to https://t.co/gTWOnOwQ...

12. llama_index (Group Score: 47.4 | Individual: 31.3)

Cluster: 2 tweets | Engagement: 27 (Avg: 36) | Type: Tech

🚀 The @posthog team has just rolled out LlamaIndex support for their LLM Analytics, and we built a demo to showcase what’s possible. Using LlamaIndex, LlamaParse, and OpenAI, our Agent Workflow compares product specifications and matches users with the most suitable option for their use case 🛠️ 🦔 Thanks to PostHog’s observability integration, the demo automatically tracks OpenAI usage, including: •Token consumption •Cost breakdown •Latency metrics

🎥 Check out the video below to see it in action 👇

👩‍💻 GitHub: https://t.co/elk5VKi8IF 📚 Docs: https://t.co/IZI3w6BYKy 🦙 LlamaCloud: https://t.co/wZjhFV29gN

See 1 related tweets

  • @jerryjliu0: RT @llama_index: 🚀 The @posthog team has just rolled out LlamaIndex support for their LLM Analytics,...

13. Cloudflare (Group Score: 46.2 | Individual: 28.4)

Cluster: 2 tweets | Engagement: 606 (Avg: 359) | Type: Tech

We’re open-sourcing ecdysis: a Rust library for graceful service restarts. No dropped connections, no manual socket handoffs—just seamless transitions when you update your code. https://t.co/l3pddMycxl

See 1 related tweets

  • @eastdakota: RT @Cloudflare: We’re open-sourcing ecdysis: a Rust library for graceful service restarts. No droppe...

14. Reuters (Group Score: 45.9 | Individual: 23.4)

Cluster: 2 tweets | Engagement: 92 (Avg: 122) | Type: Tech

Indian IT stocks set to lose $50 billion in worst week since pandemic on AI fears https://t.co/HBD9sQmDt7 https://t.co/HBD9sQmDt7

See 1 related tweets

15. AlexFinn (Group Score: 45.2 | Individual: 45.2)

Cluster: 1 tweets | Engagement: 6615 (Avg: 1189) | Type: Tech

We have entered a new age

An open source model just released that is:

• Better than Opus 4.6 for coding • Faster than Sonnet • State of the art for tool calling

I will be running Opus level superintelligence on my desk. For free. This quite literally changes everything

I will now be able to have a super intelligent AI model powering my OpenClaw that will search through X and Reddit 24/7/365 finding challenges to solve, then building apps out to solve those challenges, then shipping the apps live

All autonomously

A full, autonomous, software factory on my desk running 24/7 for free.

Imagine what happens when people realize what's now possible.

Totally secure, private, unlimited, free in your home super intelligence.

Nothing will be the same

16. guohao_li (Group Score: 44.0 | Individual: 44.0)

Cluster: 1 tweets | Engagement: 560 (Avg: 71) | Type: Tech

RT @PrimeIntellect: Introducing Lab: A full-stack platform for training your own agentic models

Build, evaluate and train on your own envi…

17. blackboxai (Group Score: 42.3 | Individual: 34.8)

Cluster: 2 tweets | Engagement: 72 (Avg: 35) | Type: Tech

MiniMax M2.5 is now live on BLACKBOX AI.

A frontier model designed for real world execution with strong reasoning, reliable tool use, and complex multi step workflows.

Engineered for demanding workloads. Ready for production scale orchestration.

Switch instantly in the BLACKBOX CLI.

Run /𝚖𝚘𝚍𝚎𝚕

Select 𝚋𝚕𝚊𝚌𝚔𝚋𝚘𝚡𝚊𝚒/𝚖𝚒𝚗𝚒𝚖𝚊𝚡-𝚖𝟸.𝟻

See 1 related tweets

  • @MiniMax_AI: RT @blackboxai: MiniMax M2.5 is now live on BLACKBOX AI.

A frontier model designed for real world e...

18. jm_alexia (Group Score: 41.8 | Individual: 41.8)

Cluster: 1 tweets | Engagement: 5054 (Avg: 468) | Type: Tech

RT @karpathy: New art project. Train and inference GPT in 243 lines of pure, dependency-free Python. This is the full algorithmic conten…

19. aakashgupta (Group Score: 40.7 | Individual: 40.7)

Cluster: 1 tweets | Engagement: 1236 (Avg: 449) | Type: Tech

Brendan Eich’s story is wild.

He built JavaScript in 10 days. May 1995, almost no sleep, because Netscape needed a scripting language before Navigator 2.0 shipped in September. He was 34. The prototype was called Mocha. For all of 1995 and most of 1996, he was the only developer working full-time on the engine.

That 10-day sprint now runs 98.8% of all websites on earth. JavaScript has been the most-used programming language for 13 consecutive years. 66% of all developers use it today. Every time you open Gmail, YouTube, or Netflix, you’re running code that traces back to those 10 sleepless nights in Mountain View.

He co-founded Mozilla in 1998 and helped spin it into an independent foundation after AOL gutted Netscape in 2003. Firefox went from zero to 30% browser market share. He proved browsers didn’t have to be a Microsoft monopoly.

Then Mozilla made him CEO in March 2014. Eleven days later, he resigned under public pressure over a political donation from six years earlier. The board tried to keep him in a different role. He walked entirely.

Here’s where most people’s story would end. He was 53, wealthy, had nothing left to prove. Instead he started Brave from scratch, raised 2.5MfromFoundersFund,andbuiltabrowserthatblockseveryadandtrackerbydefault.Herana2.5M from Founders Fund, and built a browser that blocks every ad and tracker by default. He ran a 35M ICO for Basic Attention Token in 2017. He built his own search engine doing 20 billion queries a year.

Brave crossed 100 million monthly active users in September 2025 and hit 100Minannualizedrevenue.Desktopmarketsharedoubledinasingleyear,from0.8100M in annualized revenue. Desktop market share doubled in a single year, from 0.8% to 1.3%. Against Chrome’s 20 billion search deal with Apple alone, and Google sitting at 65% global share.

The pattern across 30 years is the same every time. Eich builds something, gets told it won’t work or won’t scale, and then the thing he built becomes infrastructure that outlasts the people who doubted it.

The 10-day prototype became the language of the internet. The open-source side project became the second most popular browser on earth. The post-cancellation startup just crossed 100 million users.

Absolute legend.

20. aakashgupta (Group Score: 39.6 | Individual: 20.0)

Cluster: 2 tweets | Engagement: 38 (Avg: 449) | Type: Tech

Most PMs say they do user research. What they actually do is talk to 3 people, confirm what they already believed, and call it "validated."

The reason research stays shallow is time. Synthesizing 20 interviews manually takes 40+ hours. So teams cut corners. They interview 5 users, pattern-match to their hypothesis, and ship.

AI flips the bottleneck. When Claude can process 50 transcripts and surface contradictions between what users say and what they do, the constraint moves from "can we afford to do research" to "can we afford to ignore what the research found."

That second constraint is way harder to manage. And most product teams aren't ready for it.

See 1 related tweets

  • @aakashgupta: The constraint on good product decisions was never talent or frameworks. It was research bandwidth. ...
Discuss on Twitter • View on GitHub