- Published on
今日科技推文精选 - 2026-02-20
- Authors
- Name
- geeknotes
今日科技聚焦：印度人工智能峰会成为行业焦点。信实工业（Reliance）宣布了1100亿美元的基础设施投资承诺，而正处于1000亿美元融资关键阶段的OpenAI则披露了与塔塔集团（Tata）的合作伙伴关系。与此同时，Google DeepMind推出了具备突破性智能体编程能力的Gemini 3.1 Pro。然而，现场也出现了一些变数：比尔·盖茨宣布退出，且竞争对手公司的CEO们避免了公开展示团结。在法印关系不断加强的背景下，专家提醒称，评估环节的缺失仍导致大多数人工智能原型难以投入生产应用。
1. FT (Group Score: 211.8 | Individual: 49.7)
Cluster: 6 tweets | Engagement: 3073 (Avg: 298) | Type: Tech
Bill Gates has pulled out of a keynote speech at a high-profile global AI summit in India, as the Gates Foundation continues to manage the fallout from its chair’s interactions with Jeffrey Epstein. https://t.co/AzTAThF8j1 https://t.co/ennmj9VVQ8
See 5 related tweets
- @WSJ: Bill Gates canceled his keynote address at an AI summit in India as the Microsoft co-founder faces g...
- @Reuters: Bill Gates pulled out of India's AI Impact Summit hours before his scheduled keynote address on Febr...
- @Reuters: Bill Gates pulls out of India’s AI Impact Summit just hours before his scheduled keynote, weeks afte...
- @business: Microsoft co-founder Bill Gates backed out of a keynote address at a high-profile Indian AI summit j...
- @business: Bill Gates pulled out of a keynote at India’s AI Summit just hours before he was set to speak, as sc...
2. mark_k (Group Score: 182.9 | Individual: 30.0)
Cluster: 9 tweets | Engagement: 452 (Avg: 446) | Type: Tech
The new Gemini 3.1 Pro from @GoogleDeepMind has officially dropped, bringing significant performance leaps across the board!
The agentic coding capabilities are the standout highlight of this release. The model sets an incredibly high bar for complex software engineering tasks, scoring 80.6% on SWE-Bench Verified and leading with 54.2% on the highly diverse SWE-Bench Pro (Public).
Beyond coding, the model shows massive gains in reasoning and scientific knowledge. It achieves 77.1% on ARC-AGI-2 abstract reasoning puzzles and a stellar 94.3% on GPQA Diamond, solidifying its place as a top-tier model for advanced problem-solving.
See 8 related tweets
- @LiorOnAI: Google just doubled reasoning performance in three months.
Gemini 3.1 Pro, the upgrade to their fla...
- @BlindGoose1337: RT @GeminiApp: Gemini 3.1 Pro is here: A smarter model for your most complex tasks.
Building on the...
@cryptopunk7213: fucking fantastic breakdown of google's new gemini 3.1 model:
#1 ranked intelligence model but at...
@demishassabis: RT @JeffDean: Today, we’re continuing to push the boundaries of AI with our release of Gemini 3.1 Pr...
@rohanpaul_ai: Gemini 3.1 Pro just dropped by @GoogleDeepMind with better coding and long context support.
Massive...
3. business (Group Score: 122.8 | Individual: 33.7)
Cluster: 5 tweets | Engagement: 600 (Avg: 106) | Type: Tech
OpenAI is partnering with the Tata Group in India on artificial intelligence technologies, including data center infrastructure that could become one of the largest in the country https://t.co/2yvJDKpiK2
See 4 related tweets
- @ReutersBiz: Tata Group announces a partnership with OpenAI to create India's first AI-optimized large-scale data...
- @Reuters: India's Tata signs up OpenAI as customer for data centre business https://t.co/y2xlbzzFSz https://t....
- @rohanpaul_ai: OpenAI secured 100MW of AI-ready data center compute in India from Tata with plans to scale to 1GW
...
- @TechCrunch: OpenAI taps Tata for 100MW AI data center capacity in India, eyes 1GW https://t.co/oA3TkQXsoc...
4. Forbes (Group Score: 100.5 | Individual: 26.1)
Cluster: 4 tweets | Engagement: 712 (Avg: 199) | Type: Tech
See Sam Altman’s Awkward Moment With Anthropic’s CEO At India’s AI Summit—As Bill Gates Drops Out Amid Epstein Scrutiny https://t.co/zn7IPclkcd https://t.co/EJfDRD98R8
See 3 related tweets
- @Reuters: Bill Gates pulls out of India AI summit amid Epstein scrutiny https://t.co/SRl52STmVE https://t.co/S...
- @Reuters: Bill Gates pulls out of India AI summit; anger mounts over organisational lapses https://t.co/L8VSEK...
