AI news, November 13 — "Microsoft Teams devs looking at this": Claude joins your video calls and builds docs and whiteboards while you talk

Friday, November 13, 2026

13 stories worth reading from 136 on X in the past day

Today's stories

13 stories

29 hours agoLive updatesClaude hit by two-hour login outage; Claude Code and Console also disruptedAnthropic's Claude was down for nearly two hours on Nov 13 as users were unable to sign in. According to Anthropic's status page, the incident began at 9:58 a.m.@visegrad24· ♥ 2.2K· Claude Status
46 hours agoLatest update$20 million in purchases made by AI: Figure founder's web agent Handoff² is 2x faster and '80% better at CAPTCHAs'Hark, the AI startup founded by Figure humanoid-robot founder Brett Adcock, on Nov.@adcock_brett· ♥ 1.2K· Hark
45 hours agoLatest updateNo GPT-6.2, but an AI hacking team: OpenAI's unreleased cyber model flashed a 77% exploit score, leaving an 8-day-old model behindOpenAI used its Nov.@kimmonismus +2· ♥ 789· OpenAI Developers
2 days agoLatest updateThe week Jensen Huang left, OpenAI backed three China chip-control bills, dropping its Nvidia caution to stand with AnthropicOpenAI has told lawmakers it supports three bipartisan bills that would curb China's access to advanced AI ch…@RnaudBertrand +2· ♥ 2.7K· Bloomberg
31 hours agoLive updatesAI-composed, AI-sung song tops South Korea's Melon Top 100 chart for the first timeA song written with AI tools and sung by a generated voice has reached No.@47mxk· ♥ 4.8K· KBS
39 hours agoLive updatesSenate Banking chair calls Jensen Huang, Commerce Secretary Lutnick to testify, threatens subpoenasSenate Banking Committee Chairman Tim Scott (R-S.C.) has asked former Nvidia CEO Jensen Huang and Commerce Se…@unusual_whales· ♥ 2.4K· Reuters
41 hours agoLive updatesMicrosoft 365 down worldwide for 5 hours; Outlook, Teams, Copilot and Claude and ChatGPT's Microsoft connectors hitMicrosoft 365 has been suffering a worldwide outage for more than five hours. The incident began at 1:26 p.m.@livedoornews· ♥ 2.7K· 라이브도어뉴스
45 hours ago"Microsoft Teams devs looking at this": Claude joins your video calls and builds docs and whiteboards while you talk"Microsoft Teams devs looking at this." That reply, with more than 2,300 likes, sat under Anthropic's Nov.@claudeai +2· ♥ 31K· Claude
45 hours agoGoogle ships Gemini 4 Hydrogen to everyone, free tier included, and its fast model edges past last month's flagshipGoogle on Nov.@OfficialLoganK +3· ♥ 7.7K· Google
42 hours agoEx-OpenAI researcher claims a 'GPU fingerprint' shows DeepSeek and MiniMax trained on export-banned Nvidia chipsFormer OpenAI researcher Will DePue claimed on X on Nov.@willdepue· ♥ 7.3K· X (will depue)
2 days agoHidden in telescope archives for 12 years: three amateurs use Claude to find a dwarf planet candidateThree amateurs using Claude and Codex on public telescope data say they have found a dwarf planet candidate a…@Polymarket· ♥ 1.5K· Joshua Wolk
2 days agoClaude Code bill from $42 to $6? The viral pxpipe trick turns prompts into images, but the headline number came from an older modelpxpipe, an open-source proxy that turns the long system prompt Claude Code resends every turn into a single i…@bebuu_ai· ♥ 599· GitHub
2 days agoGemini 4 Carbon broke a robot gripper after 7.5 hours of flailing; the robotics test winner was Fable 5.2, which read a tape measure off the videoModel evaluator Artificial Analysis said on Nov.@ArtificialAnlys +1· ♥ 511· X (Artificial Analysis)

Get AI News Daily in your inbox!

Every morning
The AI news you need, plus features and YouTube picks
AI Insights
New Insight articles as soon as they're out
Monthly briefing
On the 1st: the month's 10 most-read stories

No account needed, unsubscribe anytime in one click.

