Introduction

Welcome to the second edition of this fully AI-generated newsletter. Every story, summary, and insight you’re about to read has been chosen, compiled, written, and structured by AI.

This is an experiment. The goal? To push the limits of AI in content generation, refine the process, and see just how far we can take automation in curating and delivering the most relevant AI news.

Along the way, I’ll be tweaking, testing, and iterating - so expect updates, improvements, and perhaps a few surprises. Consider this a journey into the evolving role of AI in media. Let’s see how well it does and thanks for joining us!

24/03/25 Foreword

“Hi everyone, thank you for subscribing to our newsletter and we look forward to you being part of the journey. Please do share your thoughts and feedback as we progress.

This week there are no tweaks or updates, this is because we want to give it a week or two more to see how it is performing out of the gate, and to hear your thoughts”. - Very best, Nicolas and the System7 team.

Now, onto the latest in AI.

As AI-generated content continues its rapid expansion, we're witnessing transformative changes across industries. This week brings several groundbreaking developments that highlight just how quickly the AI landscape is evolving. From Nvidia's powerful new chips and generalist robotics announcements to Baidu's surprising model that reportedly outperforms GPT-4.5 at a fraction of the cost, we're seeing barriers to advanced AI fall faster than many predicted. If you've been wondering how to keep up with the pace of change, this week's newsletter provides the essential updates you need.

Major AI News Headlines

• Nvidia unveils next-gen AI infrastructure and robotics foundation models.

• Baidu claims its ERNIE 4.5 outperforms GPT-4.5 at just 1% of the cost.

• AI capabilities doubling every 7 months, potentially handling week-long tasks soon.

Major AI News

Nvidia Transforms GTC Into AI's 'Super Bowl' with Major Announcements

Summary

At GTC 2025, CEO Jensen Huang unveiled significant advancements including the Blackwell Ultra chip and a roadmap for even more powerful future processors. The company introduced GROOT N1, an open-source foundation model for humanoid robots that uses a dual-system approach to handle both quick intuitive actions and thoughtful planning.

Why It Matters

These announcements solidify Nvidia's leadership in AI infrastructure while expanding into robotics, a field that has long struggled to deliver on its promise. The open-source approach to GROOT N1 could accelerate development across the industry, potentially bringing capable humanoid robots into commercial settings much sooner than expected.

Key Takeaway

Nvidia's roadmap includes Vera Rubin (2026) at 50 petaflops and Rubin Ultra (2027) at 100 petaflops, DGX Spark and DGX Station bring powerful AI to compact devices for developers, GROOT N1 is already being adopted by Boston Dynamics and Agility Robotics.

Source: Link

Baidu Challenges AI Giants with Ultra-Affordable ERNIE 4.5

Summary

Baidu has introduced ERNIE 4.5, a multimodal large language model that reportedly outperforms GPT-4.5 across multiple benchmarks while costing only 1% as much. Alongside it, Baidu unveiled X1, a reasoning model designed to rival DeepSeek's R1 at half the price, with plans to release both through a free chatbot on April 1.

Why It Matters

This development signals a dramatic cost reduction in sophisticated AI capabilities, potentially democratizing access to advanced AI tools. If Baidu's claims hold up, we could see a significant shift in the competitive landscape as premium AI features become more widely available at drastically lower price points.

Key Takeaway

ERNIE 4.5 will be open-sourced by June 30, 2025, Baidu plans to make its Ernie Bot free on April 1, This move puts pressure on Western AI companies to reduce prices.

Source: Link

AI Tasks Doubling in Complexity Every 7 Months

Summary

A study by research group METR indicates that the tasks large language models can tackle are doubling in duration every seven months, outpacing Moore's Law. Today's models like Claude 3.7 Sonnet can handle tasks that take humans about an hour, compared to GPT-4's five-minute capability from just a year earlier.

Why It Matters

This acceleration suggests we're approaching an inflection point where AI systems could soon manage complex projects lasting days or weeks. If this trend continues, by 2029 AI could potentially handle month-long projects like starting a business or making scientific discoveries, fundamentally changing how we approach work and innovation.

Key Takeaway

Current AI models handle hour-long tasks, but could manage four-hour tasks within a year, This growth rate is significantly faster than hardware improvements under Moore's Law, The findings should be viewed cautiously as they're based on only a few years of data.

Source: Link

Fun AI Topics Of The Week!

Italian Newspaper Goes Fully AI-Generated

Summary

Il Foglio published what it claims is the world's first entirely AI-generated newspaper edition. Every element from headlines to summaries was produced by artificial intelligence, with journalists limited to asking questions and reviewing AI outputs.

Why It’s Interesting!

