AI Post β€” Artificial Intelligence
849K subscribers
2.7K photos
2.2K videos
3 files
4.35K links
πŸ€– The #1 AI news source! We cover the latest artificial intelligence breakthroughs and emerging trends.

Manager: @rational
Download Telegram
"Cowork" was built at Anthropic in under two weeks. 100% of the code written by Claude Code πŸ‘€

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
1πŸ‘25❀10πŸ‘€9πŸ”₯6
πŸ“’ MIT Technology Review has released its list of 10 Breakthrough Technologies for the coming year, curated by the editorial team and it’s an intriguing mix:

1. Sodium-ion batteries
2. Code generation
3. Next-generation nuclear power
4. AI companions
5. Gene therapy for infants
6. Species revival via genetic engineering
7. AI model interpretability
8. Commercial space stations
9. Genetic scoring of embryos
10. Hyper-scale data centers

What stands out is the blend of the very real and the very aspirational. On one end, we have tangible, industrial-scale efforts already underway, new nuclear reactors, gigawatt-class data centers, battery chemistries moving toward commercialization. On the other, ideas that feel closer to marketing narratives than settled science. Many biologists, for instance, strongly dispute that engineering a β€œhairy elephant” meaningfully resurrects the woolly mammoth.

Still, the meta-signals are revealing.

β€’ AI dominates: 4 out of 10 entries are directly about artificial intelligence.
β€’ Genetic engineering follows with 3 slots.
β€’ Energy gets 2.
β€’ Space gets just 1.

That distribution says a lot about where editors and arguably capital and attention, believe momentum is concentrating.

Even more interesting is what’s missing: no climate adaptation tech beyond energy, no robotics as a category, no new materials (outside batteries), no breakthroughs in transportation or manufacturing. Whether that’s because progress there is incremental, unfashionable, or simply harder to hype is an open question.

And finally, there’s the tone. Longtime readers will notice the familiar, slightly dissatisfied undercurrent, a mix of excitement, skepticism, and quiet irritation with overpromised futures. That tension may be the most honest part of the list.

Sometimes the breakthroughs are less revealing than the editorial mood behind them.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
1❀31πŸ‘4πŸ‘€3πŸ€ͺ2
πŸ“ˆ The market is estimated at $5.8B in 2025, projected to hit $7.2B next year, and then grow at a neat 24.6% CAGR to $52B by 2035.

The numbers themselves are less interesting than their perfect smoothness, real markets don’t grow like spreadsheets. Still, one thing is hard to dispute: AI is now a core part of software development, embedded in IDEs, pipelines, and workflows.

That raises the real question, not market size, but salaries. AI boosts individual productivity, especially for routine tasks, which puts pressure on junior roles. Senior engineers, meanwhile, shift from β€œwriting code” to owning decisions, architecture, and AI output.

Forecasts may be fictional. The impact on the profession won’t be.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
1❀20πŸ‘5πŸ™2
This media is not supported in your browser
VIEW IN TELEGRAM
Manus has launched Meeting Minutes, a powerful new feature that records in-person conversations and transforms them into structured, actionable notes.

With speaker recognition, task assignment, and built-in workflows for creating deliverables like slides or social posts, Meeting Minutes bridges the gap between ideas and execution all in one seamless flow.

Available now for all Manus users.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀26πŸ€”7πŸ‘5
πŸ”₯ Google Research has released MedGemma 1.5

A major upgrade that boosts performance across CT, MRI, histopathology, X-ray timelines, and medical text

Delivering gains up to +14% accuracy in imaging tasks and +22% in EHR QA. Alongside it comes MedASR, an open medical speech-to-text model with up to 82% fewer transcription errors than general ASR

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀22πŸ‘3πŸ”₯3
This media is not supported in your browser
VIEW IN TELEGRAM
Free open source AI motion capture tool for everyone

Download and use for free, link below:

https://github.com/freemocap/freemocap

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ”₯34❀9πŸ‘8
Media is too big
VIEW IN TELEGRAM
This is the most important video for AI investors right now

Eric Schmidt: β€œAI’s natural limit is electricity, not chips”

The former $GOOGL CEO dropped the truth. Forget the chip shortage, power is bottleneck. Find out who controls the power… and monetize every last megawatt

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀21πŸ‘10πŸ”₯5πŸ€ͺ5
⚑️ OpenAI locks in massive AI compute via Cerebras

OpenAI has signed a three-year deal worth $10B+ with Cerebras, securing ~750 MW of cloud computing power to train and run next-gen AI models.

