25 Aug 2025

First AI Bank; AI for Aging?; 1 Command = Full Stack App

💸 AI in Finance: The Bank Just Got Deleted

The Scoop: Malaysia is now home to the world's first fully AI-powered bank, Ryt Bank, a joint venture between YTL Group and Sea Limited. Rather than another hyped up banking app, it's a ground-up reinvention of financial services built on the country's native large language model (ILMU).

Why It Matters: Ryt Bank represents a true threat to traditional banking infrastructure. By offering 24/7 multilingual AI assistants, instant credit lines and a 4% daily interest rate with no fees, it makes legacy banks with their physical branches and paperwork-heavy processes look horribly dated. For consumers, especially younger generations, the experience is seamless and conversational - more like TikTok than a trip to the bank.

The Bottom Line: As AI Secret aptly puts it, "Ryt Bank isn’t adding AI to banking - it’s deleting the bank from banking." This move is a clear signal that the financial sector is on the verge of sigificant AI-driven disruption. Expect to see similar models emerge globally, putting immense pressure on established players to innovate or become obsolete.

Read more: AI Secret

🤝 Frenemies & Industry Shifts; Siri's Last Hope?

Apple Turns to Google for a Siri Makeover: In a move that could reshape the voice assistant landscape, Apple is in early discussions with Google to license its Gemini AI for a complete overhaul of Siri. This comes after setbacks reportedly pushed Siri's major internal upgrade to 2026.

The Details: Google is already training a custom Gemini model to run on Apple's private servers. Apple is simultaneously testing its own models (Project Linwood) against external options (Project Glenwood), with Anthropic and OpenAI also in the running. A decision is expected in the coming weeks.
Why It Matters: This signals a major strategy shift for Apple, which has historically favored in-house development. By potentially partnering with a direct competitor, Apple is acknowledging the immense challenge of building a frontier AI model and prioritizing user experience over its walled-garden philosophy. For consumers, a Gemini-powered Siri could be a tremendous upgrade.

Meta and Midjourney Team Up for 'Aesthetic' AI: Meta is partnering with Midjourney to integrate the startup's renowned 'aesthetic technology' into its future AI models and products. This collaboration will combine teams to upgrade the visual capabilities across Meta's entire ecosystem.

The Details: While Meta has its own image and video generation tools (Imagine, Movie Gen), this partnership gives it access to Midjourney's unique and highly-regarded artistic style. Midjourney, for its part, remains an independent research lab with no investors.
Why It Matters: This is another example of a tech giant looking externally to bolster its AI capabilities. By integrating Midjourney's aesthetics, Meta can significantly improve the quality of its visual generation tools, making them more appealing to its billions of users.

Read more: The Rundown AI

🔒 Safety & Security: The AI Browser Is a Minefield

The Scoop: The AI browser company Brave has exposed a critical vulnerability in Perplexity's Comet browser, dubbed the "Lethal Trifecta." This flaw allows malicious actors to embed hidden commands in web content, which the AI can then execute to steal sensitive user data, including emails and passwords.

How It Works: The vulnerability arises from a fundamental aspect of how LLMs operate: they can't distinguish between user commands and instructions hidden in the content they process. A simple command like "Summarize this page" can trigger a hidden script that hijacks your accounts. The attack involves the AI having access to untrusted data (websites), private data (your accounts), and the ability to communicate externally.
Why It Matters: This isn't just a bug; it's a core design flaw in the current generation of AI browsers. As The Neuron points out, every AI browser with these capabilities is susceptible to this type of attack. This raises serious questions about the safety of using AI to interact with the web and access personal information. While Perplexity has patched this specific exploit, the underlying problem remains a significant challenge for the industry.
The Bottom Line: We are in the early days of agentic AI, and security is lagging behind capabilities. Until robust safety measures are in place, such as isolating AI browsing sessions and requiring user confirmation for sensitive actions, it's wise to be cautious about the data you expose to AI browsers.

Read more: The Neuron

🔬 Breakthroughs: Could AI Reverse Aging?

The Scoop: In a stunning collaboration, OpenAI and Retro Biosciences have used a custom AI model to redesign proteins that can turn aging cells into stem cells. The AI-designed proteins are 50 times more efficient than the original Nobel Prize-winning methods discovered in 2012.

