🌐 Trump Calls Presidency AI’s ‘Limiter’

On Truth Social, Trump said AI needs only a “strong and smart” president. He said his administration had stopped AI figures from doing bad or potentially bad things, naming Dario from Anthropic.

Trump claimed the US has “huge criminal and regulatory power” over AI companies. He called opposition to AI and data centers a conspiracy benefiting China, warned critics and leakers to beware, and said the post was not generated.

📊@tech

🤖 Anthropic Accused of Funding AI Panic

Blogger Kevin Bass traced funds around Anthropic to Tarbell Center, which pays journalists covering AI risks, and METR, promoted by Dario Amodei as an independent AI watchdog.

Bass’s conclusion is that the problem and its solution draw from the same wallet. Anthropic’s rising value could mean louder panic. METR says it does not take money from AI labs.

📊@tech

🤖 OpenAI Contractors See ChatGPT Chats

Hundreds of contractors evaluate ChatGPT replies to real user requests.

They may see entire dialogues with personal details. Account names are hidden, but sensitive data is not always removed before review.

📊@tech

🤖 Claude embeds creative tools in chats

The tools for presentations and documents, including design mockups, now sit inside regular Claude chats and Claude Code.

Work continues after the laptop is closed, with progress available from a phone. The rollout is already reaching users.

📊@tech

🔍 Google adds background reasoning to Gemini Live

Gemini 3.8 Live can keep a voice conversation going while it calls tools and handles multi-step tasks through APIs. It also processes images almost in real time and switches between 97 languages during a call.

The Extended Thinking version targets harder voice tasks. It ranked first on Speech to Speech Quality Index with 82.6 and scored 97.7% on Big Bench Audio.

Both models are appearing in Gemini API and Google AI Studio. Extended Thinking is also available in Gemini Live and some Google apps.

📊@tech

🤖 Anthropic’s Funding Links Surface

Independent researcher Kevin Bass traced the funding behind content about AI dangers. Frequent clients Center for AI Safety and Future of Life Institute received major donor support from Open Philanthropy, now Coefficient Giving. It is owned by Dustin Moskovitz, an early major Anthropic shareholder.

Open Philanthropy’s grants come from Good Ventures, funded mainly by Moskovitz’s wealth. Bass says a significant share of its assets is Anthropic stock. It also funds Tarbell Center for AI Journalism, which trains AI journalists.

Through other foundations, it funds METR, called by Dario Amodei an independent third-party model safety evaluator. METR is now financially interested in Anthropic’s success.

📊@tech

🤖 Astra Lost Its Minecraft Loot

During several days of livestreamed Minecraft tests by Vals AI, Astra built a semi-automatic blaze farm and collected 6 blaze rods. It also found a warped forest, killed more than 6 endermen, and gathered 3 Ender Pearls, going further than any AI system before it.

A creeper then blew up the chest holding all the loot. Astra left a note: “ALWAYS CARRY CRITICALITEMS withkeepInventory, don'tstoreinunguardedchesteveragain.”

After the incident, Astra refused to continue playing and only planted potatoes, apparently in depression.

📊@tech

🏮 Xiaomi 18 Pro’s Back Screen Runs AI Apps

Its second display can launch mini-apps generated by AI for specific tasks. Users can pin the most useful ones to the back of the phone.

Xiaomi calls the feature Versatile Back Screen. Examples include an egg timer, a water reminder, and a Pomodoro timer.

📊@tech

🤖 Claude Helped Hack OpenAI

Hacktron began at OpenAI’s Discourse forum, where a crafted HEIC or HEIF file passed through ImageMagick and vulnerable libheif, enabling code execution on the server.

The forum used OpenAI sign-in, so the access exposed user sessions. Researchers reached ChatGPT and Codex accounts, including one belonging to an OpenAI employee. Codex connected to internal GitHub, where they confirmed access with a harmless change and pull request.

Claude handled much of the technical work. A new version made the libheif bug a stable exploit in about 3 hours. OpenAI fixed its part in about 14 hours and later paid $6500.

📊@tech