- @BBCWorld: Bill Gates pulls out of India's AI summit amid Epstein files controversy https://t.co/1sScFvYKtX...
5. yupp_ai (Group Score: 98.0 | Individual: 30.0)
Cluster: 5 tweets | Engagement: 113 (Avg: 56) | Type: Tech
📢 New Model Drop: Gemini 3.1 Pro is now live on Yupp!
Congratulations to the @GoogleDeepMind team on today's release. We're looking forward to seeing what our community does with the latest Gemini - and how it performs on our user-preference leaderboards. https://t.co/cHnUPzyixE
See 4 related tweets
- @mark_k: Gemini 3.1 Pro is now live on Vertex AI! @GoogleAI is definitely delivering a release today. ✊ https...
- @addyosmani: RT @GoogleDeepMind: Gemini 3.1 Pro is here.
We’ve significantly improved the model’s overall intell...
- @GeminiApp: RT @Google: Today, we’re releasing Gemini 3.1 Pro.
It’s the same core intelligence that powers Gemi...
- @mark_k: Gemini 3.1 Pro coming today from @GoogleDeepMind 👌
Officially confirmed by Logan: https://t.co/y0Xm...
6. business (Group Score: 92.3 | Individual: 55.1)
Cluster: 3 tweets | Engagement: 2281 (Avg: 106) | Type: Tech
Sam Altman and Dario Amodei, CEOs of rivals OpenAI and Anthropic, refused to hold hands at an AI summit in India.
The two tech chiefs were positioned next to each other in a lineup of business and political leaders, and instead stood with their hands in the air and avoided eye contact https://t.co/GIItJ5hu7f
See 2 related tweets
- @MarioNawfal: 🇮🇳 Sam Altman (OpenAI) and Dario Amodei (Anthropic) couldn't even pretend to get along at the AI Imp...
- @TechCrunch: Altman and Amodei share a moment of awkwardness at India’s big AI summit https://t.co/F5p3QtiHK6...
7. Reuters (Group Score: 85.5 | Individual: 34.4)
Cluster: 4 tweets | Engagement: 969 (Avg: 155) | Type: Tech
'There are only a small number of years left for AI models surpassing the cognitive capabilities of most humans for most things,' Anthropic's CEO Dario Amodei said at the India AI Summit https://t.co/rLpHsedfuj
See 3 related tweets
- @business: Anthropic CEO Dario Amodei says AI models could surpass human cognitive capabilities within just a f...
- @kimmonismus: "AI has been exponential for the last 10 years. (...) There are only a small number of years left fo...
- @rohanpaul_ai: "There are only a small number of years before AI surpass the cognitive capabilities of most humans ...
8. business (Group Score: 83.2 | Individual: 24.7)
Cluster: 5 tweets | Engagement: 240 (Avg: 106) | Type: Tech
OpenAI is close to finalizing the first phase of a new funding round that is likely to bring in more than $100 billion, sources say https://t.co/rPLxh4moTn
See 4 related tweets
- @rohanpaul_ai: 🚨 BREAKING: OpenAI is close to locking in the 1st phase of a new funding round, likely raising over ...
- @TechCrunch: OpenAI reportedly finalizing 850B valuation https://t.co/T3a8TiLi7O...
- @rohanpaul_ai: RT @rohanpaul_ai: 🚨 BREAKING: OpenAI is close to locking in the 1st phase of a new funding round, li...
- @Cointelegraph: RT @MSBIntel: 🚨 ABSOLUTELY SURREAL.
OPENAI IS ABOUT TO RAISE $100 BILLION IN A SINGLE FUNDING ROUND...
9. business (Group Score: 76.0 | Individual: 44.8)
Cluster: 2 tweets | Engagement: 689 (Avg: 106) | Type: Tech
Billioniare Mukesh Ambani announced that his retail-to-telecom conglomerate will invest as much as $110 billion in building AI-related infrastructure https://t.co/W8dTkzwHps
See 1 related tweets
- @Forbes: Reliance Industries—controlled by Asia’s richest person Mukesh Ambani—is doubling down on its AI exp...
10. kdnuggets (Group Score: 66.6 | Individual: 29.0)
Cluster: 3 tweets | Engagement: 5 (Avg: 2) | Type: Tech
Agentic AI doesn’t fail because models aren’t powerful enough.
It fails because agents lack shared meaning.
This webinar demonstrates how Databricks Agent Bricks and AtScale’s semantic layer work together to give AI agents deterministic metrics, governed access, and predictable behavior at scale.
If you’re building conversational analytics, internal GPTs, or autonomous agents on Databricks, this session outlines what it takes to move from demo to production.