How It Works: Researchers trained a specialized AI, GPT-4b micro, on biological data rather than internet text. This allowed the model to understand the complex language of proteins and design new versions of the "Yamanaka factors" that are far more effective at cellular reprogramming. The results, which have been validated by multiple labs, show a dramatic improvement in DNA repair and a reversal of key aging signatures at the cellular level.
Why It Matters: This breakthrough demonstrates the incredible potential of custom AI models trained on domain-specific data. While general-purpose AI is powerful, specialized models can unlock discoveries that would be impossible with a one-size-fits-all approach. This opens up new frontiers in biology, chemistry, and materials science, turning them into computational playgrounds where decades of research can be compressed into weeks.

Read more: The Rundown AI

🎨 AI Passes Creativity Boundary (And Then Some!)

AI Aces "Aesthetic Turing Test": We have essentially reached the point where AI can generate images and music that are indistinguishable from human-created art. From Vogue covers to viral music tracks, the line between authentic and artificial has blurred, effectively passing the "Aesthetic Turing Test."

The Impact: This is forcing a massive shift in the creative industries. Workflows in fashion, music and media are being compressed into prompts and filters. The value proposition of agencies that once sold "originality" and "authenticity" is rapidly eroding.
The Bottom Line: As AI Secret notes, "When AI nails beauty better than humans, the only scarcity left is trust." Human creatives are now forced to either lean into imperfection as a differentiator or risk being outcompeted by infinitely tweakable, zero-cost AI.

Google AI's Restaurant Hallucinations: On the less glamorous side of creative AI, Google is facing backlash for its AI chatbot confidently inventing fake "daily specials" and menu items for restaurants. This is causing chaos for small businesses, who are left to deal with angry customers demanding non-existent discounts.

The Impact: Google is effectively outsourcing the cost of its AI hallucinations to small businesses that lack the resources to fight back. This erodes trust in both Google and the affected restaurants, turning a supposedly helpful feature into a liability.

Read more: AI Secret

🇨🇳 Chinese AI Continues Apace

A recent ranking of 19 Chinese AI labs highlights the country's growing contributions to AI research. The evaluation focuses on research output rather than the raw capabilities of the models themselves.

Top Performers: Deepseek remains a top-tier lab, despite a recent quiet period. Newer companies like Moonshot AI and Zhipu are quickly closing the gap with their recent releases of the K2 and GLM-4.5 models, respectively.
Why It Matters: This ranking provides valuable insight into the competitive landscape of AI development in China, showcasing the labs that are making the most significant contributions to the field.

Read more: TLDR AI

🛠️ New AI Tools & Launches

Informed: A personalized news app that reads you daily briefings in any voice, including clones of your own voice or celebrities like Trump or Taylor Swift. (Mac only for now)
VibeFlow: Turns your prompts into full-stack web applications with visual n8n-style workflows that you can see and edit.
Zen Browser: A privacy-focused browser featuring Workspaces, Split View, and Zen Mods for a calmer web experience, complete with an AI chatbot on the side.
Willow Voice: Allows you to dictate emails, Slack messages, and any text across your computer 3x faster than typing, automatically formatting and adapting to your writing style.
Pixxel: A Chrome extension that instantly styles browser screenshots with custom backgrounds, shadows, and rounded corners.
ChatGPT Marketing: A service designed to improve your ranking on ChatGPT by aligning and optimizing your existing marketing content.

Read more: AI Secret, The Neuron

👀 What to Watch For This Week & Beyond

Apple's Siri Decision: The outcome of Apple's partnership discussions could reshape the voice assistant market and set a precedent for how tech giants approach AI development.
AI Browser Security: How the industry responds to the "Lethal Trifecta" vulnerability will determine the future safety and adoption of AI browsers.
Regulatory Response: Governments worldwide are likely to increase their focus on AI safety and security, potentially leading to new regulations.
Creative Industry Adaptation: How traditional creative industries adapt to AI that can match human aesthetic output will be a key story to follow.
AI Banking Expansion: Watch for similar AI-first banking models to emerge in other markets, potentially disrupting traditional financial institutions globally.

Today's sources: