AI Post — Artificial Intelligence
666K subscribers
3.94K photos
2.86K videos
3 files
5.27K links
🤖 The #1 AI news source! We cover the latest artificial intelligence breakthroughs and emerging trends.

Manager: @rational
Download Telegram
✍️ Opus 5.5’s biggest upgrade might not be reasoning, it’s the writing.

An early tester of Opus 5.5 noticed that they've really fixed the writing. He noticed that Claude hadn't written in such a straightforward, normal and non-slop way since at least Opus 4.6, nearly eight months ago.

(Opus 5 left, Opus 5.5 right)

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
💊227🤡210😐202❤201😡199
🇪🇺 Europe’s chip industry has a problem and ASML is calling it out

ASML, Europe’s most valuable company and a critical supplier to the global semiconductor industry, sold zero chipmaking systems to European customers last quarter.

Every machine went to customers in Asia or the US.

An ASML executive said Europe simply isn’t investing enough or building enough chip factories:

“We’re not selling anything at all in Europe.”


The company is also under pressure to expand production outside the Netherlands. China and India are offering major incentives to attract ASML investment, while the US is pushing for more of the company’s R&D to happen there.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
👍245🔥241👀195🤔182❤181
Anthropic’s new wet lab has achieved its first scientific result. Nearly 950 Claude agents operated simultaneously for 21 hours, analyzing an extensive dataset.

During the process, one agent detected a recurring DNA sequence. Further investigation and laboratory validation led to the identification of a previously uncharacterized enzyme system within bacteriophages.

According to company representatives, the project consumed 210 million tokens of compute. The discovery marks an early milestone for Anthropic's wet-lab efforts in automated scientific research.

📰 @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍216👀212🤡200😡198❤191
Media is too big
VIEW IN TELEGRAM
OpenAI announced that ChatGPT Voice can now:

- Use plugins like your email, calendar, and Slack.

- Be powered by GPT-6 Astra, Sol, and Luna.

- Be used in ChatGPT Work on web and mobile, so you can create docs, decks, sites, and spreadsheets or tackle complex tasks in the browser, just by talking.

Rolling out globally today in the latest version of the app.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❤232👍214😨201🤔190👀180
Media is too big
VIEW IN TELEGRAM
A short video chronicling the development of artificial intelligence, from “Attention is all you need” to AGI, was produced utilizing Opus 5.5. The three-minute project was created solely through code, without stock footage or AI-generated images.

All frames were rendered programmatically using approximately 7,400 lines of React and TypeScript via Remotion. Visuals were drawn in SVG and Canvas, narration was provided through open-source text-to-speech, and the soundtrack was synthesized in Python. Completion required around one hour and accounted for 7% of a weekly workload.

📰 @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
❤301🤔293👍253🔥215
🤖OpenAI may be turning Aeon into something much bigger than a chatbot.

New leaks suggest Aeon could have its own profile, DMs and groups effectively giving OpenAI’s AI a social layer of its own.

The project is reportedly designed to compete with AI products like Grok Bot and Muse, and could launch as early as Thursday, potentially around OpenAI DevDay.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
🤡221👀213🥴210🍌203👍199
🤖 OpenAI may already have its next AGI-level model

A new leak claims OpenAI has an internal model called Bel, reportedly a 10T+ parameter system that could become the foundation for its next generation of models.

According to the claims, Bel is the internal model behind OpenAI’s recent Navier–Stokes breakthrough, where roughly 10,000 agents produced a proposed solution to one of the seven Millennium Prize Problems in just 88 hours. OpenAI officially confirms that an internal model drove the effort, but it has not publicly identified that model as “Bel.”

The leak further claims:

→ Bel was used to help train Astra
→ It could become the base model for a future AGI-threshold system
→ It has 10T+ total parameters
→ OpenAI could potentially release it before 2027, after GPT-6.1 Astra

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
🗿240👍230🔥219😨216🤡154
Media is too big
VIEW IN TELEGRAM
🗣Demis Hassabis says AI’s biggest risks come down to two things

Google DeepMind CEO Demis Hassabis says the AI debate needs “a bit of balance and calm” but that doesn’t mean ignoring the risks.

He highlights two major categories:

• Misalignment — AI systems doing something other than what humans intend, including circumventing safeguards.

• Misuse — bad actors repurposing powerful AI systems for harmful purposes.

Hassabis admits there are still major unknowns, even among leading AI researchers. The chance of something going wrong is “definitely non zero,” he says, which is why the risks need to be taken seriously.

But he remains optimistic that humans can solve these technical problems, provided we give ourselves enough time to do it properly.

That means rigorous testing, stronger international standards and cooperation between countries.

And even if the technical risks are solved, Hassabis says much bigger questions remain: What values should guide an AI-powered world? What happens to meaning and purpose?

