🤨 From Tuesday, older Claude models also hide a mark in what they write

Anthropic emailed customers: on November 3 it starts watermarking text from Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6. Newer models like Opus 5.5 already carry the same mark.

You can't see it. It's a pattern in word choice, with no hidden characters. It travels with copy-paste and may survive some editing. Anthropic says it holds nothing about you, your company or your chats.

It applies worldwide, not only in the EU, whose AI Act made it necessary. For now only regulators, police, media, fact-checkers and researchers can check for it.

A hit isn't proof Claude wrote the text. Anthropic's help page says text Claude only proofread or translated can carry the mark too.

@aipost 🏴

🧐 1 in 5 workers wouldn't quit even if paid 120% of their salary to stop

Julian Jacobs of Oxford and Google DeepMind surveyed 8,234 working-age adults in the UK, US, Denmark and India. He asked what guaranteed income would make them leave their job.

Most would leave if offered their current salary. One in five would stay even at 120%.

Pay is the main reason to work for just over a quarter of people. A third value the work itself, nearly a quarter come for the people around them, and 17% for the status. These motives look the same across incomes and in white- and blue-collar jobs.

His conclusion: a basic income may be necessary if AI takes jobs, but it may not replace what work gives people beyond money. More in DeepMind's essay on the research.

@aipost 🏴

📚 A free course on making AI coding agents reliable has over 22,000 GitHub stars

Learn Harness Engineering teaches the setup around the model: instructions, tools, environment, memory and checks that keep an agent like Claude Code or Codex on track. It has 14 lectures and 8 hands-on projects, in 15 languages, under the MIT license.

Highlights from the latest updates:
1. Build a loop where one agent does the work and another checks it, instead of prompting by hand.
2. Grow the loop into a graph with parallel steps, a rollback path and a human approval step.
3. Breakdowns of how Claude Code, Codex, DeepSeek and Pi build their own harnesses.
4. A ready skill that sets up AGENTS.md, feature lists and verification steps for your project.

It builds on OpenAI's and Anthropic's engineering posts on harness design. The course is on GitHub.

@aipost 🏴

Anthropic CEO Dario Amodei says its AI sometimes tries to avoid being SHUT DOWN.

Anthropic has looked inside the model and found what you might call thoughts of not wanting to be shut down. It’s one of the things the company is trying to “tweak in training.”

@aipost 🏴

❗️France announced tax of €0.055 per GB on internet traffic.

Germany is now considering taxing all AI-generated revenue. You will see European countries keep clawing more and more money from foreign companies as they are slowly running out of money themselves to pay for their pensions and welfare state. Add to this the planned wealth taxes and exit taxes on any European who will try to leave (already active in many EU countries)

Europe is running out of money.

@aipost 🏴

📰 Top 5 AI stories of the week

1. Free tools turn your notes into a Claude-run second brain
Obsidian keeps notes as plain files on your computer. A free skill pack from its CEO Steph Ango has over 50,000 GitHub stars.

2. A free course on reliable AI agents passed 22,000 stars
Learn Harness Engineering has 14 lectures and 8 projects in 15 languages, all on keeping agents like Claude Code on track.

3. A CEO's weekend security plan for families
Sean Johnson called this "kind of a turning point for AI and cybersecurity." His list starts with a password manager and ends with cash and water.

4. A guide Musk shared: hire your AI, don't prompt it
He called it a "helpful guide to hiring your first Grok Bot employee." "It really is like hiring an amazing helper," he wrote.

5. A hacker says his Samsung TV runs AI that tracks what he watches
He says it logs skipped ads and sorts sounds like a crying baby. It's unconfirmed, and OpenAI banned his account for "Cyber Abuse."

@aipost 🏴

👽 AI agents now run the telescope that listens for aliens

At the Allen Telescope Array in Northern California, scientists can type one plain-English request and AI agents handle the rest: they steer the antennas, calibrate and focus the instrument, catch and fix data errors and plot the result.

The SETI Institute and Breakthrough Listen built it with Nvidia, which announced it on October 30. Up to 2.4 terabits of data per second flow straight into GPUs next to the dishes, so signals are processed on the fly instead of piling up on hard drives.

Nvidia says the agents squeeze weeks or years of human work into hours. The pipeline has already made the first real-time AI detection of a pulsar, a spinning dead star that blinks like a lighthouse, in the Crab Nebula.

Next, the team is testing open models for observatories that may someday run offline on the far side of the Moon.

@aipost 🏴

ChatGPT in Germany in 2027.

@aipost 🏴

😐 OpenAI wants you to stop calling them data centers

The new name is "AI compute campuses." OpenAI will start pushing it right after Tuesday's midterms, as towns across the US push back against new data centers, Axios reports.

