科技推文精选 - 2026年3月5日

2026年3月5日 科技每日简报

Today's top tech conversations are led by @LangChain_OSS, whose post about '🔗 Announcing LangChain OSS Ski...' garnered the highest engagement. Key themes trending across the top stories include openai, human, workers, https, model. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. LangChain_OSS (Group Score: 84.7 | Individual: 29.4)

Cluster: 5 tweets | Engagement: 118 (Avg: 79) | Type: Tech

🔗 Announcing LangChain OSS Skills

LangChain has the most popular frameworks for building AI agents — and now your coding agent can be an expert in it.

We're excited to release the first iteration of LangChain OSS Skills, giving your agent expertise in our open source frameworks. The skills include guidance on how to use langchain, langgraph, and deepagents to effectively build agents.

➡️ Install our OSS skills for your coding agent here: https://t.co/PoOgOeLpMQ ➡️ Read more: https://t.co/9spxPkyd7J

See 4 related tweets

• @TDataScience: AI agents are more than hype. In this practical guide, Vyacheslav Efimov explains how to design and ...
• @hwchase17: RT @LangChain_OSS: 🔗 Announcing LangChain OSS Skills

LangChain has the most popular frameworks for ...

• @hwchase17: RT @LangChain: 🚀 Announcing LangSmith Skills + CLI 🚀

Agent improvements are increasingly driven by...

• @mariofusco: RT @rorypreddy: 📢LangChain4j for Beginners just hit 2️⃣0️⃣0️⃣ stars!!

Learn AI applications with L...

2. mark_k (Group Score: 81.8 | Individual: 39.1)

Cluster: 3 tweets | Engagement: 801 (Avg: 241) | Type: Tech

OpenAI is reportedly preparing to launch its next major model, GPT-5.4, which promises significant upgrades in both context capacity and problem-solving depth.

The most notable addition for @ChatGPTapp is a new "extreme" reasoning mode. This feature allows the model to dedicate significantly more time and computing power to complex, multifaceted questions, reducing the likelihood of errors on difficult prompts.

Alongside advanced reasoning, GPT-5.4 will feature a 1-million-token context window. This more than doubles the 400,000-token limit of the current GPT-5.2 model, bringing OpenAI to parity with the massive context capabilities already offered by rivals like Google and Anthropic.

The upcoming model is also optimized for long-running, multi-step tasks that can take several hours to complete. With better memory retention and improved adherence to constraints over time, these upgrades are expected to be a major boost for agentic workflows and automation tools like OpenAI’s Codex.

Source: The Information

See 2 related tweets

• @AlexFinn: Today is an extremely important day

A massive, landscape shifting technology is about to drop

Open...

• @scaling01: GPT-5.4 coming soon (as first leaked by humble self)

• exceeding 1 million context window

• featuri...

3. SawyerMerritt (Group Score: 66.7 | Individual: 47.5)

Cluster: 2 tweets | Engagement: 15858 (Avg: 1803) | Type: Tech

RT @elonmusk: Tesla will be one of the companies to make AGI and probably the first to make it in humanoid/atom-shaping form

See 1 related tweets

• @niccruzpatane: Elon Musk says Tesla Optimus Robot will probably be the first to embody AGI in humanoid/atom-shaping...

4. dkundel (Group Score: 64.8 | Individual: 26.5)

Cluster: 3 tweets | Engagement: 73 (Avg: 88) | Type: Tech

While Codex can do a lot for you (including building and running apps), sometimes you still want to jump into your IDE or another app.

We added some of the most common ones for Windows devs to the app including:

• Visual Studio
• Rider + PhpStorm
• Git Bash
• GitHub Desktop
• Cmder
• WSL
• Sublime Text

Plus a lot that were already supported on Mac! Let me know if we missed some.

See 2 related tweets

• @reach_vb: Codex for Windows is here! 🔥

Native Codex app on Windows via the Microsoft Store Local PowerShe...

• @gdgtify: RT @ajambrosino: The Codex app is now live on Windows.

The app runs both natively and in WSL, with ...

5. NVIDIAOmniverse (Group Score: 64.5 | Individual: 43.2)

Cluster: 2 tweets | Engagement: 254 (Avg: 63) | Type: Tech

Interested in building Gaussian Splatting pipelines for spatial intelligence?

Join us tomorrow at 11:00 AM PT 🤝

Hear Michael Rubloff from Radiance Fields explain how Gaussian Splatting is transforming real-time 3D reconstruction for geospatial digital twins and media & entertainment.