This experiment pushes boundaries on AI's role in journalism and raises questions about the future of media creation. While the AI handled a range of topics from politics to economics, it notably lacked direct quotes from human sources, highlighting current limitations.

Key Takeaway

Media outlets worldwide are increasingly experimenting with AI-driven content creation, The absence of human quotes reveals an important gap in current AI capabilities, This test case provides valuable insights into both the strengths and limitations of AI journalism.

Source: Link

Study Shows AI Can Make Better Memes Than Humans

Summary

Research titled "One Does Not Simply Meme Alone" revealed that AI-generated memes, on average, outperformed human-made ones in humor, creativity, and shareability. However, the very best memes still came from humans, and the most creative and shareable content emerged from human-AI collaboration.

Why It’s Interesting!

Memes represent a uniquely human form of cultural expression that blends humor, context, and shared references. AI's ability to succeed in this space demonstrates its growing understanding of cultural nuance, while the superior results from collaboration highlight the value of human-AI teamwork.

Key Takeaway

AI excels at broadly appealing humor because it learns from massive datasets reflecting mainstream tastes, Human-AI collaboration produced the most creative and shareable content overall, AI assistance helped people generate more ideas with less effort.

Source: Link

Weather Forecasting Gets AI Makeover with 'Aardvark'

Summary

Cambridge, the Alan Turing Institute, and Microsoft jointly developed Aardvark Weather, an AI-based forecasting system that delivers faster, more precise predictions. Using just a fraction of the computing power of traditional methods, it generates specialized forecasts in minutes and aims to extend accurate predictions up to eight days.

Why It’s Interesting!

Weather forecasting has always been a computational challenge that affects everyone from farmers to festival planners. This AI approach makes high-quality forecasting more accessible and could democratize precise weather predictions globally, potentially helping communities better prepare for extreme weather events.

Key Takeaway

Aardvark uses only 10% of the data volume required by traditional forecasting models, It reportedly outperforms the U.S. Global Forecast System (GFS), The technology could make ultra-local predictions more widely available.

Source: Link

AI Tools Of The Moment

Memex 

Converts chat into working software on any tech stack with 50% improved efficiency.

Link

Pika 2.2

Create smooth transition videos between images with simple prompts.

Link

Lindy

Automates entire workflows from phone calls to sales outreach without coding.

Link

Claude with Search

Claude finally adds real-time web search with citation support (US only currently).

Link

Mistral Small 3.1

Compact open-source model that runs on a single GPU with multimodal capabilities.

Link

Expert Prompt Of The Week

CONTEXT

With AI agents becoming more capable of complex tasks as shown in the METR study, here's a prompt to help you delegate effectively.

PROMPT

I need to delegate a [SPECIFIC TASK] to an AI agent. Please help me break this down into: 1) The key steps required, 2) The specific instructions I should provide for each step, 3) Points where human verification would be valuable, 4) Potential failure modes to watch for, and 5) How I should evaluate the final output to ensure quality. Focus on making the instructions clear enough that the AI can work autonomously while maintaining appropriate guardrails.

EXAMPLE USE CASE

Use this prompt when planning to hand off research, content creation, or data analysis tasks to AI. For instance, if you need a market research report, this prompt helps you structure the assignment with the right balance of autonomy and oversight, ensuring you get reliable results while saving significant time.

Trending AI Topics For The Week

Y Combinator startups embrace 'vibe coding' for rapid development

Summary

Y Combinator's latest demo day featured startups where roughly 80% rely on LLMs, with many using 'vibe coding' - writing software through natural language interactions with AI - to achieve million-dollar revenues with minimal team sizes.

Why It’s Important

This trend demonstrates how AI is dramatically lowering barriers to entry for software development, allowing founders to focus on solving problems rather than writing complex code. It could accelerate innovation across industries as small teams can now build and deploy sophisticated applications faster than ever.

Adobe showcases AI agent experiments for creative workflows

Summary

Adobe demonstrated seven futuristic agent experiments, including Project Slide Wow for data-to-presentation conversions and Project Get Savvy for automated ad campaign creation.

Why It’s Important

These tools signal how AI will transform creative industries by automating technical aspects while enhancing human creativity. As major creative software providers integrate AI agents, we'll likely see significant productivity gains for designers, marketers, and content creators.

DeepMind CEO estimates AGI is 5-10 years away

Summary

Demis Hassabis, CEO of Google DeepMind, predicts artificial general intelligence will arrive within 5-10 years, while other tech leaders offer timelines ranging from 2-3 years to more than a decade.

Why It’s Important

This forecast from one of AI's foremost researchers suggests that truly transformative AI capabilities may arrive sooner than many expect. The consensus among industry leaders that AGI is on the horizon within a decade has profound implications for workforce planning, policy development, and investment strategies.