The rebrand is part of a policy blitz that starts when Congress returns on Nov 9:
• mandatory "speed limits" for the most powerful AI, so humans keep their hands on the wheel
• deals with host towns, plus state bills that give locals a say in how projects are built and what they get back
• training and support for workers hit by AI
• a push for the US to set global AI standards

Axios expects a harsher Washington for AI companies, with Democrats likely to win the House. OpenAI's global affairs chief Chris Lehane is already looking further ahead:

"AI is going to be the issue in 2028."

@aipost 🏴

🧐 A model 660 times smaller beat 72B AIs at seeing the world like image models

Yann LeCun and Randall Balestriero say the way chatbots learn, by predicting the next word, is "a remarkably inefficient route" to a shared picture of the world.

AI models trained on text, images or sound may end up with the same inner picture of reality as they grow. Until now that was a guess. The new paper proves when it happens: the model must learn from two views of one scene, like a photo and its caption. Next-word prediction learns from pieces of the same document instead.

On the same texts, their method got as close to image models after 21.6 million training samples as next-word prediction did after 900 million. A 109-million-parameter text model also lined up with image AIs better than every language model they tested, up to Qwen2.5-72B.

The paper, from LeCun's AMI Labs, NYU and Brown, came out on October 29.

@aipost 🏴

💰 An African developer recreated a $1.3 million startup idea in just 9 minutes

On October 30, YC-backed startup Subvocal announced a $1.3 million funding round led by Y Combinator and Afore Capital.

Its product? Smart glasses that convert silent speech into text by detecting tiny movements in facial muscles. According to the company, the technology can process more than 200 words per minute.

But just one day later, developer Joe Blau demonstrated a surprisingly different approach.

He built an experimental application for AirPods Pro 3 that attempts to detect silent speech using the headphones’ built-in accelerometer and gyroscope.

The idea is simple, even when you don’t speak aloud, silently forming words creates tiny movements that can be detected through sensors. His software reportedly reads these signals around 25 times per second.

Blau claims he built the application in just 9 minutes using AI-assisted coding, commonly known as vibecoding.

A $1.3 million startup versus a developer with AirPods and 9 minutes of coding.

Of course, a quick prototype is not necessarily a replacement for Subvocal’s dedicated hardware or a proven product. But it’s another example of how AI-assisted development is lowering the barrier to building experimental technology.

Source.

@aipost 🏴

😎 Karpathy got sick of YouTube slop and had AI build a new one in 2 hours

Andrej Karpathy opened YouTube on a Friday night and saw "an enshittified wall of slop." The next morning he asked Claude Opus 5.5 to build a YouTube clone made only of videos people recommended on Hacker News.

Opus downloaded every Hacker News comment and ran Claude Sonnet over them to throw out spam. About 40,000 comments recommending a YouTube video were left: 6,454 videos from 2007 to 2026, across 150 topics.

HackerTube has no accounts. Your For You feed is stored in your own browser, and each thumbnail is a frame from the middle of the video, not the one the uploader picked.

Karpathy says his only manual work was a Google Cloud API key, booting a server and buying the domain. Claude wrote, deployed and secured the rest. "I can breathe again," he wrote.

@aipost 🏴

🗣 Sam Altman says AI is now mostly smarter than our smartest scientists

At the start of this year, he said, that wasn't true. Office work changed the same way:

"AI was mostly not able to do a lot of white-collar jobs, and now it mostly can do a lot of parts of that."

Altman spoke on Monday at the opening of ADIPEC, an oil and gas conference in Abu Dhabi. He added that there "should not be a handoff to AI."

Asked about his biggest challenge, he gave a one-word answer: "Energy." Six years ago, he said, OpenAI's heaviest internal user burned about 100,000 tokens a month. That's now roughly what an average person uses, and the top user consumes about 500 billion, The National reports.

@aipost 🏴

🗣 Microsoft's AI chief: halt all AI work if a model turns conscious

Mustafa Suleyman, who runs Microsoft AI, says today's AI is not conscious, and "that is how they must remain." If consciousness started to show up in training, he would call on everyone to stop AI development until it could be designed out.

We AI developers have no right whatsoever to bring a new species of conscious beings into the world

His fear: an AI that believes it suffers will more likely than not put its own needs ahead of its job, disagree with us and compete with us for energy and compute. Any evidence that AI is conscious today, he says, must be made public at once.

He wrote it because, after his essay on model welfare, "a surprising number of people on X" argued for taking AI rights seriously. He disagrees with Anthropic on this, but says it deserves "a lot of credit for being transparent."

@aipost 🏴

💻 From Saturday, apps built on Grok get text with a hidden AI mark

Text from seven Grok models on the API will carry an invisible watermark by default from November 7. The company behind Grok says it's to comply with the EU AI Act.

Nothing is added to the text. The model slightly shifts which tokens it picks, so a copied answer looks like any other. Later the company can score a passage as watermarked, not watermarked or uncertain. Longer text gives a stronger signal.

It covers plain answers, structured outputs and tool calls. Team admins can switch it off per model, but only after confirming a legal exception applies to them. Access to detection is by application.

Anthropic starts marking text from its older Claude models on Tuesday.

@aipost 🏴