See 2 related tweets
- @DataScienceDojo: 🚨 Most AI coding agents fail before they reach production. Not because the model is weak. Not becaus...
- @kdnuggets: ADK from Google addresses a critical gap in the agentic AI ecosystem by providing a framework that s...
11. aakashgupta (Group Score: 64.6 | Individual: 32.6)
Cluster: 2 tweets | Engagement: 62 (Avg: 476) | Type: Tech
.@AnkythShukla made a bold claim.
"More than 95% of AI prototypes fail when it comes to production. And the reason is evaluations."
This matters more than most teams realize. Here is why:
The AI prototype-to-production gap is real and it is widening. Teams build demos that impress leadership, ship them, and then watch support tickets explode. The root cause is almost always the same: no evaluation framework existed before launch.
What Shukla is pointing to:
AI outputs are non-deterministic. The same prompt produces different results. You cannot test AI features the way you test traditional code.
Most teams only do offline evals (testing before launch). Almost nobody runs online evals (monitoring in production) or human evals (spot-checking real user interactions) properly.
Without an evaluation rubric - clear criteria for what "good" looks like - you are shipping blind. No baselines. No metrics. No rollback criteria.
The solution is not more engineering. It is a product management skill: defining success criteria, building scoring rubrics, running LLM-as-judge pipelines, and monitoring quality continuously post-launch.
The structural insight: AI evals is not a QA function. It is a core PM competency. The teams that figure this out ship AI products that survive production. The teams that skip it burn six months of work and roll back.
Shukla has taught this to thousands. Today he is giving it away for free.
Full breakdown with @aakashgupta.
See 1 related tweets
- @aakashgupta: Most AI products don't fail at the demo stage. They fail after launch.
The prototype works. Stakeho...
12. aakashgupta (Group Score: 63.4 | Individual: 63.4)
Cluster: 1 tweets | Engagement: 4171 (Avg: 476) | Type: Tech
A 15-year-old girl immigrates to New Jersey from China. Doesn’t speak English. Her parents, both educated engineers back in Chengdu, are now working as cashiers and restaurant cooks. She gets a job washing dishes at a Chinese restaurant to help the family survive.
She gets into Princeton on a full scholarship. Her reaction is so disbelieving she asks two different advisors to verify the acceptance letter is real. Then her mom gets sick, so the family opens a dry cleaning shop in Parsippany. Every weekend for seven years, Fei-Fei Li leaves Princeton’s physics department to run the register, handle inspections, talk to customers, manage billing. Monday through Friday: quantum mechanics problem sets. Saturday and Sunday: sorting other people’s laundry. She later called herself the “CEO” of the dry cleaning business. She kept running it remotely through half of her PhD at Caltech.
In 2007, she proposed building an image dataset so massive her own mentor told her she’d taken the idea “way too far.” Pre-ImageNet, the entire AI field was working with datasets containing a few hundred images. She built one with 15 million. Most researchers at the time believed algorithms were the bottleneck. She bet on data when nobody else would.
By 2012, a team ran a neural network on that dataset and halved the existing error rate overnight. AlexNet on ImageNet became the moment the deep learning era started. Every computer vision product shipping today traces its lineage back to that dataset.
Fast forward to 2024. She starts World Labs. Four months in, 1 billion valuation. Today, 5 billion.
The bet investors are making: that the woman who gave AI its eyes with 2D image recognition is about to give it spatial awareness of the 3D physical world. Her new model, Marble, generates persistent 3D environments from text or images. Unlike video generators that fake depth frame by frame, Marble creates actual geometric space where objects stay where you left them.
The investor list tells you everything. AMD and NVIDIA both wrote checks. When the two biggest competing chipmakers both fund the same startup, they’re telling you this workload is coming whether their competitor funds it or not. Autodesk put in $200 million and signed on as strategic advisor, which means they see spatial AI integrating directly into CAD and design workflows within 18 months.
From dry cleaner to ImageNet to a $5 billion spatial intelligence company. Fei-Fei Li has now placed two bets that the rest of the field thought were too early and too big. The first one created modern computer vision. The second one is trying to give machines the ability to understand physics.
If she’s right again, this is the last major unlock before embodied AI actually works.
13. MIT_CSAIL (Group Score: 59.0 | Individual: 36.7)
Cluster: 2 tweets | Engagement: 380 (Avg: 124) | Type: Tech
A free MIT course about building scalable & high-performance software systems: https://t.co/BxL5RulDp5 (v/@MITOCW)
Here, MIT prof. & CSAIL principal investigator Charles Leiserson discusses why performance matters in computing (Lecture 1). https://t.co/x3tkqWF0Nt
See 1 related tweets
- @MIT_CSAIL: RT @MIT_CSAIL: A free MIT course about building scalable & high-performance software systems: ht...