What’s being bought
β€’ Long-term access to Cerebras’ specialized AI compute at hyperscale
β€’ Capacity aimed at both training and inference, reducing reliance on traditional GPU clouds

Cerebras at scale
β€’ Already supplies compute to Meta and Mistral AI
β€’ Expanding data centers in Atlanta, Vancouver, and Montreal
β€’ Planning a Middle East facility, pending U.S. export approvals

Strategic backstory
β€’ OpenAI explored acquiring Cerebras in 2024
β€’ Sam Altman was already an investor at the time
β€’ Signals a shift from ownership ambitions to deep, capital-heavy partnerships

Capital & politics
β€’ $1.1B raised in Oct 2025 at an $8.1B valuation
β€’ Led by 1789 Capital, where Donald Trump Jr. is a partner

This isn’t just a supply deal, it’s OpenAI locking in sovereign-scale compute as a strategic moat, betting that control over power and silicon will matter as much as models themselves.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
❀11πŸ‘8
Media is too big
VIEW IN TELEGRAM
Not all robots walk on two legs.

Aletta is a robot that makes drawing blood completely automated.

We live in the future

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀34πŸ‘€12πŸ”₯8πŸ‘2😨1
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“’ Google introduces Personal Intelligence

A new Gemini beta that securely connects Gmail, Photos, YouTube, and Search to deliver a more personalized experience.

- Suggest decor using your Google Photos
- Plan trips by reasoning over flight and hotel emails
- Offer motivation based on your recent activities

Privacy first
- Off by default
- You choose which apps to connect
- Can be turned off anytime

Rolling out today in the US for Google AI Pro and Ultra subscribers. More regions coming soon.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ”₯14😨11❀8πŸ‘7πŸ€”3
This media is not supported in your browser
VIEW IN TELEGRAM
Good news and bad news.

1. Good news for folks in Southeast Asia who will start making big bucks online.

2. Bad news for Western Instagram influencers and OnlyFans girls: you'll need a real job soon.

Are you entertained?

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
😁54πŸ₯΄11😐9❀7πŸ‘7
πŸ”₯ GPT-5.3 is coming. Codename: Garlic

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ€”15❀9🫑5πŸ₯΄2πŸ‘1
❗️Grok AI bans β€œundressing” real people in some countries

X has introduced strict restrictions for Grok AI. The model will no longer be able to edit images of real people or depict them in revealing clothing, including bikinis. The ban applies to all users, including paid subscribers.

The decision was announced just hours after California’s attorney general said an investigation had been launched into the spread of sexualised AI deepfakes, including images of children created using this model.

Previously, Elon Musk had defended X, claiming critics β€œwant to suppress free speech”, and had himself posted AI-generated images of public figures including UK Prime Minister Keir Starmer in a swimsuit.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀18😐11😁6🫑6πŸ‘3
Media is too big
VIEW IN TELEGRAM
πŸ—£Geoffrey Hinton says LLMs work like the human brain, not like normal computer software

They aren't built on explainable lines of code, but rather on trillions of connection strengths that learn from data. "It's largely a mystery". We don't know what the individual neurons are doing

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘€14πŸ‘8πŸ€”8❀5πŸ₯΄3
This media is not supported in your browser
VIEW IN TELEGRAM
Steve Jobs foresaw AI in 1985.

Makes you think, how far could we have gone if he stayed in the arena today.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘29πŸ”₯18πŸ₯΄8❀7πŸ€ͺ5
This media is not supported in your browser
VIEW IN TELEGRAM
Nvidia posted this:

β€œWhat is the next generation of AI? Physical AI.”

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀26πŸ‘Ύ11πŸ€”6πŸ‘4
Media is too big
VIEW IN TELEGRAM
πŸ‡―πŸ‡΅ Japan Self-Defense Forces Deploy Vision 60 Robot Dogs

In the 2026 airborne drills, the JSDF introduced robots into combat training for the first time. Four Vision 60 robot dogs performed reconnaissance, detecting danger zones to support paratroopers before ground and helicopter units advanced.

Developed by US-based Ghost Robotics, these robots are designed for reconnaissance and light transport in dynamic environments. They are among the unmanned systems procured by various militaries, including the US, South Korea, and Japan.

Source: JGSDF

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀16πŸ‘8πŸ‘Ύ2
This media is not supported in your browser
VIEW IN TELEGRAM
AI video is now good enough for longform content that looks completely real.

This YouTube video has 3.6M views. The elderly person giving life advice doesn't exist.

Not a single comment notices

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ€”27❀15πŸ‘9πŸ€ͺ7πŸ”₯3
This media is not supported in your browser
VIEW IN TELEGRAM
The future is having 20 different livestreams with 20 different fake AI identities you've created to appeal to 20 different niches of people to sell 1 single product all at the same time.

Buckle up!!

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ€ͺ25πŸ‘€8πŸ‘7❀6
This media is not supported in your browser
VIEW IN TELEGRAM
OpenAI just unveiled Open Responses, an open-source spec built on the original Responses API that allows developers to build multi-provider, interoperable LLM interfaces.

Designed to support real-world workflows, Open Responses makes it easy to integrate different LLMs without constantly rewriting backend logic.

@aipost 🏴
Please open Telegram to view this post
VIEW IN TELEGRAM
❀15πŸ‘4πŸ‘€2