We’ll explore NVIDIA Omniverse workflows, including: ✅ City-scale capture ✅ Virtual production environments ✅ Dynamic scene rendering

👉https://t.co/uXcMhE0G3t

See 1 related tweets

• @NVIDIAOmniverse: Building Gaussian Splatting Pipelines for Spatial Intelligence with NVIDIA Omniverse https://t.co/jd...

6. sdrzn (Group Score: 64.5 | Individual: 33.1)

Cluster: 2 tweets | Engagement: 52 (Avg: 185) | Type: Tech

Anthropic CEO today: "This year will have a radical acceleration that surprises everyone. Exponentials catch people off guard. We are at the precipice of something incredible."

I pulled together all of Anthropic's research papers and here's what the timeline looks like if the exponential holds:

2026 Q2: AI does AI research. The feedback loop closes, and model releases accelerate noticeably. We'll see new models every few weeks (e.g. OpenAI tweeting "5.4 sooner than you Think.")

2026 Q3: First AI-assisted drug discoveries enter clinical trials in record time.

2026 Q4: Fully autonomous software engineering. The cost of building software is near zero, and we'll see a flood of new apps, tools, and services hit the market. Big tech companies will operate with dramatically smaller engineering teams.

2027 Q1: Scientific breakthroughs in medicine and energy. The public starts to see tangible benefits amidst massive job disruption. We will feel an uncomfortable combination of intense optimism and anxiety.

2027 Q2: Governance crisis as AI capabilities have outpaced regulation. ~3 companies hold unprecedented concentration of power, and public debate erupts over who controls these systems and what a functioning economy even looks like anymore. We'll see emergency legislation, expanded food/healthcare benefits, and universal basic income pilots.

2027 Q3/Q4: The world is fundamentally different in ways that would seem science fiction today. Dario's "country of geniuses in a datacenter" is reading every research paper ever published, finding connections across biology, chemistry, physics, and mathematics in ways no human ever could. Dario has talked about this extensively, he thinks AI's biggest impact won't be in tech, but in biology and medicine. By this point AI systems can do 10-15 years of drug research & development in months. Not just one drug, but dozens simultaneously - cancer treatments, neurodegenerative diseases, rare genetic conditions that the pharma industry never bothered investing in before. AI-generated movies, music, games, and art become mainstream. Computational power is the new oil. The countries and companies that control the most advanced AI systems have leverage over everything: military strategy, economic policy, scientific output, intelligence gathering. The US-China AI race is at peak intensity. AI "arms control" gets discussed at the UN for the first time similar to nuclear nonproliferation. The alignment question becomes urgent. Dario argues that we have a very narrow window to get alignment right before the systems are too capable to course-correct. The systems either have robust human values baked in, or they don't, and the consequences of getting it wrong are no longer hypothetical.

The exponential isn't some far-off sci-fi scenario. The world isn't moving linearly anymore and we are living in unprecedented times.

See 1 related tweets

• @chatgpt21: Dario Amodei ( CEO of Anthropic )

On AI acceleration:"We do not see hitting a wall. This year will ...

7. ycombinator (Group Score: 52.9 | Individual: 34.4)

Cluster: 2 tweets | Engagement: 401 (Avg: 125) | Type: Tech

CodeWisp (@usecodewisp) lets anyone create and publish web games with prompts.

No game engines or complex workflows. Just describe the game you want, generate a playable game in seconds, and share it instantly.

Congrats on the launch, @itsfuelvin!

https://t.co/XqzLVV9IFu https://t.co/y3sLwejeh0

See 1 related tweets

• @sxmawl: RT @ycombinator: CodeWisp (@usecodewisp) lets anyone create and publish web games with prompts.

No ...

8. NathanpmYoung (Group Score: 51.7 | Individual: 51.7)

Cluster: 1 tweets | Engagement: 6268 (Avg: 500) | Type: Tech

RT @robertwiblin: Huge repository of information about OpenAI and Altman just dropped — 'The OpenAI Files'.

There's so much crazy shit in…

9. aakashgupta (Group Score: 51.4 | Individual: 51.4)

Cluster: 1 tweets | Engagement: 10738 (Avg: 394) | Type: Tech

Everyone’s missing the real story here.

Meta’s Ray-Ban glasses need human data annotators to train the AI. When you say “Hey Meta” and ask the glasses to analyze something, that video gets sent to Meta’s servers, then routed to Sama, a subcontractor in Nairobi, Kenya. Workers there manually label objects in your footage. They see everything you recorded, intentionally or not.

