AI news, November 8 — Claude deletes a UChicago professor's year of lecture notes while he's at lunch… all he asked was to fix page numbers

Sunday, November 8, 2026

7 stories worth reading from 12 on X in the past day

Listen to today's news

Every story, read aloud.

0:000:00

Today's stories

7 stories

18 hours agoLatest updateDid Qwen 4 Flash beat Opus 5.5? Result from one game-balance benchmark draws 900 likes; its maker says "not correlated with general capabilities"A screenshot saying an early version of Alibaba's next-generation Qwen 4 beat Anthropic's Claude Opus 5.5 on a benchmark spread on X on Nov. 8.@Layton_Gott· ♥ 905· X (@Layton_Gott)
31 hours agoLatest update"I can't hear sounds": Claude explains in a video how it makes music with code and math, Japanese user's post draws 5,700 likesA post saying Claude, which tells users "I can't hear sounds" when asked to compose, made music from code and…@550Atelier· ♥ 5.7K· X (@550Atelier)
24 hours agoLatest updateAnthropic CEO "doesn't know his home address" in sworn testimony: record draws 1M views, with pushback that putting addresses on public record is the real riskA court-record screenshot of Anthropic CEO Dario Amodei saying under oath that he does not know his own home…@Kalshi· ♥ 7.7K· X (Kalshi)
26 hours agoClaude deletes a UChicago professor's year of lecture notes while he's at lunch… all he asked was to fix page numbersUniversity of Chicago law professor William Baude said Anthropic's Claude deleted a year of his constitutiona…@Dexerto· ♥ 14K· X (William Baude)
2 days agoClaude Code put account emails in web requests' User-Agent; Anthropic shipped a guard module Nov. 5 but calls it a stopgapAn issue spread among developers saying Anthropic's coding agent Claude Code sent the signed-in account's ema…· DEV Community
38 hours agoClaude Max '20x' label heads to court: class action says weekly caps leave 6-8x ProA class action over whether Anthropic's Claude Max 20x plan delivers what it advertises is ongoing, and a pos…· Reddit
32 hours agoHinton: 'Chernobyl' AI disaster 'later this year or next'; Polymarket summary draws 2,400 likesA post saying Geoffrey Hinton warned "either later this year or next year we'll see Chernobyl" for AI spread…@Polymarket· ♥ 2.4K· X (Polymarket)