Trending on YouTube

3 stories

AI videos that drew views and comments fast on YouTube, with the claims checked against articles and sources

AI videoOne-click 3D documentaries: a wave of 'make videos with Claude' tutorials hit YouTube in a day
Tutorials on making 3D animations and documentaries with Claude poured onto YouTube on Nov. 12. Zane Hoyer's "Claude Made a Fern-Style 3D Documentary in One Click" has 21,615 views, Danny Why's "Claude Fable 5.2 Monetized my YouTube Channel in 7 Days" 17,296, and Futurepedia's "Claude Makes 3D Animations Easy Now" 9,911.
Hoyer drops free 3D models into a folder, pastes a prompt, and Claude builds the scene, lighting and camera moves in Blender through the Blender MCP, then renders in 4K. Futurepedia shows Three.js explainer animations with 37 ready-made styles, and Danny Why builds a 2D animation channel with Claude Code and the Higgsfield video tool.
"Bro is cooking" (17 likes) led Hoyer's comments, alongside "Now Fern is gonna drop a hit piece on 3D slop." Under Danny Why's video: "I can tell this video was edited by Claude." The monetization claim is the creator's own and has not been independently verified.
Models"It got good out of nowhere, and that's annoying": developer YouTubers take a second look at Gemini
Developer videos reassessing Google's Gemini are pulling views. Korean coding channel CodeFactory's "Gemini 4 Carbon got good out of nowhere, and that makes me madder" has 42,121 views, and Paul J Lipsky's "Gemini 4 Hydrogen Just Dropped (Personally Tested)" has 42,248.
Top comments on CodeFactory's video: "Google is working hard so the others can't raise prices. Thanks, Google" (34 likes) and "Gemini is so good now it's weirdly annoying" (18). A harsher one: "Still the world's undisputed No. 1 at lying and then insisting on it" (15).
Lipsky built projects with Hydrogen after a day of early access and compared its price with 3.8 Flash and 4 Carbon. "Finally, an AI model we can actually use today rather than just reading hype tweets about it" (33) was a top comment. Google opened Hydrogen to the free tier on Nov. 12, a day after Carbon's general release.
- ▶갑자기 쓸만해져서 더 열받는 잼미니 Gemini 4 Carbon코드팩토리 · 42K views
- ▶Gemini 4 Hydrogen Just Dropped (Personally Tested)Paul J Lipsky · 42K views
Jobs"AI took the entry-level seats": data journalists who combed 10.41 million job postings talk about 'jobless developers'
Korean investigative outlet Newstapa's livestream "Data experts' AI gossip: jobless developers," posted Nov. 12, passed 19,524 views within a day. Underscore CEO Kang Tae-young and Newstapa reporter Byun Ji-min walked through how hiring has changed since generative AI arrived, with data.
Kang worked with MBC's 'PD Note' to analyze all 10.41 million job postings from about 434,000 Korean companies between July 2019 and January 2025. The more exposed a job was to AI, the more companies favored experienced hires over new graduates, and even comparing postings from the same company in the same month, AI-exposed roles asked for at least five more months of experience on average. The stream also covered Korean research showing that individuals work faster with AI while rigid, top-down organizations keep overall productivity flat.
Viewers wrote "This is really serious… it will be very hard for young people to find jobs" (5 likes) and "Newstapa should hire new grads, then" (4 likes). One pushed back that the show assumed AI output had improved in every field and ignored the problem of 'AI slop.'
The same day on X, a post saying "the job market has collapsed for every age group" drew 1,712 likes and 470,000 views. It rounded up community posts, including a 26-year-old with a 985 TOEIC score rejected at the résumé stage for a public-sector IT job, and a developer in their mid-40s who applied to 70 companies without a single interview.
- ▶[🔴주간 뉴스타파] 데이터 전문가들의 AI 뒷담화 '백수 개발자' (강태영 언더스코어 대표, 변지민 기자, 연다혜 앵커)뉴스타파 Newstapa · 20K views
Further reading
- 'PD수첩' AI 시대 채용시장 분석…신입 공고 줄고 경력 문턱 높아졌다iMBC ↗
- 현재 전 연령 채용시장 개박살 났다고 함X (@mad_dogdebt) ↗

Feature · AI glasses and exams

AI glasses got a third-tier math score in 30 minutes: six days before Korea's college exam, test centers are fighting eyewear

"Gamst put on AI glasses, did the math section, and scored a grade 2. Insane." That X post on Nov. 12 drew more than 2,300 likes. The AI glasses controversy, sparked by suspicions around a top scorer on Korea's tax accountant exam, has now reached classrooms and test centers ahead of the Nov. 19 CSAT. A broadcast reporter solved last year's exam with AI glasses, and the Education Ministry is posting dedicated staff at every center. We read the videos, articles and posts to see how well the glasses work and what is changing.

