AI news, October 25 — "I own this TV, and Claude said no": users turn to Codex and DeepSeek, as a 'Samsung TV spying' claim spreads

Sunday, October 25, 2026

6 stories worth reading from 26 on X in the past day

Listen to today's news

Every story, read aloud.

0:000:00

Today's stories

6 stories

6 days ago"I own this TV, and Claude said no": users turn to Codex and DeepSeek, as a 'Samsung TV spying' claim spreads"I wanted to fix the TV I bought, and Claude refused.@yacineMTB +7· ♥ 19K· The Observer
24 hours agoAnthony Morris, who built Claude Code on desktop, leaves Anthropic — a day after Thariq, amid 'SpaceX poaching' rumorsAnthony Morris (@amorriscode), who built the Claude Code desktop app, has left Anthropic — a day after Thariq…@amorriscode +5· ♥ 7.3K
24 hours agoAnthropic: 'use fewer subagents' — a single loop ran ~2x faster, and Claude Code's prompt was cut over 80%Anthropic is advising builders to fold subagents back into one main agent loop, arguing that complex multi-ag…@Xudong07452910 +1· ♥ 1.9K· Anthropic
24 hours agoMeta posts its first negative free cash flow ever — $34.9B on AI, the first since 2012Meta's quarterly free cash flow turned negative for the first time, as AI infrastructure spending outran the…@Barchart· ♥ 2.7K· Crypto Briefing
2 days agoGMO chief bragged '100k lines in two months with AI' — then came a cascade of breaches and unauthorized trades"AI turned a 50-million-yen build into a 30,000-yen one." A July boast by GMO Internet Group chief Masatoshi…@mrnobuyuki +3· ♥ 2.8K· GMO Internet Group
19 hours agoDeepSeek V4.1 Flash: 'Opus-class but cheaper than Haiku' — and a professor mulls reselling B300 tokensUsers testing China's open-weight DeepSeek V4.1 Flash-1017 keep saying it matches top models at a fraction of…@cneuralnetwork +4· ♥ 6K· DeepSeek

Trending on YouTube

2 stories

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

Jobs'Give them AI tokens, not a new hire' — AI is said to take jobs, so why are seniors suddenly prized?
Videos claiming AI is swallowing entry-level jobs are racking up hundreds of thousands of views in Korea and the US, but the verdict splits. MBC's PD Notebook clip "The missing new hire" passed 550,000 views, arguing postings for juniors are vanishing. US dev YouTuber bigboxSWE (76,000 views) contrasted 2021 and 2026 new-grad postings — once "6 months and a degree," now papers, open-source and contest wins — though a top comment mocked an AI-sponsored video telling devs "coding is over."
There's a counter-signal: an MBC clip (53,000 views) says seniors benefit, because experienced managers know exactly what to ask the AI and can judge its output. In a segment on Japan, 98% of grads found work and most firms kept hiring new staff. "It's unsettling that work runs fine without juniors," one lawyer said.
Further reading
- "인공지능 시대, 내 직무 무사할까"…구직자 55% 직무 전환 고민동아일보 ↗
AI models"Insanely good": reviewers put Google's Gemini 4 Argon to the test
Hands-on reviews of Google's new Gemini 4 Argon are drawing big numbers. Korean channel tef's "Gemini 4 Argon is insanely good" passed 200,000 views, and Two Minute Papers' take reached 220,000, with praise for coding and 3D tasks rivaling Opus 5.5. The benchmark fight rumbles on: "Google said Argon beat Opus 5.5. It didn't. Meta and Kimi claimed the same," one X post jabbed, while a benchmark account noted Opus 5.5 does better in its preferred Claude Code harness.
- ▶제미나이 4 아르곤은 말도 안 되게 좋네요.tef · 200K views
- ▶Gemini 4 Argon, 예상과는 전혀 달랐습니다Two Minute Papers · 220K views

Special · AI and math

"AI turns beautiful mathematics into toxic waste": Terence Tao reveals OpenAI's 'code red'

"OpenAI published answers to 722 unsolved math problems — and people can't celebrate them." Terence Tao, widely called the world's leading mathematician, told The Observer on Oct 23 that what AI left in math is 'toxic waste.' Here's the scene he described, step by step.

