科技推文精选 - 2026年3月14日

2026年3月14日科技每日简报

Today's top tech conversations are led by @CopilotKit, whose post about 'Introducing Open Generative UI...' garnered the highest engagement. Key themes trending across the top stories include google, https, gemini, search, system. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. CopilotKit (Group Score: 148.8 | Individual: 37.8)

Cluster: 6 tweets | Engagement: 176 (Avg: 20) | Type: Tech

Introducing Open Generative UI Repo 🌟

We built an open-source version of @claudeai's new feature for your own AI agents!

It's a template for building rich, interactive AI-generated UI with CopilotKit and @LangChain LangGraph.

Ask the agent to visualize algorithms, create 3D animations, render charts, or generate interactive diagrams — all rendered as live HTML/SVG inside a sandboxed iframe.

👾 Check it out: https://t.co/8Mh6QzWqvR

See 5 related tweets

@ataiiam: ✨Introducing: Open Generative UI

Open-source implementation of Claude's new Generative UI capabili...

@techdaily24: Claude AI can now generate interactive charts, diagrams, and visualizations. This new capability is ...
@ataiiam: Every agentic app should be utilizing these types of interactive, generative UIs within their chat. ...
@CopilotKit: RT @CopilotKit: Introducing Open Generative UI Repo 🌟

We built an open-source version of @claudeai'...

@CopilotKit: RT @ataiiam: @claudeai Generative UI is the way

We built CopilotKit to make this trivially easy to...

2. Meer_AIIT (Group Score: 92.7 | Individual: 34.1)

Cluster: 4 tweets | Engagement: 87 (Avg: 159) | Type: Tech

🚨google maps just became a real AI product and most people will miss why this matters

everyone is building AI chatbots that answer questions about the internet

google did something different

they connected Gemini to 300 million real places with real reviews from 500 million real people

so the AI actually knows the physical world around you

here's what "Ask Maps" lets you do now:

you type a question like "where can i charge my phone without waiting in a long coffee line"

Maps pulls from real place data and real reviews

gives you a custom map with actual options

and you can book or navigate from that same screen

the old way was opening maps, searching, reading 40 reviews, comparing then deciding

now you just ask one question and get a usable answer

the second part is even wilder

they rebuilt the entire driving navigation with 3D views that show you actual buildings and terrain around you in real time

Gemini analyzes Street View and aerial photos to build out what your route actually looks like before you drive it

so you can preview your parking spot and see which side of the street the entrance is on before you even leave

this is the template for how AI should work inside products people already use every day

just the thing you already open 10 times a week... now way smarter

rolling out now in the US and India on Android and iOS

See 3 related tweets

@techdaily24: Google Maps is introducing a new "Ask Maps" AI feature, powered by Gemini, alongside significant upg...
@rohanpaul_ai: RT @Meer_AIIT: 🚨google maps just became a real AI product and most people will miss why this matters...
@techdaily24: Ask Maps by Google is a new Gemini-powered conversational AI feature integrated into Google Maps. It...

3. cryptopunk7213 (Group Score: 88.1 | Individual: 38.4)

Cluster: 3 tweets | Engagement: 2514 (Avg: 760) | Type: Tech

holy shit Meta might ditch ai efforts and go with google gemini instead

Meta to delay their new AI model launch and use gemini to power Meta AI - HUGE fucking win for google:

Meta's avocado model underperformed frontier models from openai, google and anthropic (shitty reasoning, coding etc)
this comes after Meta spent $20B hiring a new AI team thats produced... no ai models.
looking at licensing google gemini (google just licensed to Apple for $1B per year)

Google is fast-becoming the preferred model for the largest companies in the world. Meta has 3.6 BILLION MAUs

if this happens google will single-handedly have the largest AI distribution of any company.

See 2 related tweets

@ns123abc: 🚨BREAKING: META delays AI model “Avocado” after it underperforms Google, OpenAI, Anthropic, xAI

be...

@zerohedge: RT @cryptopunk7213: holy shit Meta might ditch ai efforts and go with google gemini instead

Meta to...