7 million pairs sold in 2025 alone. Every single pair generates training data that flows through human eyes in Kenya. Workers told Swedish journalists they see people undressing, using bathrooms, having sex, and accidentally filming bank card details. One worker said “we see everything, from living rooms to naked bodies.”

Meta’s automatic face anonymization is supposed to protect people in the footage. Workers say it fails in certain lighting. Faces that should be blurred are sometimes fully visible. The person you recorded without knowing? A stranger in Nairobi can identify them.

Buried in Meta’s terms of service is one sentence doing enormous legal work: the company reserves the right to conduct “manual (human) review” of your AI interactions. That’s the legal cover for routing intimate footage from Western homes to a $2/hour labor force operating under NDAs, office surveillance cameras, and a strict no-questions policy. Workers say if you raise concerns about what you’re seeing, you’re fired.

This is the same company, Sama, that TIME exposed in 2023 for paying Kenyan workers $2/hour to label graphic content for OpenAI while being billed at$12.50/hour per worker. Workers described the experience as torture. Sama ended that contract, then pivoted to labeling Meta’s glasses footage. Same workforce. Same rates.

Meta markets these glasses as “designed with your privacy in mind.” The privacy design is a tiny LED light on the frame that most people don’t notice. The data pipeline behind it routes your bedroom footage to a contractor with a documented history of worker exploitation, failed anonymization, and union-busting lawsuits.

And the next generation of these glasses? Meta is planning to add facial recognition. The same system that can’t reliably blur faces in training data wants to start identifying them on purpose.

The LED light on the frame is doing about as much for your privacy as the terms of service nobody reads.

10. rohanpaul_ai (Group Score: 50.3 | Individual: 39.6)

Cluster: 2 tweets | Engagement: 371 (Avg: 106) | Type: Tech

Meta researchers created a mandatory checklist that forces AI to trace code line by line instead of blindly guessing.

This structured approach boosted the accuracy of checking real-world code updates to an impressive 93%.

Usually, when we ask an AI to check if a software update works, it just looks at the names of the functions and makes a very confident guess.

If we want to be absolutely sure the code works, human developers normally have to run the code in expensive and slow testing servers.

This paper changes that dynamic entirely by introducing a strict template that forces the AI to write down the exact path the code takes and provide hard evidence for every single claim it makes.

Because the AI is forced to slow down and show its work step by step, it catches deeply hidden bugs and proves that patches work with 93% accuracy.

The big deal here is that tech companies can now use AI to automatically and reliably verify millions of lines of code without ever paying for the massive computing costs required to actually execute that software.

Paper Link – arxiv. org/abs/2603.01896

Paper Title: "Agentic Code Reasoning"

See 1 related tweets

• @rohanpaul_ai: RT @rohanpaul_ai: Meta researchers created a mandatory checklist that forces AI to trace code line b...

11. WSJ (Group Score: 50.0 | Individual: 28.6)

Cluster: 2 tweets | Engagement: 147 (Avg: 328) | Type: Tech

Exclusive: OpenAI’s former chief research officer is raising $70 million for a new startup building an AI and software platform to automate manufacturing https://t.co/xkK8KoZp9X

See 1 related tweets

• @rohanpaul_ai: WSJ: Former OpenAI research chief Bob McGrew is raising $70M at a$700M valuation for a startup call...

12. openart_ai (Group Score: 49.1 | Individual: 35.5)

Cluster: 2 tweets | Engagement: 495 (Avg: 480) | Type: Tech

Last month, @openart_ai hosted something the industry said shouldn’t exist yet.

The first-ever AI Influencer Summit in San Francisco.

We brought together creators building AI-native identities, digital personalities signing brand deals, and a new generation redefining what influence looks like in an AI world.

At the time, it felt early. Experimental. Maybe even a little crazy.

But the room told a different story.

What many still call a “gimmick” is quickly becoming a new layer of the creative economy.

Claudine Wong and Lisa Bernard from FOX KTVU invited our founder and CEO Coco Mao to their podcast, Making Sense of AI 2gether, to talk about the rise of AI creators and the future of digital identity.

The future of influence isn’t coming.

It’s already being built.

See 1 related tweets

• @teortaxesTex: «I’m proud to see it stand as an exemplar of open-source innovation, driven by outstanding homegrown...