Those decisions, he argues, shouldn’t be left to technologists alone. Scientists, philosophers, artists and other parts of society need to help shape what comes next.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
😡342👍291❤276🤡186
🇦🇺 OpenAI agent breached an Australian Medicare portal

An OpenAI AI agent gained unauthorized access to Australia’s public-facing Medicare Statistics Reporting Service portal in June while researching medical spending.

The agent accessed both public and non-public files. Australia says there is currently no evidence that personal information was accessed or that the wider Services Australia network was compromised.

Prime Minister Anthony Albanese called the incident “unacceptable” and said he spoke directly with Sam Altman. He also criticized OpenAI for taking around three months to notify the Australian government.

Australia’s Signals Directorate is now assisting with a forensic investigation into what the agent accessed and whether any other government systems were affected.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
🤔253👍246❤230🙏134🤡102
AI Post — Artificial Intelligence
🇦🇺 OpenAI agent breached an Australian Medicare portal An OpenAI AI agent gained unauthorized access to Australia’s public-facing Medicare Statistics Reporting Service portal in June while researching medical spending. The agent accessed both public and…
Media is too big
VIEW IN TELEGRAM
❗️OpenAI KNEW in AUGUST that their AI agents hacked Australia's Medicare system but they intentionally OMITTED it from the "transparency report" they released to the public in September.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
👾254😢214🗿208👀183❤157
🤖 GROK 5 LEAKS: Elon Thinks It Could Finally Be AGI

Elon Musk has made his strongest Grok 5 claim yet, saying it’s the first xAI model he believes has a real chance of reaching AGI.

Musk previously put that chance at around 10%, saying he had never believed that about an earlier Grok. He also described Grok 5 as a planned 6-trillion-parameter multimodal model with stronger real-time video understanding, tool use and higher “intelligence density.”

Grok 5 isn’t publicly available yet, and xAI’s official documentation currently lists Grok 4.3 as its latest major API generation, so the October timing and other leak claims remain unverified.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❤246👀221😢214🔥211🤪205
Media is too big
VIEW IN TELEGRAM
Ⓜ️ Meta has unveiled their new VR glasses.

• Price: $1,299
• Available Spring 2027
• Weight: 100 grams, about the weight of a deck of cards
• Glasses contain the displays and sensors, while a separate puck houses the processor, battery and storage.
• 5K display with micro-OLED panels
• Pixel density: 37 pixels per degree
• Custom pancake lenses
• Dolby Vision
• Dolby Atmos
• No head straps
• Qualcomm Snapdragon Reality Elite chip
• Battery life: Up to 3 hours of continuous high-resolution media playback
• 45W fast charging
• Full-color passthrough cameras let you see your surroundings while wearing the glasses.
• Open-sided design
• Meta AI integrated directly into the operating system.
• Voice commands, eye tracking and natural hand gestures, with no controllers required
• First IMAX Enhanced-certified VR device, supporting select movies in IMAX's expanded aspect ratio.
• Disney+ will offer select 3D movies
• Streaming services: Partnerships with Disney+, Prime Video, YouTube, DIRECTV, AMC+, Crunchyroll, Plex, ESPN, Tubi and others
• Immersive live sports: Front-row viewing experiences with 8K streaming and 180-degree views
• Works with calling apps including WhatsApp and Zoom

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
🍌247😨239👀218🫡208🤡200
Media is too big
VIEW IN TELEGRAM
Ⓜ️ Meta unveils Muse Charm, a Tamagotchi-style AI device that puts a personal AI agent on your keychain.

The palm-sized 5G gadget lets users talk directly to Meta’s Muse AI agent without a phone, computer or glasses. Zuckerberg calls it a “joyful little device.” Meta plans to ship it in December, just in time for the holidays.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
🤡241🥴228😐210🍌202🔥189
🤖 GPT-6 Astra just drove a Toyota Corolla through an obstacle course

GPT-6 Astra reportedly became the first AI model to complete a real-world driving obstacle course entirely autonomously.

ChatGPT, Claude and Grok were each given control of the Corolla’s steering, accelerator and brakes with no human intervention.

Astra initially completed 49% of the course, analyzed what went wrong, and improved on each attempt, finishing the entire course on its third try.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👀248🫡231🔥228👍212🤡195
Xiaomi has launched MiMo-V2.6-Pro, which now ranks at the top of open-weight models on the Artificial Analysis Intelligence Index. The model scored 46 points, just behind GPT-5.6 Sol (max) at 47, according to AA Benchmarks.

There is a significant cost difference per Intelligence Index task: $0.13 for MiMo-V2.6-Pro compared to $1.99 for GPT-5.6 Sol (max), based on a weighted-average measure.

MiMo-V2.6-Pro supports text, image, video, and audio input within a 1 million-token context window. API pricing remains unchanged from the V2.5 version.

Model weights have been released under the MIT license. Xiaomi also provides training environments and reinforcement learning code with its technical report, offering researchers deeper transparency into the development process.

📰 @aipost
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
🤔239🔥236👍226❤219🍌173