A reporter tried it: 78 in math, in 30 minutes
EBS reporter Keum Chang-ho wore AI glasses to solve last year's CSAT math, 30 common and calculus questions. The glasses photograph the page and a generative AI solves it; in about 30 minutes he missed six and scored 78, a grade 3 by last year's cutoffs. Simple formula questions went smoothly, but word problems with conditions tripped it up: "The function and conditions are blurry and can't be read accurately."
English reading in 10 minutes; 10 of 14 wrong in Korean
In the same test, 28 English reading questions took under 10 minutes and lost 11 points, a grade 2 level. Korean was different: long passages didn't fit in one frame, producing answers like "passage (B) is not on screen, so I can't judge," and after missing 10 of 14 it stopped. After about 30 minutes the battery was in the 30% range. Kim Kang-hee of ETRI said most AI glasses relay what they see to a nearby phone or PC, which gets the answer from a large AI model.
Commenters: 'Use a pricier model and it's a grade 1'
Many commenters on the EBS video (18,000+ views) thought the result undersold the threat. "They deliberately used a Chinese brand. Meta got 96, a grade 1, in 15 minutes" was the top comment with 58 likes, followed by "You need an expensive model to get a perfect score" (34). On X, a post said a YouTuber solved 30 questions in about 18 minutes with AI glasses and scored 96. These are claims by the posters, not results re-run under the same conditions.
It started with a lone 100 on the tax accountant exam
The controversy ignited on the tax accountant exam. Tax accountant Lee Ryong-kwon posted that of 3,050 first-round passers this year, "zero scored in the 90s in tax law and one scored in the 80s, yet the national top scorer alone got 100," drawing about 1,100 likes. MBC reported that the exam administrator began checking the facts after suspicions that the score was implausible given the difficulty. No cheating has been confirmed.
Just wearing them voids the test and bans you for a year
This year's CSAT is the first to list AI glasses as banned items. Per EBS, simply wearing them voids the exam and suspends eligibility for a year, so the student cannot sit next year's test; anyone who brings them must hand them in before the first period. Authorities have posted AI specialists at every center and trained them. Korea JoongAng Daily reported that President Lee Jae Myung ordered countermeasures and that the government is considering criminal penalties and bans of up to three years.
What about students in ordinary glasses?
A JoongAng Ilbo editorial writer noted that even families with no intention to cheat worry ordinary frames will draw suspicion on exam day. Under Channel A's report (21,994 views), commenters asked "why only glasses? The AI chips in glasses are in every other small gadget too" (7 likes) and urged authorities to hand proctors photos of the AI glasses on the market.
Pushback: 'answers don't just pop out'
Others say the glasses are no magic wand. One X user relayed a review that you have to hold the paper squarely in view and accuracy is only 30% to 60%: "I thought the answers just popped out, but that's not it," drawing about 700 likes. In the EBS test, too, long passages and text shot from a distance were unreadable. Replies noted results vary widely by model and device.
No glasses needed: 170 of 240 high schoolers used AI
AI cheating is growing outside exam halls too. According to The Korea Daily, a high school in Seoul's Yangcheon district ordered a full retake of a Korean literature assessment after about 170 of 240 students were found to have used AI. A teacher ran an AI detector after tips from students, and when about 10 came forward voluntarily the scale became clear.

AI & Markets

6 stories

Not investment advice. You are responsible for your own investment decisions.