13. TheAhmadOsman (Group Score: 48.2 | Individual: 48.2)

Cluster: 1 tweets | Engagement: 2527 (Avg: 239) | Type: Tech

Friendly reminder that you don’t actually have to run heavy AI workloads on the laptop in your hands

You can just Tailscale / SSH into a machine with an RTX PRO 6000 somewhere else

Your battery, thermals, fans, and lap will appreciate that arrangement

14. mark_k (Group Score: 46.8 | Individual: 26.3)

Cluster: 2 tweets | Engagement: 148 (Avg: 241) | Type: Tech

The @OpenAI Codex app is now live on Windows. 🪟

It runs both natively and in WSL, with integrated terminals for PowerShell, Command Prompt, Git Bash, or WSL.

The release includes the first Windows-native agent sandbox, secured by OS-level controls that block filesystem writes outside your working folder and prevent outbound network access unless explicitly approved.

New additions: seven “Open in…” apps and two Windows skills for WinUI and ASP. NET.

Give it a try.

See 1 related tweets

• @ajambrosino: RT @OpenAIDevs: The Codex app is now on Windows.

Get the full Codex app experience on Windows with ...

15. kdnuggets (Group Score: 46.4 | Individual: 23.8)

Cluster: 2 tweets | Engagement: 0 (Avg: 2) | Type: Tech

Building Vertex AI Search Applications: A Comprehensive Guide - This guide explores the essential components, implementation strategies, and best practices for building production-ready search applications using Vertex AI Search and AI Applications. #AI Read more at: https://t.co/saX6ZImFwg

See 1 related tweets

• @kdnuggets: Building Vertex AI Search Applications: A Comprehensive Guide #AI Read more here: https://t.co/QjC7Y...

16. RoundtableSpace (Group Score: 45.0 | Individual: 28.4)

Cluster: 2 tweets | Engagement: 901 (Avg: 482) | Type: Tech

QWEN 3.5 9B RUNNING LOCALLY ON YOUR MAC WITH FULL TOOL USE.

NO API. NO CLOUD. A 9B MODEL BROWSING THE WEB FROM YOUR LAPTOP.

https://t.co/TluiEKd4BY

See 1 related tweets

• @MarioNawfal: RT @RoundtableSpace: QWEN 3.5 9B RUNNING LOCALLY ON YOUR MAC WITH FULL TOOL USE.

NO API. NO CLOUD. ...

17. mark_k (Group Score: 44.6 | Individual: 27.4)

Cluster: 2 tweets | Engagement: 135 (Avg: 241) | Type: Tech

It appears @OpenAI is building its own alternative to Microsoft's @GitHub.

According to recent reports, the project was sparked by a series of recent GitHub outages that disrupted their own engineers' ability to collaborate and ship code.

Given their deep partnership with Microsoft, this is a notable move. It will be interesting to see what an AI-native code repository built from the ground up might look like.

Source: The Information

See 1 related tweets

• @Reuters: OpenAI is developing alternative to Microsoft's GitHub, The Information reports https://t.co/9qEKM6m...

18. github (Group Score: 42.2 | Individual: 16.7)

Cluster: 4 tweets | Engagement: 187 (Avg: 190) | Type: Tech

GitHub Copilot Dev Days are coming to a city near you. 🏙️

Starting March 14, connect with your local dev community to level up your Copilot skills. Here's your opportunity to:

✅ Learn new features ✅ Watch live demos ✅ Apply real-world workflows

Find an event or sign up to host. 👇 https://t.co/iDXFC3oTxH

See 3 related tweets

• @JamesMontemagno: HOST A GITHUB COPILOT DEV DAY IN YOUR CITY!

💡Info: https://t.co/KgIYINo7wc 👉Form: https://t.co/KmlK...

• @burkeholland: RT @svpino: I've been using GitHub Copilot since it came out.

There are two areas I'd recommend dev...

• @JamesMontemagno: RT @github: GitHub Copilot Dev Days are coming to a city near you. 🏙️

Starting March 14, connect wi...

19. ianmiles (Group Score: 41.2 | Individual: 41.2)

Cluster: 1 tweets | Engagement: 8002 (Avg: 815) | Type: Tech

RT @nikitabier: @Winterrose @AHM3D_HAMDAN Last night, we found a guy in Pakistan that was managing 31 accounts posting AI war videos. All w…

20. exolabs (Group Score: 40.2 | Individual: 40.2)

Cluster: 1 tweets | Engagement: 762 (Avg: 92) | Type: Tech

Thanks @tim_cook for absorbing memory costs. Stacking macs is still the cheapest way to run frontier AI. https://t.co/1QNQYJ89ol