4. CNBC (Group Score: 68.4 | Individual: 29.0)

Cluster: 4 tweets | Engagement: 262 (Avg: 55) | Type: Tech

AI agents could easily send college grad unemployment over 30%, ServiceNow CEO says https://t.co/56B3eD3sPq

See 3 related tweets

@danprimack: ServiceNow CEO: AI agents could send college grad unemployment over 30% in the next couple years. h...
@gregisenberg: "AI agents could EASILY send college grad unemployment over 30%" - ServiceNow CEO https://t.co/kz6Wt...
@Cointelegraph: 🚨 UPDATE: ServiceNow CEO Bill McDermott says unemployment among recent college graduates could rise ...

5. UnslothAI (Group Score: 51.4 | Individual: 28.8)

Cluster: 2 tweets | Engagement: 1388 (Avg: 1262) | Type: Tech

We collaborated with @NVIDIA to teach you about Reinforcement Learning and RL environments.

Learn: • Why RL environments matter + how to build them • When RL is better than SFT • GRPO and RL best practices • How verifiable rewards and RLVR work

Blog: https://t.co/Jng3urMPyw https://t.co/CmEj1S3QAe

See 1 related tweets

@akshay_pachaar: NVIDIA and Unsloth just dropped one of the best practical guides on building RL environments from sc...

6. ednewtonrex (Group Score: 51.1 | Individual: 24.6)

Cluster: 3 tweets | Engagement: 2683 (Avg: 841) | Type: Tech

Adobe CEO resigns, most likely due to Adobe floundering amidst the AI wave.

IMO Adobe made a big mistake leaning hard into generative AI. Their userbase hate it.

They will thrive if they reject gen AI, and return to putting humans first. That should be the focus of the next CEO

See 2 related tweets

@Reuters: Adobe's longtime CEO to exit role amid AI disruption, shares fall https://t.co/QV4du6nqJn https://t....
@business: Adobe CEO to Step Down in Face of Investor Concerns Over AI. Listen for more on Bloomberg Intelligen...

7. business (Group Score: 50.5 | Individual: 18.0)

Cluster: 3 tweets | Engagement: 38 (Avg: 128) | Type: Tech

Amazon plans to use chips from startup Cerebras Systems alongside its own Trainium processors, a combination that the companies say will be able to better run AI software https://t.co/fErTETzY4R

See 2 related tweets

@WSJ: Amazon Web Services plans to deploy processors designed by Cerebras inside its data centers, the lat...
@Reuters: Cerebras Systems, Amazon strike deal to offer Cerebras AI chips on Amazon's cloud https://t.co/GYgda...

8. aakashgupta (Group Score: 50.5 | Individual: 27.8)

Cluster: 2 tweets | Engagement: 94 (Avg: 156) | Type: Tech

The math problem in this tweet is so hard that one of history’s greatest mathematicians said if aliens threatened to destroy Earth unless we solved it, humanity should surrender.

That was his assessment of R(5,5), one specific Ramsey number. The search space for just that one problem contains roughly 10^271 possible graphs. A brute force solution would take longer than the age of the universe by a factor so large the number itself is meaningless.

For decades, every time a mathematician wanted to make progress on any single Ramsey number, they had to design a custom search algorithm from scratch. One researcher, one algorithm, one result. The best previous bounds on some of these numbers hadn’t moved in over ten years.

AlphaEvolve just improved five of them at once. One system. One deployment. Five new records, plus it matched or recovered every previously known result across the board, including cases where the original researchers never even published how they did it.

This is the same system that broke a 56-year-old record in matrix multiplication, the operation underneath essentially all of modern AI training. The same system that optimized Google’s data center scheduling, recovering 0.7% of their worldwide compute resources. The same system that simplified the chip designs powering Google’s own AI hardware.

One meta-algorithm that searches for search strategies. Pointed at pure math, it cracks decade-old conjectures. Pointed at infrastructure, it saves hundreds of millions in compute costs. Pointed at chip design, it finds simplifications human engineers missed.

When the tool that discovers better algorithms can also improve the training pipeline of the AI models powering itself, the ceiling on what it can solve next keeps rising.

See 1 related tweets

@demishassabis: Ramsey numbers are notoriously hard. Amazing to see AlphaEvolve improve bounds for 5 classical Ramse...

