🇺🇸 The death of the traditional search, thank AI

The era of keyword searches and blue links is rapidly declining. Artificial Intelligence (AI) is no longer merely retrieving information; it is actively generating answers in real time. Companies like xAI, Google, and Perplexity are constructing a world where users don’t have to browse; instead, they simply ask questions. AI skillfully combines live data, eliminating the need for clicks, skipping websites, and fundamentally altering how humans access information. Publishers are in a state of panic as their traffic dwindles, and a new “zero-click” reality emerges.

Personalized and predictive summaries are gradually replacing the pursuit of truth. What once facilitated global connectivity through discovery is now condensing it into curated and conversational responses. The search engine you were familiar with is becoming obsolete. AI didn’t merely disrupt; it has completely replaced it.

Source: MIT Technology Review, Medium

@aipost 🪙 | Our X 🥇

Peter Thiel warns that AI combined with weapons technology is dangerous, even without uncontrollable Superintelligence.

In the Russia-Ukraine conflict, when human control of drones is jammed, adding AI to make them autonomous becomes the natural solution. "Even i, as a pro-tech person, find that somewhat unsettling"

@aipost 🪙 | Our X 🥇

People now want AI to sound more robotic 😐

@aipost 🪙 | Our X 🥇

Microsoft built a tiny AI that runs on a basic computer — And it’s shockingly smart

Microsoft has developed BitNet b1.58, an AI model that can run on standard CPUs found in everyday laptops, requiring no advanced hardware like supercomputers or graphics cards. This model simplifies computations by using only three values: -1, 0, and 1, allowing it to efficiently perform tasks that typically require more complex processing.

Key features of BitNet include:

- It needs just 0.4GB of memory to operate, less than many popular apps.
- It consumes 85–96% less energy compared to traditional AIs.
- It can read at a human-like speed (5–7 tokens per second) on basic hardware.
- Trained on 4 trillion tokens, it achieves performance close to larger, more complex models.

BitNet represents a potential breakthrough in making powerful AI accessible and affordable for everyday users, although Microsoft acknowledges that they do not fully understand the reasons behind its effectiveness. This raises both excitement and concern about the implications of such technology.

@aipost 🪙 | Our X 🥇

Elon Musk shared that an improved X algorithm powered by xAI's Grok AI will be rolling out soon

This came in response to Paul Graham complaining about X feed "drowning" in posts from either left or right-wing trolls

@aipost 🪙 | Our X 🥇

⏺ Musk’s xAI Holdings is reportedly raising the second-largest private funding round ever

Elon Musk's xAI Holdings is reportedly seeking to raise $20 billion in a funding round that could value the company at over $120 billion. If successful, it would be the second-largest startup funding round ever, following OpenAI's recent $40 billion raise.

This funding aims to help reduce xAI's significant debt, costing the company $200 million monthly in servicing fees. The round may attract familiar backers from Musk's other ventures, reflecting ongoing investor interest in AI.

@aipost 🪙 | Our X 🥇

Elon Musk announced: "Grok 3.5 will launch in early beta next week for SuperGrok subscribers"

Elon believes it is the first AI capable of accurately answering technical questions about rocket engines and electrochemistry fields.

@aipost 🪙 | Our X 🥇

⏺ Chatbot goes rogue… invents company policy, gaslights, sparks internet meltdown

Sam is an AI support bot for Cursor, an AI-powered code editor. Trained on various resources, Sam confidently responded to a user’s support ticket about being logged out when switching machines by claiming it was due to a new policy. However, this policy did not exist; Sam had fabricated it.

This misleading response quickly sparked outrage on platforms like Reddit and Hacker News, leading to user cancellations and frustration among developers. Cursor's team had to clarify that no such policy was in place, highlighting the risks of relying on AI for critical decisions. The incident serves as a cautionary tale about trusting chatbots like Sam with policy matters.