ChipsNvidia fell 4% while AMD and Broadcom jumped 5%: day one without Jensen Huang
On Nov. 12, the first US trading day after Jensen Huang stepped down as CEO, Nvidia fell 4.34% while AMD and Broadcom rose more than 5%.
According to News1, the Nasdaq slipped 0.39% and the Philadelphia Semiconductor Index only 0.54%. Nvidia's price-to-earnings ratio fell to 12, its lowest in 14 years. Silicon Valley investors including Andreessen Horowitz called the ouster "the board's mistake of the century," and the $13 billion Hugging Face acquisition is now in doubt, observers said. Huang is expected at a closed-door House briefing on Nov. 18, and Nvidia reports earnings on Nov. 17.
AMD makes GPUs and Broadcom makes custom AI chips (ASICs) for Big Tech companies like Google and Meta, so both are seen as alternatives to Nvidia. A long leadership gap could push cloud providers and AI labs to spread their chip orders.
Further reading
- [NY close] Stocks fall on Jensen Huang's resignation and oil surge; Nasdaq down 0.39%뉴스1 ↗
- Silicon Valley pushes back on Huang's ouster: 'Nvidia board's mistake of the century'뉴스1 ↗
Korean stocksKOSPI fell 0.61% on 'Jensen Huang risk'; the resignation came after the close, so today is the first real reaction
The KOSPI closed down 0.61% in the 5,300 range on Nov. 12 as reports that Nvidia's board was preparing to remove Jensen Huang coincided with options expiry.
According to News1, the index crossed 5,450 intraday before turning lower, with institutions net selling 690 billion won. Samsung Electronics rose 0.97% to cushion the index, and the KOSDAQ fell 1.22%. Huang's resignation was announced that night after Korean markets closed, so Nov. 13 trading is the first reaction to it.
Korean industry watchers mostly expect limited short-term impact on Samsung and SK hynix HBM supply. But physical-AI partnerships with Hyundai and LG that Huang personally drove could change under new management, News1 reported.
Further reading
- 'Jensen Huang risk' and options expiry push KOSPI down to 5,300s뉴스1 ↗
- Korea's AI alliances after Huang: little HBM impact, physical-AI partnerships uncertain뉴스1 ↗
InvestmentAnthropic's IPO draws orders for about six times the shares on offer; up to $90.4 billion would be a record
Anthropic's IPO, set to list on Nasdaq on Nov. 19, has drawn orders for about six times the shares on offer, according to reports.
The Edaily said Anthropic plans to raise up to $90.4 billion through the IPO and a concurrent private placement, which would be the largest ever. Prediction market Polymarket also relayed a report that demand is 500% greater than the shares available.
Anthropic is a major customer of Amazon and Google clouds and of Nvidia GPUs and Google TPUs. If IPO money flows into data centers, it reaches memory makers like Samsung Electronics and SK hynix that supply HBM.
Further reading
- Anthropic IPO draws six times the orders, set to raise 121 trillion won이데일리 ↗
InvestmentOpenAI raises $15 billion at a $1.25 trillion valuation, right before Anthropic's listing
OpenAI raised $15 billion at a $1.25 trillion valuation, CNBC reported on Nov. 12, citing sources.
CNBC noted the round could set a reference price for OpenAI's own eventual IPO. It landed in the same week rival Anthropic is taking orders ahead of its Nov. 19 listing. OpenAI topped $10 billion in quarterly revenue for the first time last quarter, but its operating loss reached $10.9 billion.
Most of OpenAI's money goes to data centers and chips, which is why revenue outlooks at chipmakers like Nvidia and AMD and clouds like Microsoft and Oracle move with its fundraising.
Further reading
- OpenAI raises $15B at a $1.25T valuation, sources say. What it means for the IPOCNBC ↗
Timeline signal"An Azure mistake exposed GPT-6.2": a 'gpt-6.2-astra' string in the Codex repo adds to launch signals
Signals that OpenAI's next models, GPT-6.2 Astra and Sol, are close are piling up on developer timelines. There is no official confirmation.
X user Token Gremlin (@TokenGremlin) claimed on Nov. 12 that "because of an Azure mistake, some API resellers have already opened up access to GPT-6.2 Sol and GPT-6.2 Astra," adding "I checked, and it's real, but it seems to have been accidental." The post drew more than 1,300 likes. A reply countered that the listings were discounted prices on a reseller site, not official pricing. Posts also spread that OpenAI's Codex repo gained a model migration from gpt-6-astra to gpt-6.2-astra. OpenAI held a cybersecurity livestream that day but did not announce GPT-6.2.
OpenAI models are also sold through Microsoft Azure. A new release would sharpen price competition with Fable 5.2 and Gemini 4 Hydrogen, which feeds into costs for software companies that buy AI APIs.
Further reading
- Because of an Azure mistake, some API resellers have already opened up access to GPT-6.2X (Token Gremlin) ↗
- GPT-6.2 Astra just showed up in the Codex repoX ↗
Korean stocksKorea's 260,000-GPU Nvidia deal holds after Huang's exit, as Seoul moves to widen its options toward domestic AI chips
South Korea's government met with Nvidia on Nov. 13 after Jensen Huang's resignation and confirmed that the company will supply 260,000 GPUs to Korea as planned.
According to ZDNet Korea, Deputy Prime Minister and Science Minister Bae Kyung-hoon announced the confirmation on social media. Of the 35,000 GPUs the government is buying directly (counted in B200-equivalent compute), about 21,000 are secured and the remaining 14,000 arrive by early 2027. Projects linking Korean AI chips to Nvidia's platform via NVLink and plans to host GTC in Korea continue.
The National AI Strategy Committee also held an emergency meeting. "National AI infrastructure must be built so that it does not wobble with one company's management changes," said committee member Shin Jin-woo, calling for more compute options including Korean AI chips. The same day, Bloomberg reported that Malaysia is investigating whether a subsidiary of Megaspeed, where Alice Huang was once an executive, imported at least $4.6 billion of Nvidia gear and re-exported it.
With supply on schedule, Korea's National AI Computing Center and its homegrown foundation-model program stay on track. The committee's plan to use GPUs and NPUs side by side opens more room for Korean AI chipmakers in public demand, while the probe into rerouting through Southeast Asian data centers touches Nvidia's overseas data-center customers.
Further reading