9. RoundtableSpace (Group Score: 48.5 | Individual: 28.6)

Cluster: 2 tweets | Engagement: 381 (Avg: 350) | Type: Tech

SOMEONE OPEN-SOURCED THE WORLD'S LARGEST PROMPT LIBRARY WITH 152K GITHUB STARS

> browse by category or AI search
> copy expert "Act as..." prompts
> paste into ChatGPT/Claude/Gemini/Llama/Mistral
> get god-tier results instantly

GitHub: https://t.co/dxPh7UJOgr https://t.co/xrThQNQcec

See 1 related tweets

@RoundtableSpace: RT @RoundtableSpace: SOMEONE OPEN-SOURCED THE WORLD'S LARGEST PROMPT LIBRARY WITH 152K GITHUB STARS ...

10. svpino (Group Score: 48.0 | Individual: 23.4)

Cluster: 3 tweets | Engagement: 89 (Avg: 245) | Type: Tech

This is cool!

An open-source version of Claude's new feature to generate interactive charts, diagrams, and UI components.

I'm really digging how fast open-source is catching up with big labs!

See 2 related tweets

@CopilotKit: RT @svpino: This is cool!

An open-source version of Claude's new feature to generate interactive ch...

@MillieMarconnni: RT @MillieMarconnni: 🚨 RIP Excel charts and PowerPoint diagrams.

Claude can now build interactive v...

11. edzitron (Group Score: 47.7 | Individual: 30.0)

Cluster: 2 tweets | Engagement: 83 (Avg: 654) | Type: Tech

Tomorrow's Premium: The Hater's Guide To The SaaSpocalypse. SaaS's biz model is collapsing under the weight of greed and the belief that software companies can grow forever. 15% of software loans may default in the next few years. Here's $10 off annual. https://t.co/1yICOXdFs3 https://t.co/zyNTwPZMow

See 1 related tweets

@edzitron: RT @edzitron: Tomorrow's Premium: The Hater's Guide To The SaaSpocalypse. SaaS's biz model is collap...

12. business (Group Score: 46.7 | Individual: 15.8)

Cluster: 4 tweets | Engagement: 83 (Avg: 128) | Type: Tech

Uber's co-founder Travis Kalanick has launched a new venture that will focus on creating “gainfully employed robots” for the food, mining and transport industries. https://t.co/a757nASde1

See 3 related tweets

@theinformation: .@Jessicalessin exclusively reports that Uber co-founder Travis Kalanick is plotting a new self-driv...
@theinformation: RT @Jessicalessin: Scoop from me: Travis Kalanick is in the early stages of starting a new AI and ro...
@zerohedge: *TRAVIS KALANICK PLANS NEW AI ROBOTICS COMPANY: INFORMATION https://t.co/JaJ4YRs6HE...

13. ibuildthecloud (Group Score: 46.5 | Individual: 46.5)

Cluster: 1 tweets | Engagement: 15582 (Avg: 173) | Type: Tech

RT @MyLordBebo: 🎂 How to know if the caller is an AI, just ask for a cupcake recipe!

lol … https://t.co/nNAOUfHIaq

14. jerryjliu0 (Group Score: 46.4 | Individual: 38.8)

Cluster: 2 tweets | Engagement: 201 (Avg: 54) | Type: Tech

Existing "OCR" technology for digitalizing PDFs has been around for ~30 years. Reading printed characters on a page and converting them into meaningful representations is a hard problem!

Existing approaches were either dependent on pattern matching to specific document templates, or on specialized ML models for specific data distributions. They constantly needed template/model refitting and broke on the long-tail of varied docs.

Today, vision models are capable of much higher general accuracy without constant retraining, but they still need careful orchestration to make sure that they're able to attend to specific elements (tables, charts), and output semantically correct outputs.

Our OCR platform LlamaParse is built on this "agentic OCR" foundation. A network of specialized agents will parse apart even the most complicated documents and reconstruct the outputs in a semantically meaningful way. We're excited to reach a world where raw parsing accuracy is not just 80% over "easy" docs, but 100% accurate over literally any document that exists.

Check it out: https://t.co/FeOoTjeKjf LlamaParse: https://t.co/TqP6OT5U5O

See 1 related tweets

@llama_index: RT @jerryjliu0: Existing "OCR" technology for digitalizing PDFs has been around for ~30 years. Readi...

15. dwarkesh_sp (Group Score: 46.0 | Individual: 32.3)

Cluster: 2 tweets | Engagement: 569 (Avg: 1309) | Type: Tech

.@dylan522p gives a deep dive on the 3 big bottlenecks to scaling AI compute: logic, memory, and power.