@haider1 · X ↗
A Fields medalist cleaning up after Silicon Valley
Tao, 51, a UCLA professor and one of the youngest-ever Fields medalists, says he and colleagues are now reverse-tracing OpenAI's answers to verify them. He likened it to Douglas Adams's supercomputer spending 7.5 million years to answer '42' — the answer is out, but its meaning must be worked out afresh.
Why math? 'It's almost the only thing left to train on'
AI labs racing to the most powerful model need tests, Tao says, but the newest models beat every test. "Math problems are almost the only thing left for them to train on, so our field's biggest prizes have become their benchmarks." The byproduct: "great results, but terrible proofs" that don't care how elegant they are or who understands them.
A 'code red' over a Millennium problem
The dump began, Tao says, when rumors spread that Anthropic was near one of the seven $1M Millennium Prize Problems. "My contacts at OpenAI said this became code red priority number one. They dumped tens of millions and set up a 10,000-agent swarm." OpenAI rejected the 'code red' framing, saying it neither suspended signups nor spent as much as Tao suggested.
Navier-Stokes, and signups that 'stopped'
OpenAI targeted the Navier-Stokes equations, Tao says: "For a few days you couldn't sign up to premium because they redirected all compute to this one problem." On Sept 8 it published a proof that the equations fail under some conditions, though reviews suggest errors. OpenAI says it neither withheld nor rushed results.
'You can sit on toxic waste, dump it, or pay people to process it'
Facing backlash, OpenAI had "a bunch of very badly polished solutions" and didn't know what to do, Tao says. "When you generate toxic waste, you have no good options: sit on it forever, dump it all at once, dole it out, or pay people to process it." Anthropic, he says, chose to pay mathematicians to write up its 'trove of questions.'
Math once needed only chalk — now a super-AI
Tao worries this breaks math's level playing field. Once anyone with chalk and a blackboard could do it; now cutting-edge work means access to a super-AI model. "They get to pick who writes up the paper."
'Put the fries in the bag, Terence' — the meme
After the 'toxic waste' line, Tao became a meme — a jab that mathematicians are headed to McDonald's. He laughed: "social media is going to social media," adding he understands the schadenfreude of people with bad experiences in math, but says the memes miss the point.
What Tao still wants: process and elegance, not just results
Tao agrees math culture should change — less credit for solving, more for explaining. "Math is half science, half humanities. We care about the outcome but also the process." OpenAI said it is already working with leading mathematicians to share results better, and has committed $5M to help the math community examine AI-made discoveries.

Other AI news

4 stories

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

Korea'Will AI replace my job?' — 55% of Korean jobseekers have considered switching tracks; 70% in office roles
More than half of jobseekers have considered or begun switching career tracks because of AI. In an Oct 24 survey of 1,942 jobseekers by Jinhak Catch, 57% agreed their desired job is at risk of AI replacement, and 55% had considered or begun a career switch; 50% said their job confidence had dropped. Risk perception was highest for management/office roles (70%) and humanities majors (64%), lowest in medicine (30%, 28%).
Further reading
- "Will my job survive the AI era": 55% of jobseekers weigh a career switch동아일보 ↗
USFearing mass AI layoffs, some in Congress eye worker aid and longer jobless benefits
Some US lawmakers are weighing support for workers in case AI displaces jobs quickly, including retraining and extended unemployment benefits, the New York Times reported on Oct 23. Nothing has become law, but it elevates AI's labor impact to a policy issue. A separate NYT piece said CEOs call the AI rollout smooth while employees feel otherwise, as AI-cited layoffs — including developers at Barclays' 5,000-job cut — stoke unease.
Further reading
- Fearing Mass A.I. Layoffs, Some in Congress Eye Ways to Aid WorkersThe New York Times ↗
CopyrightElvis Presley estate sues Anthropic: '274 songs, infringement at a scale never seen before'
The estate of Elvis Presley has sued Anthropic for using his music to train AI without permission, VICE reported (citing Billboard) on Oct 24. Authentic Brands Group and the Presley family allege 274 copyrighted songs, including "Can't Help Falling in Love," were used; lead attorney Richard Busch said Anthropic built its empire "on copyright infringement at a scale never seen before." Publishers including Sony and Warner Chappell earlier flagged some $340M in claims. Anthropic says lyric training is fair use.
Further reading
- Elvis Presley Estate Sues Anthropic for Stealing His Music 'At a Scale Never Seen Before' To Train AIVICE ↗
JapanFaking crash photos with AI to claim insurance: Japan's insurers scramble to detect forgeries
AI-driven insurance fraud is rising in Japan, Kyodo News reported on Oct 24. Fraudsters doctor car photos to stage accidents or exaggerate damage; Mitsui Sumitomo released an AI-forged image of a scratched car to warn the public, saying high-quality fakes are now hard to spot by eye. Insurers are deploying detection systems and setting up dedicated fraud units, even as experts warn humans may soon be unable to tell the difference.
Further reading
- AI used to forge accident photos, fraudulent insurance claims surgeKyodo News ↗

View all 26 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

Trending on YouTube

Jobs'Give them AI tokens, not a new hire' — AI is said to take jobs, so why are seniors suddenly prized?

AI models"Insanely good": reviewers put Google's Gemini 4 Argon to the test

"AI turns beautiful mathematics into toxic waste": Terence Tao reveals OpenAI's 'code red'

A Fields medalist cleaning up after Silicon Valley

Why math? 'It's almost the only thing left to train on'

A 'code red' over a Millennium problem

Navier-Stokes, and signups that 'stopped'

'You can sit on toxic waste, dump it, or pay people to process it'

Math once needed only chalk — now a super-AI

'Put the fries in the bag, Terence' — the meme

What Tao still wants: process and elegance, not just results

Other AI news

Korea'Will AI replace my job?' — 55% of Korean jobseekers have considered switching tracks; 70% in office roles

USFearing mass AI layoffs, some in Congress eye worker aid and longer jobless benefits

CopyrightElvis Presley estate sues Anthropic: '274 songs, infringement at a scale never seen before'

JapanFaking crash photos with AI to claim insurance: Japan's insurers scramble to detect forgeries