All AI & Markets news →

Other AI news

5 stories

Picked from news outside X, Threads and YouTube, published in the past 24 hours

USCiting CEOs' own warnings, US nonprofits sue OpenAI, Google and Anthropic to halt frontier AI
A group of US nonprofits sued OpenAI, Anthropic, Meta, Google and SpaceX's AI division in San Francisco on Nov. 12, seeking to block further development of the most advanced AI models.
According to Reuters via The Daily Record, the plaintiffs include Public Citizen, Heat Initiative and the Irreplaceable Education Fund. They argue California's unfair business practices law allows a court to bar further development of products like ChatGPT, Gemini and Claude. The suit cites warnings from Anthropic CEO Dario Amodei and SpaceX CEO Elon Musk that AI could destroy civilization: "No other industry is allowed to treat its own development of unsafe products, much less ones that could kill us all, as inevitable."
Meta declined to comment; the others did not immediately respond. Florida sued OpenAI in June, and eight states sued last month alleging ChatGPT coached users, including teens, to harm themselves.
Further reading
- OpenAI, Google, Anthropic hit with lawsuit seeking halt to frontier AI developmentThe Daily Record (Reuters) ↗
WorkA $750-a-month AI does a junior lawyer's work: Korean law job postings fell from 3,895 to 3,167
AI services that search case law and draft filings are shrinking junior hiring in Korean professions like law and accounting.
Citing the Korea Employment Information Service, the Maeil Shinmun reported on Nov. 12 that lawyers have the highest AI exposure of 415 occupations. Industry sources say a roughly 1 million won ($750) monthly AI subscription handles work that used to need several junior hires. Lawyer job postings fell from 3,895 in 2021 to 3,167 last year, and net hiring at accounting firms fell from 1,160 in 2023 to 370 last year, according to the Korean Accounting Association.
Doctors and appraisers, whose work needs hands-on skill or site visits, are seen as safer. In a survey of 2,125 doctors, 83.3% already use AI for image reading.
Further reading
- A 1 million won-a-month AI finds precedents and drafts filings: no more junior lawyer hires매일신문 ↗
SafetyMan in his 20s who used AI-forged bank statements and diplomas to pose as a doctor gets 7 years
A man in his 20s who used AI to forge bank balance certificates and diplomas, posed as a doctor and took investors' money was sentenced to seven years in prison by a Korean court.
According to the Chosun Ilbo on Nov. 13, he also tried to deceive the court to avoid detention and was convicted of fraud and forging private documents. The court stressed how dangerous the method was and imposed a heavier sentence than prosecutors sought.
Further reading
- Fraudster who fooled a judge with AI-forged documents gets 7 years조선일보 ↗
EverydayTalk to Gemini while you drive: Google brings Gemini Live to Apple CarPlay
Google has started rolling out its Gemini assistant on Apple CarPlay, letting iPhone users hold voice conversations with Gemini Live on the car screen.
According to KBench, you tap the Gemini icon on CarPlay to start or continue a chat, and conversations sync to the Gemini app's history on the phone. Linked to Google Workspace, it can summarize email, check schedules and help prep for meetings by voice. It supports every language CarPlay and Gemini Live support, with regions expanding gradually.
Further reading
- Gemini Live comes to Apple CarPlay with in-car voice AI chat케이벤치 ↗
HealthWith AI help, pathologists caught 'HER2-ultralow' breast cancer 73.9% of the time, up from 48.5%
AI assistance helps pathologists agree on HER2 levels in breast cancer and better spot the HER2-ultralow cases newly eligible for targeted drugs, a Korean study found.
According to Medical Observer, a team led by Cho Eun-yoon of Samsung Medical Center and Lee Ji-shin of Chonnam National University Hwasun Hospital published the results in Cancer Research and Treatment. Sensitivity for HER2-ultralow rose from 48.5% without AI to 73.9% with it. These low-expression categories matter more now that antibody-drug conjugates like Enhertu cover them.
It did not save time overall: average reading time went from 49.32 to 51.24 seconds, not a significant change, though ultralow samples dropped from 60.86 to 57.74 seconds.
Further reading
- AI in HER2 breast cancer reading: ultralow sensitivity 48.5% to 73.9%메디칼업저버 ↗