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

4 stories

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

Jobs"I asked 100 devs what AI is doing to their jobs": HealthyGamerGG video hits 42,000 views; top comment asks whether a 2024 report can be trusted
A video by mental-health coaching channel HealthyGamerGG, 'I Asked 100 Devs What AI Is ACTUALLY Doing To Their Jobs,' passed 42,000 views less than a day after upload on Nov. 7.
Its chapters cover emerging problems from AI, mental health in the AI era, the DORA report, stability of AI code, code review and company strategy, focusing on developers' anxiety and communication with management.
Comments were split. The top comment (200 likes) asked how far a 2024 report can be trusted, calling it "basically decades ago" in AI-model terms. Another, "You know it's bad for software engineers when Dr. K is focusing on software engineers," drew 192. A self-described working developer said an earlier video "missed the point" and sounded like their management (75).
- ▶I Asked 100 Devs What AI Is ACTUALLY Doing To Their JobsHealthyGamerGG · 42K views
Dev toolsAnthropic dropped Git, OpenAI made Git faster: Better Stack video at 54,000 views, daily code changes 2,500 to 30,000
Better Stack's video 'Anthropic Replaced Git. OpenAI Did the Opposite.' reached 54,000 views within about half a day of its Nov. 7 upload.
According to the description, Anthropic went from merging about 2,500 code changes a day to over 30,000. It built Scion, a new version-control system for AI agents, reporting 200x lower CPU and memory use than its previous Git mirroring. OpenAI built Pando, a Git backend that cut median push latency by nearly 80% without changing how developers use Git. The video also shows how to run multiple AI agents in parallel with Git worktrees.
The most-liked comment (113) said 30,000 changes a day "doesn't make a better UX"; a critical one called it "an Anthropic skill issue" (13).
- ▶Anthropic Replaced Git. OpenAI Did the Opposite.Better Stack · 54K views
SecurityClaude Code's "Don't Ask Again" left passwords on GitHub: SavvyNik video at 44,000 views, secrets in 1.1% of settings files
Security and Linux channel SavvyNik (110,000 subscribers) posted "Claude Code's 'Don't Ask Again' Is a Security Nightmare.." on Nov. 8, and it reached 44,000 views and 909 likes within a day.
The video covers research Apiiro published on Nov. 5. When a user approves a shell command in Claude Code with "Yes, and don't ask again," the whole command is saved to .claude/settings.local.json, so a password typed in the command stays in the file. If that file is committed to a public GitHub repo, the password goes public. Of 104,185 public copies, 1,122 (about 1.1%) held a secret, and across git history and other files the researchers counted 2,539 secrets in 1,994 repositories.
Anthropic says that when Claude Code first writes the file it adds an entry to the developer's global git excludes file, but that setting does not travel with forks or copies. Apiiro did not test whether the exposed credentials still work, and the same problem was raised in a GitHub issue in August 2025. Suggested fixes: rotate exposed keys first, stop tracking personal settings files, and reference secrets via a secret manager or environment variable instead of typing them into commands.
Comments were sardonic. "You need to add 'make no mistake' in the prompt" drew 74 likes, "Who is letting an AI touch their API keys to begin with" drew 67, and a comment wondering how many people will fix it by telling an AI to rotate the compromised keys drew 75.
- ▶Claude Code's "Don't Ask Again" Is a Security Nightmare..SavvyNik · 44K views
Further reading
- Claude Code approval files found exposing company secrets on GitHub, says Apiiro researchersITWire ↗
- [BUG] Sensitive tokens stored in .claude/settings.local.json and not ignored by git by defaultGitHub ↗
SecurityAI broke into a PHP server in 17 minutes: MBC News video at 60,000 views, about 110,000 Korean hosts suspected vulnerable
MBC Newsdesk's Nov. 7 report "A skull appears in 17 minutes... PHP server easily breached in AI penetration test" reached 60,600 views and 201 likes in a day.
According to MBC, the AI found an old-version PHP server and started the attack, and a skull image appeared after 17 minutes 42 seconds. The AI also wrote a report listing the attack commands it used.
MBC first reported the flaw on Oct. 22. Security firm Calif found the attack path while reviewing source code with an OpenAI model and reported it, and Google-owned Wiz independently reported two more. The key issue is that code can be executed without logging in. The Korea Internet & Security Agency issued an urgent notice that PHP versions below 8.5.12 are exposed, and AhnLab counted about 110,000 suspected vulnerable PHP hosts in Korea.
The top comment (44 likes) said it feels like the reason for personal data to exist is disappearing. Others said some servers surely still haven't patched (23 likes), and that many cannot upgrade because newer PHP versions break functions their sites rely on (11).
- ▶17분 만에 해골문양 등장‥PHP 서버, AI 모의해킹에 속수무책으로 뚫려 (2026.11.07/뉴스데스크/MBC)MBCNEWS · 61K views
Further reading
- 17분 만에 해골문양 등장‥PHP 서버, AI 모의해킹에 속수무책으로 뚫려MBC 뉴스 ↗
- PHP 소프트웨어 '치명적 결함' AI로 발견‥"즉시 업데이트해야"MBC 뉴스 ↗

Feature · Jobs

Rehiring the people AI replaced: how many companies reversed layoffs, and why surveys disagree

"The AI layoff gamble is backfiring. Companies that rushed to replace workers with AI are now cleaning up the mess," a post on X said on Nov. 7, drawing about 4,800 likes. Citing The Guardian, it said Meta scaled back plans to cut staff and use more AI agents after employees spent growing time cleaning up the agents' technical and security problems. We checked why the reported 'reversal' rate varies by survey.

The viral post: Meta scaled back layoff plans, per the Guardian
Unusual Whales posted on Nov. 7, citing the Guardian, that Meta scaled back plans to cut staff and use more AI agents as employees spent more time cleaning up agents' technical and security issues. We could not read the Guardian original; the post passed 450,000 views.
Survey 1, 600 HR leaders: two in three are bringing staff back
Per IBTimes, a February Careerminds survey of 600 HR leaders who oversaw layoffs found two in three companies making AI-driven cuts were already bringing staff back, and more than a third rehired over half the roles. 52% rehired within six months.
Survey 2, cost: 30.9% spent more rehiring than automation saved
In the same survey, 30.9% spent more on rehiring than they saved and 42.4% roughly broke even; about a quarter came out ahead. Orgvue estimates each dollar clawed back through workforce cuts costs $1.27 including severance, lost productivity and recruiting.
Survey 3, hiring managers: Robert Half's 32%, finance at 44%
Robert Half surveyed nearly 2,000 US hiring managers: 32% said they cut a role for AI and refilled the same position. By function: finance 44%, HR 35%, technology 32%.
Survey 4, workers: 19.1% of US respondents asked to return
An SHL survey run by Propeller Insights in June-July found 19.1% of US respondents were replaced by AI and later asked to return; 16.0% across the US and UK combined.
Cases: Commonwealth Bank, Klarna and IBM differ
Commonwealth Bank of Australia cut 45 customer-service roles in July 2025 citing an AI voice bot, reversed within weeks, and admitted it "did not adequately consider all relevant business considerations." Klarna shrank from about 5,500 to 3,400 staff, then CEO Sebastian Siemiatkowski said all-AI service was "lower quality" and rehiring began. IBM said its HR assistant automates about 94% of routine HR tasks but replaced only "a couple hundred" roles, and total employment rose.
Caveat: forecasts are not outcomes
Forrester's 55% regret figure and prediction that half of AI-attributed layoffs reverse by year-end, and Gartner's call that half of customer-service cutters rehire by 2027, are forecasts. IBTimes notes rehired jobs may return offshore or at lower pay. Because the surveys ask HR leaders, hiring managers or workers, the numbers range from about 19% to 66%.

AI & Markets

5 stories

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

InvestmentInvestors push Anthropic IPO valuation to $1.5-1.6 trillion vs. ~$2 trillion floated; quarterly revenue $17 billion
Institutional investors are pressing for an Anthropic IPO valuation of $1.5 trillion to $1.6 trillion, the Wall Street Journal reported.
Per investingLive's summary of the WSJ report, that is below the roughly $2 trillion discussed in recent months. Reasons cited: last week's tech selloff after weak Intel and Alphabet earnings, volatile markets, and political opposition to new data centers. Anthropic has not set a target valuation or share count and is expected to propose a price range. The roadshow could begin as soon as next week, with a debut targeted for the week before Thanksgiving.
Revenue for the quarter ended Sept. 30 was $17 billion, up from about $11.6 billion. About $600 million in price cuts from new model versions and contract discounts partly offset the growth. Anthropic disclosed about $895 billion in compute purchase and lease commitments, nearly $450 billion of it due by the end of 2030. A Cointelegraph post on X, citing CNBC, said some investors expect the valuation could fall to $1 trillion (about 280 likes).
Anthropic's compute commitments tie directly to long-term orders for cloud and chip suppliers, and the listing price bears on the valuation of backers such as Amazon and Google.
Further reading
SemiconductorsTSMC weighs 3D DRAM at Nvidia's request… possible new rival for Samsung Electronics and SK hynix
Taiwan's TSMC is weighing production of next-generation 3D DRAM at Nvidia's request, Seoul Economic Daily reported on Nov. 7.
Per the paper, The Information, citing two people familiar, reported that TSMC is considering building a new memory plant or converting an acquired one. Investment, scale and timing are undecided. 3D DRAM stacks memory cells vertically, a different structure from HBM, which stacks DRAM chips.
Today TSMC makes Nvidia's GPUs while SK hynix and Samsung supply the HBM. If TSMC makes memory too, it could affect Korean suppliers' bargaining power. But heat and process-integration hurdles mean it could take years to reach stable yields, so Korean positions are unlikely to shift soon.
Read as Nvidia diversifying supply amid an AI memory shortage, the move bears on how memory, foundry and packaging businesses divide roles.
Further reading
- Seoul Economic Daily: TSMC weighs entering advanced memory at Nvidia's request; possible rival for Samsung, SK hynix서울경제 ↗
Timeline signal'US approved Nvidia chip sales to Chinese firms with empty offices': X post draws 4,500 likes, no official confirmation
A post saying the US reportedly approved Nvidia chip sales to Chinese firms with empty offices and suspicious ties spread on X on Nov. 7. Watcher.Guru's post drew about 4,500 likes and 900,000 views; prediction market Kalshi posted the same claim (about 460 likes).
Both said only "reportedly" without naming an outlet or agency. It is not confirmed by an article or a Commerce Department statement.
The confirmed background dates to May. Per Reuters and others, Commerce approved H200 chip sales to about 10 Chinese companies including Alibaba, Tencent, ByteDance and JD.com, capped at 75,000 chips per company. At that point no deliveries had been made, and Chinese firms had held back after guidance from Beijing.
If true, it bears on Nvidia's China revenue outlook and on how loosely export licences are screened. Until official confirmation, treat it as unverified.
Further reading
Korea marketForeigners net-sold 7.6 trillion won of Samsung Electronics and SK hynix last week: shares fall 16% and 20% on AI-regulation worries
Foreign investors net-sold about 7.6 trillion won ($5.5 billion) of Samsung Electronics and SK hynix combined last week (Nov. 2-6), according to exchange data.
Per Ajunews and others citing Korea Exchange and Koscom figures, foreigners sold about 10.4 trillion won on the KOSPI over five sessions, led by SK hynix (5.4 trillion won) and Samsung Electronics (2.2 trillion won). Samsung closed at 188,700 won on Nov. 6, down 16.13% from the previous week; SK hynix fell 20.17%, from 1,418,000 to 1,132,000 won. Foreign ownership of Samsung dropped to 45.95%, the lowest since July 2009, about 17 years and 4 months.
Domestic investors absorbed the selling: individuals net-bought 2.2 trillion won of SK hynix and 1.9 trillion won of Samsung, and institutions bought 1.4 trillion won of SK hynix. Worries that AI data-center regulation could tighten after the U.S. midterms were cited behind the selling. Mirae Asset's Seo Sang-young expects volatility to continue.
Both companies supply high-bandwidth memory (HBM) for AI servers. With Democrats floating data-center taxes and construction delays, slower data-center spending was named as a burden on the sector. This is not investment advice.
Further reading
Timeline signalSpaceX-Anthropic compute deal has a 90-day exit: Chamath says Musk ending it would be "the king-making move," clip draws 660 likes
Investor Chamath Palihapitiya said the compute deal between SpaceX and Anthropic has a mutual 90-day out, in a clip that spread on X on Nov. 7. The post from Fireside Alpha drew about 660 likes and 170,000 views.
On a podcast, Chamath said, "If SpaceX says, I'm going to terminate this compute in 90 days, that is the king-making move," because it is a disproportionate amount of the compute Anthropic needs. He argued SpaceX could reprice the compute at the spot rate, probably double the market rate and possibly triple.
Contract figures vary by report. One report cites up to $120.5 billion through 2029 per Anthropic's IPO documents, while early IPO coverage put it at $40-45 billion. Some reports say the 90-day notice applies after non-cancellable periods end in January and March 2027.
SpaceX has not said it will terminate; Chamath was describing a scenario. Replies ranged from "I don't believe Elon would do that just for money" to "that will add to the nobody-wants-to-do-business-with-Elon lack of trust."
Further reading

All AI & Markets news →

Other AI news

2 stories

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

KoreaWith the CSAT near, 'AI glasses' alarm: ministry pushes criminal complaints and 3-year exam bans for smart-glasses cheating
Korea's Education Ministry is pushing to file criminal complaints against students who cheat on the CSAT college entrance exam with AI smart glasses and to extend the ban on retaking from one year to three.
Per i-News24 and others, the ministry said merely wearing smart glasses voids that exam and suspends eligibility the next year; refusing a proctor's check also counts as serious cheating. Each test site will get one staffer dedicated to spotting AI-glasses cheating, covering 1,312 sites nationwide.
The measures followed repeated catches of students wearing AI glasses at Yonsei and University of Seoul essay exams. The ministry sent universities a notice urging tighter supervision and a manual for proctors. One report said 82% of teachers surveyed find smart glasses hard to identify under current procedures.
Some measures are still "being pursued" or "under review," so details and timing differ across reports.
Further reading
- i-News24: Wearing 'smart glasses' voids CSAT; eligibility suspended next year too아이뉴스24 ↗
- Seoul Shinmun: AI-glasses cheating in college admissions becomes real; proctor manual sent on the 15th서울신문 ↗
USDemocrats, buoyed by midterm gains, prepare a flood of AI bills… data center taxes and a construction pause floated
Democrats, who gained in the US midterms, are preparing to push a wave of AI and data center bills in Congress, local outlets reported on Nov. 7.
According to Bloomberg Tax, Democrats have already introduced roughly half a dozen bills in both chambers to trim data center tax breaks or tax AI use and excess profits. Some senators confirmed to Bloomberg Tax that AI taxation would be a priority if Democrats retake the Senate. Sen. Bernie Sanders and Rep. Alexandria Ocasio-Cortez proposed pausing new AI data center construction until Congress passes AI safety, wealth-sharing and environmental laws. New Jersey Democrats said that if they win the House they would require data centers to supply their own clean power and call AI executives to testify.
Republicans control the current Congress, so the bills are unlikely to pass soon and would need to be reintroduced in a new session. A tally says 56% of voters called AI and data centers a major factor in an NBC exit poll, but we could not verify the original report.
Further reading
- GV Wire: Democrats planning flood of AI bills when they take control of CongressGV Wire ↗
- Bloomberg Tax: Democrats Float AI, Data Center Tax Plans as Tool for MidtermsBloomberg Tax ↗

View all 12 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

Jobs"I asked 100 devs what AI is doing to their jobs": HealthyGamerGG video hits 42,000 views; top comment asks whether a 2024 report can be trusted

Dev toolsAnthropic dropped Git, OpenAI made Git faster: Better Stack video at 54,000 views, daily code changes 2,500 to 30,000

SecurityClaude Code's "Don't Ask Again" left passwords on GitHub: SavvyNik video at 44,000 views, secrets in 1.1% of settings files

SecurityAI broke into a PHP server in 17 minutes: MBC News video at 60,000 views, about 110,000 Korean hosts suspected vulnerable

Rehiring the people AI replaced: how many companies reversed layoffs, and why surveys disagree

The viral post: Meta scaled back layoff plans, per the Guardian

Survey 1, 600 HR leaders: two in three are bringing staff back

Survey 2, cost: 30.9% spent more rehiring than automation saved

Survey 3, hiring managers: Robert Half's 32%, finance at 44%

Survey 4, workers: 19.1% of US respondents asked to return

Cases: Commonwealth Bank, Klarna and IBM differ

Caveat: forecasts are not outcomes

AI & Markets

InvestmentInvestors push Anthropic IPO valuation to $1.5-1.6 trillion vs. ~$2 trillion floated; quarterly revenue $17 billion

SemiconductorsTSMC weighs 3D DRAM at Nvidia's request… possible new rival for Samsung Electronics and SK hynix

Timeline signal'US approved Nvidia chip sales to Chinese firms with empty offices': X post draws 4,500 likes, no official confirmation

Korea marketForeigners net-sold 7.6 trillion won of Samsung Electronics and SK hynix last week: shares fall 16% and 20% on AI-regulation worries

Timeline signalSpaceX-Anthropic compute deal has a 90-day exit: Chamath says Musk ending it would be "the king-making move," clip draws 660 likes

Other AI news

KoreaWith the CSAT near, 'AI glasses' alarm: ministry pushes criminal complaints and 3-year exam bans for smart-glasses cheating

USDemocrats, buoyed by midterm gains, prepare a flood of AI bills… data center taxes and a construction pause floated