Meta’s Push to Fully Automate Ad Creation with AI

Meta plans to fully automate the ad creation process using AI by the end of 2025, enabling businesses to generate complete advertisements — from concept to video, imagery, text, and targeting — with minimal human input. This initiative aligns with CEO Mark Zuckerberg's vision to redefine advertising and sustain Meta’s core revenue stream, which heavily relies on ads.

Currently, Meta offers AI tools that personalize and modify ads. The new tools aim to allow brands to input basic goals (e.g., budget and product), with AI generating the full ad and optimizing it for platforms like Facebook and Instagram. Real-time personalization based on location and context is also planned.

Link to full article

Morgan Stanley Uses AI to Tackled One of Coding’s Toughest Problems

Morgan Stanley built its own AI tool, DevGen.AI, to modernize legacy code—something existing tools struggle with. Powered by OpenAI’s models, it translates old code (like COBOL) into plain English specs, helping developers rewrite it in modern languages. So far, it has reviewed 9 million lines of code and saved 280,000 developer hours. While full code translation still needs human oversight, the tool excels at mapping legacy code for easier rewriting. It’s part of Morgan Stanley’s broader push to modernize systems and scale AI across the business.

Link to the full article

Join Google Cloud’s Startup School: Agentic AI

Starting June 10, 2025, Google Cloud is hosting a free, hands-on program designed for founders, developers, and aspiring AI enthusiasts. Whether you're new to AI or looking to deepen your expertise, this is your chance to learn directly from Google Cloud

What to Expect:

• Live, Interactive Classes: Two sessions per week, tailored for EMEA and APAC time zones.

• Practical Learning: Explore how to build and deploy AI agents using Google Cloud's latest tools, including Gemini.

• Real-World Applications: Discover how AI agents are transforming industries and how you can leverage them in your

Why Attend?

• Gain hands-on experience with cutting-edge AI technologies.

• Learn from industry-leading experts at Google Cloud.

• Network with like-minded individuals and startups.

• Access resources to kickstart your AI journey.

How to Register:

Visit the Startup School: Agentic AI page.

The cost of Training AI

Training large language models (LLMs) has become increasingly expensive, with costs escalating dramatically over recent years. According to the Stanford 2025 AI Index Report, the financial investment required to develop cutting-edge AI models has surged, primarily due to the immense computational resources, extensive datasets, and specialized talent needed.

📊 Escalating Training Costs

The progression of training expenses for notable AI models is as follows:

• Original Transformer (2017): Approximately $930

• GPT-3 (2020): Around $4.3 million

• GPT-4 (2023): Estimated at $78 million

• Llama 3.1-405B (2024): Approximately $170 million

• Gemini 1.0 Ultra (2024): Estimated at $191

These figures underscore the exponential growth in training costs, with newer models requiring significantly more investment than their predecessors

Link to Stanford 2025 AI Index Report

Cursor’s Anysphere Reaches $500 Million Revenue, Raises $900 Million at $9.9 Billion Valuation

Anysphere, creator of AI coding assistant Cursor, raised $900M at a $9.9B valuation, led by Thrive Capital with Andreessen Horowitz, Accel, and DST Global. This is its third funding round in under a year, following a $100M raise at a $2.5B valuation in late 2024.

Cursor, built on a customized Visual Studio Code, enables natural language coding, debugging, and command generation. It now has 7M+ monthly active users, 40K+ paying teams (including OpenAI and Stripe), and generates nearly 1B lines of code daily.

Anysphere’s ARR has surpassed $500M, up from $300M in April 2025, with enterprise licenses now driving growth. Gross margins are ~74% and projected to reach 85% by 2027 as the company shifts to its own AI models.

The company rejected acquisition offers from OpenAI (which later acquired rival Windsurf), aiming to stay independent and scale further. Analysts estimate its ARR could hit $8.6B by 2029, positioning it as a leader in the fast-growing AI developer tools market.

Link to details