14. bindureddy (Group Score: 58.0 | Individual: 26.7)
Cluster: 3 tweets | Engagement: 130 (Avg: 213) | Type: Tech
🚨 Secure OpenClaw - Message Your Custom AI Agent From Telegram
We are excited for our first release of personal agents. You can now use Abacus AI's DeepAgent to vibe code a personal agent on Telegram
Create and simply message the agent!
- do tasks like respond to emails
- schedule meetings
- use the browser to book tickets
- connect to your internal systems
The agent runs on a secure environment with constraints and you can pretty much customize it exactly like you want it when you are building it
This is just v1 - we are going to be launching something WAY more exciting soon
See 2 related tweets
- @abacusai: RT @bindureddy: 🚨 Secure OpenClaw - Message Your Custom AI Agent From Telegram
We are excited for o...
- @abacusai: 🚨 Your personal AI agent is here!
Vibe code your own agent with DeepAgent, message it on Telegram, ...
15. Reuters (Group Score: 54.5 | Individual: 20.2)
Cluster: 3 tweets | Engagement: 276 (Avg: 155) | Type: Tech
French President Emmanuel Macron said that India and France would work together to build the future of artificial intelligence, at the AI Impact Summit in New Delhi https://t.co/Tu5zmQd9Br https://t.co/voU2RPxGtB
See 2 related tweets
- @sundarpichai: Enjoyed meeting with President @EmmanuelMacron at the AI Impact Summit, and discussing how France's ...
- @ReutersBiz: Watch: French President Emmanuel Macron said that India and France would work together to build the ...
16. KirkDBorne (Group Score: 54.4 | Individual: 24.9)
Cluster: 3 tweets | Engagement: 29 (Avg: 42) | Type: Tech
"Master #Python Fundamentals: Ultimate Guide for Beginners" — The Number #1 Python Book For Beginners with Extra 300+ Hands-on Practice Questions
Get this brilliant book at https://t.co/kHA2mRTU0d by @RealBenjizo ———— #DataScience #DataScientist #ComputationalScience https://t.co/uxP6OmnRld
See 2 related tweets
- @KirkDBorne: Tiny #Python Projects — 21 small fun projects for Python beginners.
Source code & videos: https...
- @KirkDBorne: 77 #Python Advanced Programming Exercises — Complete Python Concepts covered in detail. Prepare For ...
17. business (Group Score: 54.0 | Individual: 29.4)
Cluster: 3 tweets | Engagement: 446 (Avg: 106) | Type: Tech
OpenAI CEO Sam Altman says we should remain “humble” about what we still don’t know about superintelligence.
He pointed to the uncertainty around how the technology could be used by totalitarian governments while speaking at India's AI summit: https://t.co/nCcJlggj75 https://t.co/zIvbyKTVjU
See 2 related tweets
- @minchoi: Not AGI. Superintelligence.
Sam Altman just said this today at the India AI Impact Summit:
"By the...
- @minchoi: RT @minchoi: Not AGI. Superintelligence.
Sam Altman just said this today at the India AI Impact Sum...
18. BrianRoemmele (Group Score: 51.6 | Individual: 51.6)
Cluster: 1 tweets | Engagement: 7226 (Avg: 387) | Type: Tech
RT @elonmusk: You can now tap on For You and choose general topics, such as “Technology”!
Update your 𝕏 app to make this work.
19. kimmonismus (Group Score: 51.3 | Individual: 23.3)
Cluster: 4 tweets | Engagement: 1408 (Avg: 418) | Type: Tech
Holy sh*t, thats what I call an improvement!
Gemini 3.1 pro is insane:
- Arc agi 2 77%
- SWE verified 80%
- HLE 44%/51%
See 3 related tweets
@Cointelegraph: 🔥 UPDATE: Sundar Pichai announces Gemini 3.1 Pro, scoring 77.1% on ARC-AGI-2, marking a major leap i...
@demishassabis: RT @scaling01: Google is now dominating ARC-AGI-2 with Gemini 3 Flash, Gemini 3.1 Pro and Gemini 3 D...
@demishassabis: RT @arcprize: Gemini 3.1 Pro on ARC-AGI Semi-Private Eval @GoogleDeepMind
ARC-AGI-1: 98%, $0.52/t...
20. EHuanglu (Group Score: 50.4 | Individual: 25.7)
Cluster: 2 tweets | Engagement: 1702 (Avg: 658) | Type: Tech
RT @EHuanglu: this is the biggest revolution in film history
AI can literally make an entire transformer film in days.. they say Seedance…
See 1 related tweets
- @EHuanglu: this is the biggest revolution in film history
AI can literally make an entire transformer film in...