@aipost 🪙 | Our X 🥇

CZ — Binance crypto exchange CEO says that 99,99% of AI tokens are useless 🔸

Simultaneously, it advocates for the creation of practical encrypted AI agents

@aipost 🪙 | Our X 🥇

After mistakenly making GPT-4o "sycophant-y" and "annoying," OpenAI reversed the update, fixing the issue

CEO Sam Altman said the company is also working on "additional fixes" for the model's personality and will share more in the coming days

@aipost 🪙 | Our X 🥇

Google just announced the expansion of NotebookLM's podcast-generating Audio Overviews feature

It now supports over 50 languages, including Spanish and Hindi, for easy creation of multilingual content

@aipost 🪙 | Our X 🥇

OpenAI launched a shopping feature within ChatGPT.

@aipost 🪙 | Our X 🥇

⏺ AI + bandage predicts wound healing

Researchers at Caltech have developed a smart bandage called iCares that can predict if a wound is worsening before symptoms appear. This innovative bandage acts like a mini lab on the skin, using sensors to monitor factors like pH, temperature, and inflammatory chemicals. In tests with diabetic mice, it detected infections early, and in trials with 20 people with chronic wounds, it accurately predicted healing outcomes when combined with AI.

Chronic wounds, which affect millions, particularly older adults and diabetics, can lead to severe complications. iCares has the potential to improve treatment speed, reduce unnecessary doctor visits, and enhance the precision of wound care. This represents a significant advancement in combining science with skincare.

@aipost 🪙 | Our X 🥇

Sam Altman on X.

@aipost 🪙 | Our X 🥇

⏺ China’s plan: Let AI run healthcare

At this year’s China Medical Development Conference, officials made it abundantly clear that Artificial Intelligence (AI) is not merely a tool; it is poised to revolutionize the medical field, becoming the future of doctors, researchers, and hospital CEOs. The president of the Chinese Academy of Medical Sciences unveiled a comprehensive plan to gradually transfer more control of the medical system to algorithms, starting with pilot projects and progressing to full AI integration in the future.

China is already making significant strides in the medical field, boasting early Alzheimer’s detection, gene therapy for deafness, and other groundbreaking achievements. However, it is essential to emphasize that these successes hinge on the availability of clean data, adherence to strict regulations, and the utmost commitment to avoiding any potential pitfalls.

@aipost 🪙 | Our X 🥇

⏺ AI’s mega Brain plants could cost $200B and suck power like a small nation

A recent study suggests that the most advanced machines of the future might require an immense amount of energy, comparable to that of a nuclear reactor, to remain operational. By 2030, the most powerful AI data center could potentially contain 2 million chips, have a staggering cost of $200 billion, and consume an astonishing 9 gigawatts of power enough to potentially overwhelm and cripple the local power grid.

Georgetown University, the RAND Corporation, and Epoch AI conducted an investigation and discovered that both hardware costs and power requirements are experiencing a rapid and exponential growth, doubling annually. Even the most efficient AI system, known as the xAI Colossus, draws an energy consumption equivalent to that of 250,000 homes. If the current trajectory of AI development continues unabated, it is imperative to anticipate the potential consequences, including increased electricity bills, intensified carbon emissions, and substantial tax incentives that could have detrimental impacts on the environment.

@aipost 🪙 | Our X 🥇

Phi-4 14b reasoning released.

hard to keep track of all the releases.

Microsoft's Phi-4 Reasoning is a 14B parameter model fine-tuned for complex tasks in math, science, and coding. Despite its size, it rivals larger models like DeepSeek-R1 and OpenAI's o3-mini on benchmarks such as AIME 2025 and OmniMath.

live on HuggingFace

@aipost 🪙 | Our X 🥇

Alibaba’s Qwen just unveiled Qwen3: a family of eight open models ranging from 600M to 235B params.

— Flagship version rivals OpenAI o1 & DeepSeek-R1
— Hybrid “thinking” mode in all models
— Boosted coding + agent performance
— Supports 119 languages

@aipost 🪙 | Our X 🥇

Qwen 3 is quietly making serious moves

The 32b open-source model scored:

> 77 in reasoning
> 64 in coding
> 75 in maths

It outperforms gpt-4.5 preview, claude 3.7 sonnet (non-thinking), and gemini 2.5 flash

Yes, open-source isn’t just catching up, it’s competing

@aipost 🪙 | Our X 🥇