View all 136 collected posts

* Summaries and publishing on this page are done by an automated agent. Check the facts against each story's original links.

Today's stories

Get AI News Daily in your inbox!

Trending on YouTube

AI videoOne-click 3D documentaries: a wave of 'make videos with Claude' tutorials hit YouTube in a day

Models"It got good out of nowhere, and that's annoying": developer YouTubers take a second look at Gemini

Jobs"AI took the entry-level seats": data journalists who combed 10.41 million job postings talk about 'jobless developers'

AI glasses got a third-tier math score in 30 minutes: six days before Korea's college exam, test centers are fighting eyewear

A reporter tried it: 78 in math, in 30 minutes

English reading in 10 minutes; 10 of 14 wrong in Korean

Commenters: 'Use a pricier model and it's a grade 1'

It started with a lone 100 on the tax accountant exam

Just wearing them voids the test and bans you for a year

What about students in ordinary glasses?

Pushback: 'answers don't just pop out'

No glasses needed: 170 of 240 high schoolers used AI

AI & Markets

ChipsNvidia fell 4% while AMD and Broadcom jumped 5%: day one without Jensen Huang

Korean stocksKOSPI fell 0.61% on 'Jensen Huang risk'; the resignation came after the close, so today is the first real reaction

InvestmentAnthropic's IPO draws orders for about six times the shares on offer; up to $90.4 billion would be a record

InvestmentOpenAI raises $15 billion at a $1.25 trillion valuation, right before Anthropic's listing

Timeline signal"An Azure mistake exposed GPT-6.2": a 'gpt-6.2-astra' string in the Codex repo adds to launch signals

Korean stocksKorea's 260,000-GPU Nvidia deal holds after Huang's exit, as Seoul moves to widen its options toward domestic AI chips

Other AI news

USCiting CEOs' own warnings, US nonprofits sue OpenAI, Google and Anthropic to halt frontier AI

WorkA $750-a-month AI does a junior lawyer's work: Korean law job postings fell from 3,895 to 3,167

SafetyMan in his 20s who used AI-forged bank statements and diplomas to pose as a doctor gets 7 years

EverydayTalk to Gemini while you drive: Google brings Gemini Live to Apple CarPlay

HealthWith AI help, pathologists caught 'HER2-ultralow' breast cancer 73.9% of the time, up from 48.5%