And walks through the economics of labs, hyperscalers, foundries, and fab equipment manufacturers.

Learned a ton about every single level of the stack.

0:00:00 – Why an H100 is worth more today than 3 years ago 0:24:52 – Nvidia secured TSMC allocation early; Google is getting squeezed 0:34:34 – ASML will be the #1 constraint for AI compute scaling by 2030 0:56:06 – Can’t we just use TSMC’s older fabs? 1:05:56 – When will China outscale the West in semis? 1:16:20 – The enormous incoming memory crunch 1:42:53 – Scaling power in the US will not be a problem 1:55:03 – Space GPUs aren't happening this decade 2:14:26 – Why aren’t more hedge funds making the AGI trade? 2:18:49 – Will TSMC kick Apple out from N2? 2:24:35 – Robots and Taiwan risk

Look up Dwarkesh Podcast on YouTube, Apple Podcasts, or Spotify. Enjoy!

See 1 related tweets

@scaling01: RT @dwarkesh_sp: .@dylan522p gives a deep dive on the 3 big bottlenecks to scaling AI compute: logic...

16. Cointelegraph (Group Score: 44.9 | Individual: 15.5)

Cluster: 3 tweets | Engagement: 662 (Avg: 639) | Type: Tech

🇨🇳 NEW: China is shifting bank lending from real estate to AI and tech as Beijing doubles down on winning the technology race. https://t.co/bldFQdTORW

See 2 related tweets

@Reuters: Chinese banks boost loans to tech sector as Beijing ramps up AI push https://t.co/L3O2OzQ5XB https:/...
@business: China's enthusiasm for the OpenClaw AI agent has sent shares of tech companies that embraced it soar...

17. javinpaul (Group Score: 44.2 | Individual: 23.4)

Cluster: 2 tweets | Engagement: 29 (Avg: 20) | Type: Tech

I Failed 4 FAANG Interviews Before Learning This - The Ultimate Tech Interview Guide Software Engineers https://t.co/n4FPkPR2OR https://t.co/zU8iQwVlaW

See 1 related tweets

@javinpaul: I Failed 4 FAANG Interviews Before Learning This - The Ultimate Tech Interview Guide Software Engine...

18. myfear (Group Score: 43.2 | Individual: 25.2)

Cluster: 2 tweets | Engagement: 15 (Avg: 24) | Type: Tech

“AI-ready” is not adding a chatbot. It’s building a Control Plane around your Java systems.

#MCP, semantic tool contracts, OAuth 2.1, deterministic routing are what turns #Spring & #Quarkus into agent-native platforms.

Deep dive here: https://t.co/u8ayIfExTb

#Java #AI

See 1 related tweets

@myfear: RT @myfear: “AI-ready” is not adding a chatbot. It’s building a Control Plane around your Java syste...

19. RoundtableSpace (Group Score: 43.2 | Individual: 25.5)

Cluster: 2 tweets | Engagement: 188 (Avg: 350) | Type: Tech

Listen to this 2:50 clip of Ben Affleck spitting nothing but facts about AI and the future

(He built an AI company that Netflix is offering $600M to acquire)

https://t.co/RKNngZzPue

See 1 related tweets

@MarioNawfal: RT @RoundtableSpace: Listen to this 2:50 clip of Ben Affleck spitting nothing but facts about AI and...

20. tbpn (Group Score: 42.5 | Individual: 26.1)

Cluster: 2 tweets | Engagement: 54 (Avg: 197) | Type: Tech

.@travisk says AI will make human labor even more valuable and in-demand than ever before:

"Let's say the entire world - everything in our world - was automated, except for plumbers. You had machines making buildings - you would basically have like a thousand buildings a day."

"How valuable would those plumbers be?"

"Each and every plumber would be like LeBron. Why? Because plumbing would be the long pole in the tent to progress. You can't get those thousand buildings unless you have a plumber."

"And by the way, you'd get so much efficiency everywhere else that you'd need millions of plumbers."

"Humans [are going to] become more and more valuable because they will be the long pole in the tent to progress - and that progress is going to accelerate and get faster and more robust."

See 1 related tweets

@pmarca: RT @a16z: “In 2026, AI is driving a 10x increase in the productivity of the individuals who know how...